docs: 记录 RM75 QP 与 UDP 修复计划

AGENTS.md

@@ -205,6 +205,7 @@
 * 类型标注方式。
 * 注释风格。
 * 测试组织方式。
+* 新生成的 Markdown（`.md`）文档统一使用中文撰写。
 
 如果项目已有 lint、format 或 test 命令，优先使用项目已有命令，不要擅自更换工具链。
 
@@ -227,7 +228,9 @@
 
 ## Git 与提交
 
-除非用户明确要求，否则不要自动提交、推送、创建分支或修改远程仓库。
+除非用户明确要求，否则不要自动提交、推送或修改远程仓库。
+
+使用 Superpowers 执行计划时，允许 subagent 按相关 skill 创建和使用独立 worktree 及其配套本地分支；其他情况下，除非用户明确要求，不要自动创建分支。
 
 如果用户要求生成提交信息，提交信息应：
 

README.md

@@ -7,11 +7,12 @@ PICO/XR 双手柄 UDP JSON
   -> xr_rm_input/udp_controller_receiver
   -> /xr/left_controller 与 /xr/right_controller
   -> xr_rm_teleop/single_arm_velocity_teleop
-  -> 左右 RM75 笛卡尔相对位姿透传控制
+  -> Placo QP 单步逆解
+  -> 左右 RM75 七关节角透传控制
   -> /xr_rm/<arm_name>/current_pose、raw_target_pose、target_pose、cmd_vel、target_clamped 调试话题
 ```
 
-当前控制方式是“手柄相对位姿透传”遥操作：按住 `grip` 时锁定当前手柄位姿和机械臂 TCP 位姿，之后根据手柄相对位移和相对旋转生成目标 TCP，经过工作空间限幅、目标低通、姿态低通和单帧步长限制后，通过 `rm_movep_canfd` 下发目标位姿。松开 `grip`、UDP 超时或节点退出时会请求机械臂慢停。
+当前控制方式是“手柄相对位姿 + 单步 QP”遥操作：按住 `grip` 时锁定当前手柄位姿和机械臂 TCP 位姿，之后根据手柄相对位移和相对旋转生成目标 TCP。姿态目标使用旋转矩阵和 SO(3) 最短路径完成死区、滤波与限速，不经过 RPY。每个控制周期执行一次 Placo QP，并通过 `rm_movej_canfd` 下发 7 个关节目标。松开 `grip`、UDP 或关节反馈超时、节点退出时会请求机械臂慢停。
 
 ## 当前范围
 
@@ -19,7 +20,8 @@ PICO/XR 双手柄 UDP JSON
 
 - PICO/XR 手柄 UDP 数据接收，并分发到左右手柄 ROS2 话题。
 - 通过统一的 `arm_debug.launch.py` 支持左臂、右臂、双臂的 mock 调试和真机调试。
-- RM75 真机连接适配，包含 `rm_movep_canfd` 位姿透传、安全速度/加速度配置、可选初始化点位移动。
+- RM75 真机连接适配，包含关节反馈缓存、`rm_movej_canfd` 关节透传、安全速度/加速度配置、可选初始化点位移动。
+- Placo 0.9.4 RM75 QP 逆解；收到首帧有效关节反馈后才启用，求解失败时保留上一组有效关节目标。
 - 真机模式下，点击对应手柄 `trigger` 可切换并保持对应夹爪开/关状态。
 - Tkinter 启动面板 `launcher_ui.py`，用于现场快速启动、监控 topic、检查环境和清理进程。
 - 自定义 PICO 4 Ultra UDP Sender Unity 工程，负责发送左右手柄 pose、`grip`、`trigger` 和 pose 诊断字段。
@@ -67,7 +69,11 @@ src/
 │   └── msg/
 │       └── XrController.msg        # hand/grip/trigger/pose
 └── xr_rm_teleop/
+    ├── models/
+    │   ├── rm75/                   # 旧 RM75 模型资源（launch 不再选用）
+    │   └── rm75_omnipicker/        # RM75 + OmniPicker fixed URDF 与网格
     └── xr_rm_teleop/
+        ├── placo_ik_solver.py      # Placo 0.9.4 单步 QP 逆解
         ├── single_arm_velocity_teleop.py
         ├── realman_adapter.py
         └── fun_peripheral.py
@@ -90,6 +96,17 @@ source install/setup.bash
 
 真机模式还需要安装睿尔曼 Python API2。若未安装，mock 模式仍可正常使用；真机启动时会提示缺少 `Robotic_Arm` 包。
 
+遥操作节点固定由 `/home/robot/miniconda3/envs/xr/bin/python` 启动，并复用其中的 Python 3.10、Placo 0.9.4、Pinocchio 3.7.0 和 NumPy 2.2.6。`ros2`、`colcon` 和 `udp_controller_receiver` 仍使用系统 Python。禁止通过 `pip --user`、`sudo pip` 或系统安装升级 Placo、Pinocchio、EigenPy 和 NumPy。
+
+只读检查 Placo 版本：
+
+```bash
+/home/robot/miniconda3/envs/xr/bin/python -c \
+  "import importlib.metadata; print(importlib.metadata.version('placo'))"
+```
+
+输出必须为 `0.9.4`。
+
 如果希望 `launcher_ui.py` 从任意目录找到工作空间，可以设置：
 
 ```bash
@@ -150,17 +167,25 @@ sudo update-alternatives --config x-terminal-emulator
 
 打开 `launcher_ui.py`，点击 `Check Env`。如果 `install/setup.bash` 缺失，先回工作空间根目录重新执行 `colcon build --symlink-install`。
 
-第二步：跑 mock 闭环。
+第二步：分别跑左、右臂 mock 闭环。
 
-在 `Simulation` 模式运行 `One-Click Dual Mock Demo`，或分开运行：
+分两个终端依次验证左臂：
 
 ```bash
-ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=true
-ros2 run xr_rm_input sample_udp_sender --hand both --host 127.0.0.1 --port 15000 \
-  --pattern axis_sweep --seconds 60 --both-mode staggered
+ros2 launch xr_rm_bringup arm_debug.launch.py arm:=left use_mock:=true
+ros2 run xr_rm_input sample_udp_sender --hand left --host 127.0.0.1 --port 15000 \
+  --pattern axis_sweep --seconds 30
 ```
 
-`sample_udp_sender` 默认使用 `axis_sweep` 成对扫轴轨迹，并在终端打印 `XR +X/-X/+Y/-Y/+Z/-Z` 标签。`--hand both --both-mode staggered --seconds 60` 会先左后右，适合肉眼确认左右臂方向；如果只想左右同时动，可用 `--both-mode synchronized`。需要检查末端姿态时可增加 `--rotation-pattern rpy_steps --rotation-amplitude-deg 25`。
+停止左臂进程后，再分别验证右臂：
+
+```bash
+ros2 launch xr_rm_bringup arm_debug.launch.py arm:=right use_mock:=true
+ros2 run xr_rm_input sample_udp_sender --hand right --host 127.0.0.1 --port 15000 \
+  --pattern axis_sweep --seconds 30
+```
+
+`sample_udp_sender` 默认使用 `axis_sweep` 扫轴轨迹，并在终端打印 `XR +X/-X/+Y/-Y/+Z/-Z` 标签。需要检查末端姿态时可增加 `--rotation-pattern rpy_steps --rotation-amplitude-deg 25`。
 
 观察：
 
@@ -182,23 +207,18 @@ ros2 launch xr_rm_bringup arm_debug.launch.py arm:=left use_mock:=false
 ros2 launch xr_rm_bringup arm_debug.launch.py arm:=right use_mock:=false
 ```
 
-单臂真机默认执行配置文件中的 `movej(initial_joint_pose)`；现场需要跳过时，显式传入
-`move_to_initial_pose_on_connect:=false`。
+所有 YAML 默认都不会执行 `movej(initial_joint_pose)`。只有确认安全区清空后，才可在当前使用的
+`left_arm_rm75.yaml`、`right_arm_rm75.yaml` 或 `dual_arm_rm75.yaml` 中将
+`move_to_initial_pose_on_connect` 改为 `true`。
 
 第四步：双臂真机。
 
 ```bash
-ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false \
-  left_robot_ip:=192.168.192.18 \
-  right_robot_ip:=192.168.192.19
+ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false
 ```
 
-双臂默认不会自动移动到初始化点。以后需要启用时，在确认安全区清空后显式打开：
-
-```bash
-ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false \
-  move_to_initial_pose_on_connect:=true
-```
+双臂默认不会自动移动到初始化点；机器人地址、控制参数和初始化移动开关均从
+`dual_arm_rm75.yaml` 读取。
 
 ## Launch 入口说明
 
@@ -211,20 +231,9 @@ ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false \
 - `udp_host`：UDP 监听地址，默认 `0.0.0.0`。
 - `udp_port`：UDP 监听端口，默认 `15000`。
 - `udp_timer_hz`：UDP receiver 轮询频率，默认 `200.0`。
-- `left_robot_ip`：左臂 IP，默认 `192.168.192.18`。
-- `right_robot_ip`：右臂 IP，默认 `192.168.192.19`。
-- `robot_port`：RM75 TCP 端口，默认 `8080`。
-- `left_avoid_singularity` / `right_avoid_singularity`：左右臂避奇异参数，默认左 `0`、右 `1`。
-- `avoid_singularity`：非空时覆盖左右臂避奇异参数。
-- `frame_type`：`rm_movep_canfd` 坐标系类型，默认 `1`。
-- `control_rate_hz`：`rm_movep_canfd` 目标位姿发送频率，默认 `90.0`。
-- `follow`：传给 `rm_movep_canfd` 的跟随标志，默认 `false`。
-- `configure_safety_limits`：连接真机后是否配置速度/加速度安全参数，默认 `true`。
-- `enable_tool_control`：是否在遥操作节点内启用末端工具控制 topic，默认 `true`。
-- `enable_trigger_gripper_control`：是否允许用 `trigger` 点击切换对应夹爪状态，默认 `true`。
-- `trigger_close_threshold`：trigger 点击判定阈值，默认 `0.95`。
-- `configure_peripheral_on_connect`：遥操作节点连接真机后是否配置末端外设，默认 `true`；工具控制会复用同一个 RealMan 连接，避免两个进程同时抢占同一机械臂。
-- `move_to_initial_pose_on_connect`：连接后是否执行 `movej(initial_joint_pose)`；默认 `auto`，单臂配置启用、双臂配置禁用，也可显式传 `true`/`false` 覆盖。
+
+机器人 IP/端口、控制频率、CANFD、限速、工具和初始化位姿等行为参数只由对应 YAML
+配置，launch 不再提供同名覆盖项。
 
 ## 配置文件说明
 
@@ -235,7 +244,7 @@ ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false \
 
 `left_arm_rm75.yaml` 和 `right_arm_rm75.yaml` 用于 `arm_debug.launch.py arm:=left/right` 的单臂调试，因为单臂节点名是 `single_arm_velocity_teleop`。
 
-`xr_rm_bringup/config/peripherals_rm75.yaml` 保存末端工具坐标、负载和左右臂外设选择。当前配置为左臂 `scissorgripper=2`、右臂 `scissorgripper=1`，真机连接阶段会初始化外设，后续开合命令复用同一个 RealMan 连接。
+`xr_rm_bringup/config/peripherals_rm75.yaml` 保存真实控制器使用的末端工具坐标、负载和左右臂外设选择，文件内容保持原状。Placo 使用 `xr_rm_teleop/models/rm75_omnipicker` 中的一体化 fixed URDF，直接控制相对 `omnipicker_base_link` 沿 `+Z` 偏移 `0.16 m` 的 `omnipicker_tcp`，不再把外设 YAML 的工具位姿重复转换到 QP。真机连接阶段仍会初始化外设，关节反馈、关节指令、慢停和开合命令复用该单臂节点的同一个 RealMan 连接。
 
 重点控制参数：
 
@@ -245,14 +254,19 @@ ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false \
 - `target_filter_fast_threshold_m`：进入快速滤波区间的目标变化阈值。
 - `max_linear_speed`：目标位姿单帧步长限制对应的最大线速度。
 - `enable_orientation_control`：是否把手柄相对旋转映射到 TCP 姿态。
-- `orientation_filter_alpha` / `orientation_deadband_rad`：目标 TCP 姿态低通和死区。
-- `max_orientation_speed`：目标姿态单帧步长限制对应的最大角速度。
+- `orientation_filter_alpha` / `orientation_deadband_rad`：按 SO(3) 最短旋转角处理的目标 TCP 姿态滤波和死区。
+- `max_orientation_speed`：目标 TCP 姿态沿 SO(3) 最短路径的最大角速度，当前为 `0.5 rad/s`。
 - `workspace_min` / `workspace_max`：笛卡尔工作空间边界。
 - `cyl_radius_limit`：基座圆柱半径限制。
 - `xr_to_robot_matrix`：`/xr/*_controller` Project 位移到 RM75 base 坐标的映射矩阵。
-- `current_pose_poll_hz`：低频读取真机当前 TCP 的频率；控制中不再每帧阻塞读取状态。
-- `mock_initial_pose`：mock 模式初始 TCP 位姿。
-- `initial_joint_pose`：可选真机初始关节角。
+- `robot_ip` / `robot_port`：RM75 TCP 控制连接地址。
+- `realtime_push_host_ip`：连接机械臂 Wi-Fi 后本机实际 IPv4；可用
+  `ip -4 route get 192.168.192.19` 查看输出中的 `src`，当前为 `192.168.192.148`。
+- `realtime_push_port`：UDP 主动反馈端口；左臂 `8089`、右臂 `8090`，同机双臂不能重复。
+- `realtime_push_cycle_ms`：UDP 主动反馈周期，当前为厂商支持的 `5 ms`。
+- `follow` / `canfd_trajectory_mode`：`rm_movej_canfd` 的高跟随和轨迹模式参数。
+- 当前三份 YAML 默认均使用 `follow: false` 完成安全基线验证；确认关节加速度与反馈稳定后，再单独测试高跟随。
+- `initial_joint_pose`：mock 的初始关节反馈，以及显式开启初始化移动时的真机初始关节角。
 
 当前 `/xr/*_controller` 的 Project 坐标约定：
 
@@ -410,7 +424,7 @@ ros2 topic echo /xr/right_controller --field trigger
 
 1. 确认急停、网络、机械臂工作区和人员位置。
 2. `launcher_ui.py` 中先 `Ping Left RM75` 或 `Ping Right RM75`。
-3. 单臂启动，`move_to_initial_pose_on_connect:=false`。
+3. 确认对应 YAML 中 `move_to_initial_pose_on_connect: false` 后单臂启动。
 4. 手握急停，按住 `grip` 后只做小幅单轴移动。
 5. 逐个确认上/下、前/后、左/右方向。
 6. 小角度转动手柄，确认 `/xr_rm/<arm>/target_pose` 姿态和 `/xr_rm/<arm>/cmd_vel.twist.angular` 变化符合预期。

docs/superpowers/plans/2026-07-29-rm75-feedback-recovery.md

@@ -0,0 +1,964 @@
+# RM75 关节反馈与故障恢复实施计划
+
+> **执行要求：** 使用 `superpowers:executing-plans` 按任务逐项实施。只有用户明确授权 subagent 后，才允许使用 `superpowers:subagent-driven-development`、独立 worktree 或本地分支。所有步骤使用复选框跟踪。
+
+**目标：** 启动时用 `rm_get_joint_degree()` 初始化 RM75 QP；运行时以 UDP `joint_position` 作为实际反馈；短暂丢包时保持最后安全目标；持续丢包或 CANFD 错误时安全同步、停止或等待人工重新使能。
+
+**实现方式：** 继续复用现有唯一 `RealManAdapter` 连接，只增加一个同步读取关节角的方法。恢复决策仍放在 `SingleArmVelocityTeleop`，用少量布尔状态复用现有 slow-stop、Grip 重新使能和关节限速逻辑，不新增状态机类、线程、连接或依赖。
+
+**技术栈：** Python 3.10、ROS2 Humble `rclpy`、睿尔曼 Python API2、Placo、pytest、ament/colcon。
+
+**设计文档：** `docs/superpowers/specs/2026-07-29-rm75-feedback-recovery-design.md`
+
+**厂商接口依据：**
+
+- `rm_get_joint_degree() -> tuple[int, list[float]]`：<https://develop.realman-robotics.com/robot/apipython/classes/armState/>
+- `rm_movej_canfd(..., follow=False, ...)` 为低跟随：<https://develop.realman-robotics.com/robot/apipython/classes/movePlan/>
+
+## 仓库约束
+
+- 所有构建、测试和启动命令均在工作空间根目录 `/home/robot/WS_xr` 执行。
+- 所有 Git 命令均在仓库根目录 `/home/robot/WS_xr/src` 执行。
+- 每次 ROS2 构建、测试或启动前先执行 `source /opt/ros/humble/setup.bash`。
+- 不自动提交、推送、创建分支或 worktree；只有用户明确要求后才执行。
+- 验证只通过 `xr_rm_bringup/launch/arm_debug.launch.py use_mock:=true`。
+- 不连接真机，不发送真实 CANFD，不移动机械臂，不操作夹爪或末端外设。
+- 不修改依赖、锁文件、CI、格式化配置、公开入口和无关代码。
+
+## 文件范围
+
+- 修改 `src/xr_rm_teleop/xr_rm_teleop/realman_adapter.py`
+  - 增加同步关节查询。
+  - 复用 UDP 与同步查询的角度校验。
+- 修改 `src/xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`
+  - 启动同步、UDP保持/恢复、CANFD恢复和故障锁存。
+- 修改 `src/xr_rm_teleop/test/test_initial_joint_pose.py`
+  - 适配器同步查询测试。
+- 修改 `src/xr_rm_teleop/test/test_joint_control.py`
+  - 启动、超时、恢复和锁存测试。
+- 同步修改：
+  - `src/xr_rm_bringup/config/dual_arm_rm75.yaml`
+  - `src/xr_rm_bringup/config/left_arm_rm75.yaml`
+  - `src/xr_rm_bringup/config/right_arm_rm75.yaml`
+
+---
+
+## 任务一：给现有适配器增加同步关节查询
+
+**修改文件：**
+
+- `src/xr_rm_teleop/test/test_initial_joint_pose.py`
+- `src/xr_rm_teleop/xr_rm_teleop/realman_adapter.py`
+
+- [x] **步骤1：先写失败测试**
+
+在 `test_initial_joint_pose.py` 增加：
+
+```python
+def test_joint_degree_query_returns_validated_radians() -> None:
+    class FakeArm:
+        def rm_get_joint_degree(self):
+            return 0, [0.0, 10.0, -20.0, 30.0, -40.0, 50.0, -60.0]
+
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._arm = FakeArm()
+
+    snapshot = adapter.read_joint_state()
+
+    assert snapshot.positions == pytest.approx(
+        [math.radians(value) for value in [0, 10, -20, 30, -40, 50, -60]]
+    )
+    assert snapshot.read_duration_ms is not None
+
+
+@pytest.mark.parametrize(
+    "result",
+    [
+        (7, [0.0] * 7),
+        (0, [0.0] * 6),
+        (0, [0.0, 0.0, 0.0, math.nan, 0.0, 0.0, 0.0]),
+    ],
+)
+def test_joint_degree_query_rejects_sdk_errors_and_invalid_values(result) -> None:
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._arm = SimpleNamespace(rm_get_joint_degree=lambda: result)
+
+    with pytest.raises((RuntimeError, ValueError)):
+        adapter.read_joint_state()
+
+
+def test_mock_joint_query_uses_current_mock_positions() -> None:
+    adapter = MockRealManAdapter([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0])
+
+    snapshot = adapter.read_joint_state()
+
+    assert snapshot.positions == pytest.approx(
+        [math.radians(value) for value in [1, 2, 3, 4, 5, 6, 7]]
+    )
+```
+
+把现有 `test_invalid_udp_feedback_does_not_replace_snapshot` 的最终断言改为：
+
+```python
+after = adapter.get_latest_joint_state()
+assert after is not None
+assert before is not None
+assert after.positions == before.positions
+assert after.received_at == before.received_at
+assert after.motion_ready is False
+```
+
+该断言要求无效 UDP 帧保留最后已知角度，但立即禁止这些角度继续参与运动。
+
+- [x] **步骤2：运行测试并确认 RED**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_initial_joint_pose.py \
+  -k 'joint_degree_query or mock_joint_query or invalid_udp_feedback' -v
+```
+
+预期：测试失败；原因是适配器还没有 `read_joint_state()`，且无效 UDP 帧仍被标记为可运动。
+
+- [x] **步骤3：实现最小同步查询**
+
+给 `MockRealManAdapter` 增加：
+
+```python
+def read_joint_state(self) -> JointStateSnapshot:
+    return self.get_latest_joint_state()
+```
+
+给 `RealManAdapter` 增加：
+
+```python
+def read_joint_state(self) -> JointStateSnapshot:
+    self._require_arm()
+    started_at = time.monotonic()
+    result = self._arm.rm_get_joint_degree()
+    finished_at = time.monotonic()
+    if not isinstance(result, tuple) or len(result) != 2:
+        raise RuntimeError(
+            f"rm_get_joint_degree returned invalid result: {result!r}"
+        )
+    self._check_return(result, "rm_get_joint_degree")
+    return JointStateSnapshot(
+        self._joint_positions_from_degrees(
+            result[1],
+            "rm_get_joint_degree",
+        ),
+        finished_at,
+        (finished_at - started_at) * 1000.0,
+    )
+
+@staticmethod
+def _joint_positions_from_degrees(
+    values: Any,
+    source: str,
+) -> list[float]:
+    try:
+        degrees = list(values)
+    except TypeError as exc:
+        raise ValueError(f"{source} must contain 7 numeric joints") from exc
+    if len(degrees) != 7 or not all(
+        isinstance(value, Number) for value in degrees
+    ):
+        raise ValueError(f"{source} must contain 7 numeric joints")
+    positions = [math.radians(float(value)) for value in degrees]
+    if not all(math.isfinite(value) for value in positions):
+        raise ValueError(f"{source} contains NaN/Inf")
+    return positions
+```
+
+把 UDP 回调中重复的角度转换替换为：
+
+```python
+positions = self._joint_positions_from_degrees(
+    data.joint_status.joint_position,
+    "RM75 UDP feedback",
+)
+```
+
+在 UDP 回调的异常分支中，保留最后角度但标记为不可运动：
+
+```python
+with self._joint_state_lock:
+    if self._latest_joint_state is not None:
+        current = self._latest_joint_state
+        self._latest_joint_state = JointStateSnapshot(
+            list(current.positions),
+            current.received_at,
+            current.read_duration_ms,
+            current.update_interval_ms,
+            False,
+        )
+```
+
+- [x] **步骤4：运行适配器测试并确认 GREEN**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_initial_joint_pose.py -v
+```
+
+预期：该文件全部测试通过。
+
+- [x] **步骤5：检查本任务差异**
+
+```bash
+git diff --check
+git diff -- \
+  xr_rm_teleop/xr_rm_teleop/realman_adapter.py \
+  xr_rm_teleop/test/test_initial_joint_pose.py
+```
+
+预期：只有同步查询、共用角度校验、无效反馈安全标记及对应测试。
+
+---
+
+## 任务二：用启动查询结果初始化 QP 和关节命令历史
+
+**修改文件：**
+
+- `src/xr_rm_teleop/test/test_joint_control.py`
+- `src/xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`
+
+- [x] **步骤1：先写启动同步失败测试**
+
+增加：
+
+```python
+def test_startup_joint_query_initializes_qp_and_command_history() -> None:
+    positions = [0.1] * 7
+    pose = np.eye(4)
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._arm_name = "right_rm75"
+    teleop._adapter = SimpleNamespace(
+        read_joint_state=lambda: JointStateSnapshot(
+            positions,
+            time.monotonic(),
+        )
+    )
+    teleop._ik_solver = SimpleNamespace(
+        update_joint_state=lambda joints: pose
+    )
+    teleop.get_logger = lambda: FakeLogger()
+
+    teleop._initialize_joint_state()
+
+    assert teleop._latest_joint_positions == positions
+    assert teleop._last_valid_joint_target == positions
+    assert teleop._last_joint_command_target == positions
+    assert teleop._last_joint_command_velocity == [0.0] * 7
+    assert teleop._last_current_pose is pose
+
+
+def test_startup_joint_query_failure_closes_adapter() -> None:
+    class FailingAdapter:
+        def __init__(self):
+            self.close_calls = 0
+
+        def read_joint_state(self):
+            raise RuntimeError("rm_get_joint_degree failed with code 7")
+
+        def close(self):
+            self.close_calls += 1
+
+    errors = []
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._arm_name = "left_rm75"
+    teleop._adapter = FailingAdapter()
+    teleop.get_logger = lambda: SimpleNamespace(
+        error=lambda message: errors.append(message)
+    )
+
+    with pytest.raises(RuntimeError, match="code 7"):
+        teleop._initialize_joint_state()
+
+    assert teleop._adapter.close_calls == 1
+    assert "left_rm75" in errors[0]
+    assert "启动关节同步失败" in errors[0]
+```
+
+- [x] **步骤2：运行测试并确认 RED**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py \
+  -k 'startup_joint_query' -v
+```
+
+预期：测试因 `_initialize_joint_state()` 尚不存在而失败。
+
+- [x] **步骤3：增加启动同步**
+
+增加：
+
+```python
+def _initialize_joint_state(self) -> None:
+    try:
+        self._reset_joint_state(self._adapter.read_joint_state())
+    except Exception as exc:
+        self.get_logger().error(
+            f"{self._arm_name} 启动关节同步失败：{exc}"
+        )
+        self._adapter.close()
+        raise
+
+def _reset_joint_state(
+    self,
+    snapshot: JointStateSnapshot,
+) -> np.ndarray:
+    current_pose = self._ik_solver.update_joint_state(snapshot.positions)
+    positions = list(snapshot.positions)
+    self._latest_joint_positions = positions
+    self._last_current_pose = current_pose
+    self._last_valid_joint_target = list(positions)
+    self._last_joint_command_target = list(positions)
+    self._last_joint_command_velocity = [0.0] * 7
+    return current_pose
+```
+
+在现有适配器连接之后、外设初始化之前调用：
+
+```python
+self._adapter = self._make_adapter()
+self._adapter.connect()
+self._initialize_joint_state()
+self._setup_tool_control()
+```
+
+同步查询结果不得写入 `RealManAdapter._latest_joint_state`；该缓存继续只代表 UDP 反馈。
+
+- [x] **步骤4：运行启动与适配器测试**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py \
+  -k 'startup_joint_query or first_feedback' -v
+python3 -m pytest src/xr_rm_teleop/test/test_initial_joint_pose.py -v
+```
+
+预期：所选控制测试和全部适配器测试通过。
+
+- [x] **步骤5：检查本任务差异**
+
+```bash
+git diff --check
+git diff -- \
+  xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py \
+  xr_rm_teleop/test/test_joint_control.py
+```
+
+预期：启动阶段只增加一次同步读取，并初始化现有 QP/关节命令字段。
+
+---
+
+## 任务三：UDP 短暂超时保持，持续超时重新同步
+
+**修改文件：**
+
+- `src/xr_rm_teleop/test/test_joint_control.py`
+- `src/xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`
+
+- [x] **步骤1：先写超时行为测试**
+
+在测试文件增加最小构造器：
+
+```python
+def _timeout_teleop(adapter) -> SingleArmVelocityTeleop:
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._adapter = adapter
+    teleop._arm_name = "right_rm75"
+    teleop._follow = False
+    teleop._active = True
+    teleop._joint_feedback_ready = True
+    teleop._grip_rearm_required = False
+    teleop._feedback_resync_attempted = False
+    teleop._control_fault_latched = False
+    teleop._last_joint_command_target = [0.1] * 7
+    teleop._last_joint_command_velocity = [0.0] * 7
+    teleop._latest_joint_positions = [0.1] * 7
+    teleop._last_valid_joint_target = [0.1] * 7
+    teleop._last_current_pose = np.eye(4)
+    teleop._controller_start = None
+    teleop._controller_orientation_start = None
+    teleop._robot_start_transform = None
+    teleop._filtered_target = None
+    teleop._filtered_orientation_target = None
+    teleop._last_sent_target = None
+    teleop._last_sent_orientation = None
+    teleop._last_command_time = None
+    teleop._ik_solver = SimpleNamespace(
+        update_joint_state=lambda joints: np.eye(4)
+    )
+    teleop._stop_sent = False
+    teleop._feedback_resync_timeout_sec = 0.5
+    teleop._publish_stop_debug = lambda: None
+    teleop.get_logger = lambda: FakeLogger()
+    return teleop
+```
+
+增加：
+
+```python
+def test_missing_or_disabled_joint_snapshot_is_not_motion_ready() -> None:
+    assert not SingleArmVelocityTeleop._joint_snapshot_is_motion_ready(None)
+    assert not SingleArmVelocityTeleop._joint_snapshot_is_motion_ready(
+        JointStateSnapshot(
+            [0.0] * 7,
+            time.monotonic(),
+            motion_ready=False,
+        )
+    )
+
+
+def test_short_udp_timeout_repeats_last_limited_target_without_query() -> None:
+    sends = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: sends.append(
+            (list(joints), follow)
+        ),
+        read_joint_state=lambda: pytest.fail("query must not run"),
+        stop=lambda: pytest.fail("stop must not run"),
+    )
+    teleop = _timeout_teleop(adapter)
+
+    teleop._handle_stale_joint_feedback(0.2)
+
+    assert sends == [([0.1] * 7, False)]
+    assert teleop._last_joint_command_target == [0.1] * 7
+    assert teleop._grip_rearm_required
+
+
+def test_short_udp_timeout_without_active_target_stays_stopped() -> None:
+    stop_calls = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: pytest.fail(
+            "inactive control must not start CANFD output"
+        ),
+        read_joint_state=lambda: pytest.fail("query must not run"),
+        stop=lambda: stop_calls.append(True),
+    )
+    teleop = _timeout_teleop(adapter)
+    teleop._active = False
+
+    teleop._handle_stale_joint_feedback(0.2)
+
+    assert len(stop_calls) == 1
+
+
+def test_persistent_udp_timeout_queries_once_and_holds_actual_position() -> None:
+    sends = []
+    query_calls = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: sends.append(list(joints)),
+        read_joint_state=lambda: (
+            query_calls.append(True)
+            or JointStateSnapshot([0.2] * 7, time.monotonic())
+        ),
+        stop=lambda: None,
+    )
+    teleop = _timeout_teleop(adapter)
+
+    teleop._handle_stale_joint_feedback(0.5)
+    teleop._handle_stale_joint_feedback(0.6)
+
+    assert len(query_calls) == 1
+    assert sends == [[0.2] * 7, [0.2] * 7]
+    assert teleop._last_valid_joint_target == [0.2] * 7
+    assert teleop._last_joint_command_velocity == [0.0] * 7
+
+
+def test_persistent_udp_timeout_query_failure_latches_control() -> None:
+    stop_calls = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: pytest.fail(
+            "CANFD must stop after query failure"
+        ),
+        read_joint_state=lambda: (_ for _ in ()).throw(
+            RuntimeError("rm_get_joint_degree failed with code 7")
+        ),
+        stop=lambda: stop_calls.append(True),
+    )
+    teleop = _timeout_teleop(adapter)
+
+    teleop._handle_stale_joint_feedback(0.5)
+    teleop._handle_stale_joint_feedback(0.6)
+
+    assert teleop._control_fault_latched
+    assert len(stop_calls) == 1
+```
+
+删除旧的 `_fresh_joint_state()` 直接测试，并用
+`test_short_udp_timeout_without_active_target_stays_stopped` 替换旧的
+`test_stale_feedback_stops_before_active_control`。
+
+在 `test_feedback_fault_blocks_grip_until_release` 中补齐：
+
+```python
+teleop._control_fault_latched = False
+teleop._feedback_resync_attempted = False
+```
+
+- [x] **步骤2：运行测试并确认 RED**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py \
+  -k 'udp_timeout or joint_snapshot' -v
+```
+
+预期：测试因超时处理和锁存状态尚不存在而失败。
+
+- [x] **步骤3：增加参数与最小状态**
+
+参数默认值：
+
+```python
+self.declare_parameter("control_rate_hz", 90.0)
+self.declare_parameter("command_timeout_sec", 0.12)
+self.declare_parameter("feedback_resync_timeout_sec", 0.5)
+```
+
+读取并初始化：
+
+```python
+self._feedback_resync_timeout_sec = float(
+    self.get_parameter("feedback_resync_timeout_sec").value
+)
+self._feedback_resync_attempted = False
+self._control_fault_latched = False
+```
+
+在 `_validate_parameters()` 中增加：
+
+```python
+if self._feedback_resync_timeout_sec <= self._command_timeout_sec:
+    raise ValueError(
+        "feedback_resync_timeout_sec must be greater than command_timeout_sec"
+    )
+```
+
+- [x] **步骤4：增加保持、重新同步和锁存逻辑**
+
+增加：
+
+```python
+def _handle_stale_joint_feedback(self, age: float) -> None:
+    if self._control_fault_latched:
+        return
+    self._grip_rearm_required = True
+    if self._joint_feedback_ready:
+        self.get_logger().warn(
+            f"{self._arm_name} UDP关节反馈超时，保持最后安全目标。"
+        )
+    self._joint_feedback_ready = False
+
+    if (
+        age >= self._feedback_resync_timeout_sec
+        and not self._feedback_resync_attempted
+    ):
+        self._feedback_resync_attempted = True
+        self.get_logger().warn(
+            f"{self._arm_name} UDP关节反馈持续超时，"
+            "尝试rm_get_joint_degree重新同步。"
+        )
+        try:
+            self._reset_joint_state(self._adapter.read_joint_state())
+        except Exception as exc:
+            self._latch_control_fault(
+                f"UDP关节反馈持续超时且重新同步失败：{exc}"
+            )
+            return
+        self.get_logger().info(
+            f"{self._arm_name} 已通过rm_get_joint_degree重新同步，"
+            "继续保持并等待UDP恢复。"
+        )
+
+    if self._active and self._last_joint_command_target is not None:
+        self._repeat_last_joint_target()
+    else:
+        self._safe_stop(reset_active=True)
+
+def _repeat_last_joint_target(self) -> None:
+    target = self._last_joint_command_target
+    if target is None:
+        return
+    self._adapter.send_joint_target(list(target), self._follow)
+    self._stop_sent = False
+
+def _latch_control_fault(self, message: str) -> None:
+    if self._control_fault_latched:
+        return
+    self._control_fault_latched = True
+    self._grip_rearm_required = True
+    self.get_logger().error(
+        f"{self._arm_name} 控制故障已锁存：{message}"
+    )
+    self._safe_stop(reset_active=True)
+```
+
+- [x] **步骤5：在 QP 之前处理反馈状态**
+
+在 `_control_tick()` 开头用以下逻辑替换现有 `_fresh_joint_state()` 分支：
+
+```python
+if self._control_fault_latched:
+    return
+
+snapshot = self._adapter.get_latest_joint_state()
+if not self._joint_snapshot_is_motion_ready(snapshot):
+    self._grip_rearm_required = True
+    if self._joint_feedback_ready:
+        self.get_logger().warn(
+            f"{self._arm_name} 关节反馈无效或机械臂未就绪，机械臂停止。"
+        )
+    self._joint_feedback_ready = False
+    self._safe_stop(reset_active=True)
+    return
+
+feedback_age = time.monotonic() - snapshot.received_at
+if feedback_age < 0.0:
+    self._grip_rearm_required = True
+    self._joint_feedback_ready = False
+    self._safe_stop(reset_active=True)
+    return
+if feedback_age > self._command_timeout_sec:
+    self._handle_stale_joint_feedback(feedback_age)
+    return
+
+self._feedback_resync_attempted = False
+```
+
+在成功执行 `_sync_joint_feedback(snapshot)` 后，用以下逻辑替换现有首次反馈日志：
+
+```python
+if not self._joint_feedback_ready:
+    if self._grip_rearm_required:
+        message = (
+            f"{self._arm_name} UDP关节反馈已恢复，"
+            "等待Grip松开后重新使能。"
+        )
+    else:
+        message = (
+            f"{self._arm_name} 已收到首帧有效关节反馈，QP可以启用。"
+        )
+    self.get_logger().info(message)
+    self._joint_feedback_ready = True
+```
+
+用以下静态校验替换 `_fresh_joint_state()`：
+
+```python
+@staticmethod
+def _joint_snapshot_is_motion_ready(
+    snapshot: JointStateSnapshot | None,
+) -> bool:
+    return (
+        snapshot is not None
+        and len(snapshot.positions) == 7
+        and all(math.isfinite(value) for value in snapshot.positions)
+        and snapshot.motion_ready
+    )
+```
+
+XR手柄消息超时、Grip逻辑、工作空间/圆柱限位、姿态限速和关节限速保持原样。
+
+- [x] **步骤6：运行超时及反馈安全测试**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py \
+  -k 'udp_timeout or feedback_fault or joint_snapshot' -v
+```
+
+预期：所选测试通过；短暂超时不调用 QP 和同步查询，机械臂未就绪仍立即停止。
+
+- [x] **步骤7：检查本任务差异**
+
+```bash
+git diff --check
+git diff --stat
+```
+
+预期：没有新增状态机类、线程、依赖、连接或常态轮询。
+
+---
+
+## 任务四：CANFD 错误后停止、查询并等待人工恢复
+
+**修改文件：**
+
+- `src/xr_rm_teleop/test/test_joint_control.py`
+- `src/xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`
+
+- [x] **步骤1：先写 CANFD 恢复测试**
+
+用以下测试替换旧的 `test_joint_send_failure_requests_slow_stop_and_resets_control`：
+
+```python
+def test_canfd_error_stops_queries_and_requires_grip_rearm() -> None:
+    class RecoveringAdapter:
+        def __init__(self):
+            self.stop_calls = 0
+            self.read_calls = 0
+
+        def send_joint_target(self, joints, follow):
+            raise RuntimeError("rm_movej_canfd failed with code 9")
+
+        def stop(self):
+            self.stop_calls += 1
+
+        def read_joint_state(self):
+            self.read_calls += 1
+            return JointStateSnapshot([0.2] * 7, time.monotonic())
+
+    teleop = _timeout_teleop(RecoveringAdapter())
+    teleop._joint_command_max_speed = math.radians(180.0)
+    teleop._joint_command_max_acceleration = math.radians(300.0)
+    teleop._dt = 1.0 / 90.0
+
+    sent = teleop._send_joint_target([0.3] * 7)
+
+    assert not sent
+    assert teleop._adapter.stop_calls == 1
+    assert teleop._adapter.read_calls == 1
+    assert not teleop._control_fault_latched
+    assert teleop._grip_rearm_required
+    assert teleop._last_joint_command_target == [0.2] * 7
+
+
+def test_canfd_error_latches_when_joint_query_also_fails() -> None:
+    class FailingAdapter:
+        def __init__(self):
+            self.stop_calls = 0
+
+        def send_joint_target(self, joints, follow):
+            raise RuntimeError("rm_movej_canfd failed with code 9")
+
+        def stop(self):
+            self.stop_calls += 1
+
+        def read_joint_state(self):
+            raise RuntimeError("rm_get_joint_degree failed with code 7")
+
+    teleop = _timeout_teleop(FailingAdapter())
+    teleop._joint_command_max_speed = math.radians(180.0)
+    teleop._joint_command_max_acceleration = math.radians(300.0)
+    teleop._dt = 1.0 / 90.0
+
+    assert not teleop._send_joint_target([0.3] * 7)
+    assert teleop._control_fault_latched
+    assert teleop._adapter.stop_calls == 1
+```
+
+- [x] **步骤2：运行测试并确认 RED**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py \
+  -k 'canfd_error' -v
+```
+
+预期：测试失败；当前发送错误只会 slow-stop，不会查询实际关节角或锁存查询失败。
+
+- [x] **步骤3：增加统一 CANFD 恢复路径**
+
+增加：
+
+```python
+def _recover_from_canfd_error(self, send_error: Exception) -> None:
+    self.get_logger().error(
+        f"{self._arm_name} rm_movej_canfd发送失败：{send_error}"
+    )
+    self._grip_rearm_required = True
+    self._send_stop_once()
+    self._safe_stop(reset_active=True)
+    try:
+        snapshot = self._adapter.read_joint_state()
+        self._reset_joint_state(snapshot)
+    except Exception as query_error:
+        self._latch_control_fault(
+            "CANFD错误后关节同步失败："
+            f"send={send_error}; query={query_error}"
+        )
+        return
+    self.get_logger().info(
+        f"{self._arm_name} CANFD错误后已同步实际关节角，"
+        "等待UDP恢复及Grip重新使能。"
+    )
+```
+
+把 `_send_joint_target()` 的异常分支替换为：
+
+```python
+except Exception as exc:
+    self._recover_from_canfd_error(exc)
+    return False
+```
+
+让短暂超时重发也走相同错误恢复：
+
+```python
+def _repeat_last_joint_target(self) -> None:
+    target = self._last_joint_command_target
+    if target is None:
+        return
+    try:
+        self._adapter.send_joint_target(list(target), self._follow)
+        self._stop_sent = False
+    except Exception as exc:
+        self._recover_from_canfd_error(exc)
+```
+
+- [x] **步骤4：运行全部关节控制测试**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py -v
+```
+
+预期：全部关节控制测试通过。
+
+- [x] **步骤5：检查日志与差异**
+
+```bash
+rg -n "rm_movej_canfd发送失败|控制故障已锁存|rm_get_joint_degree" \
+  src/xr_rm_teleop/xr_rm_teleop
+git diff --check
+```
+
+预期：
+
+- CANFD错误和查询错误都包含机械臂名称及失败阶段。
+- 锁存分支不会每周期重复打印错误。
+- 正常QP输出和超时保持使用同一个CANFD恢复入口。
+
+---
+
+## 任务五：同步90 Hz配置并完成mock验证
+
+**修改文件：**
+
+- `src/xr_rm_bringup/config/dual_arm_rm75.yaml`
+- `src/xr_rm_bringup/config/left_arm_rm75.yaml`
+- `src/xr_rm_bringup/config/right_arm_rm75.yaml`
+
+- [x] **步骤1：只修改请求中的控制参数**
+
+三份配置的每个机械臂条目统一为：
+
+```yaml
+control_rate_hz: 90.0
+command_timeout_sec: 0.12
+feedback_resync_timeout_sec: 0.5
+```
+
+保留低跟随：
+
+```yaml
+follow: false
+```
+
+不得修改：
+
+- 工作空间与圆柱限位。
+- TCP线速度、角速度及关节速度/加速度限制。
+- `configure_safety_limits: true`。
+- `move_to_initial_pose_on_connect: false`。
+- 机械臂IP、端口、初始位姿和末端工具配置。
+- 双臂节点名 `left_arm_teleop`、`right_arm_teleop`。
+
+- [x] **步骤2：机械检查三份配置**
+
+```bash
+rg -n "control_rate_hz|command_timeout_sec|feedback_resync_timeout_sec|follow:" \
+  src/xr_rm_bringup/config/dual_arm_rm75.yaml \
+  src/xr_rm_bringup/config/left_arm_rm75.yaml \
+  src/xr_rm_bringup/config/right_arm_rm75.yaml
+```
+
+预期：共四个机械臂配置条目，每个条目均为90.0、0.12、0.5和`follow: false`；三份文件不再出现125.0。
+
+- [x] **步骤3：运行相关测试**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest src/xr_rm_teleop/test/test_initial_joint_pose.py -v
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py -v
+python3 -m pytest src/xr_rm_teleop/test/test_orientation_control.py -v
+```
+
+预期：三个命令均以0退出且无失败。
+
+- [x] **步骤4：构建ROS2工作空间**
+
+```bash
+source /opt/ros/humble/setup.bash
+colcon build --symlink-install
+```
+
+预期：`xr_rm_interfaces`、`xr_rm_input`、`xr_rm_teleop`、`xr_rm_bringup` 构建成功。
+
+- [x] **步骤5：通过统一入口进行mock启动验证**
+
+```bash
+source /opt/ros/humble/setup.bash
+source install/setup.bash
+if timeout --signal=INT 10s ros2 launch xr_rm_bringup arm_debug.launch.py \
+  arm:=right use_mock:=true udp_port:=15123
+then
+  true
+else
+  launch_status=$?
+  test "$launch_status" -eq 124
+fi
+```
+
+预期：
+
+- `udp_controller_receiver` 与 `single_arm_velocity_teleop` 正常启动。
+- 遥操作节点报告90 Hz、低跟随，并完成mock关节状态初始化。
+- 不导入厂商SDK，不建立RealMan连接，不发送CANFD，不移动机械臂，不操作夹爪。
+- 10秒后由`timeout`结束；仅该超时允许退出码124。
+
+- [x] **步骤6：最终范围与安全审计**
+
+```bash
+git diff --check
+git status --short
+git diff --stat
+git diff -- \
+  xr_rm_teleop/xr_rm_teleop/realman_adapter.py \
+  xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py \
+  xr_rm_teleop/test/test_initial_joint_pose.py \
+  xr_rm_teleop/test/test_joint_control.py \
+  xr_rm_bringup/config/dual_arm_rm75.yaml \
+  xr_rm_bringup/config/left_arm_rm75.yaml \
+  xr_rm_bringup/config/right_arm_rm75.yaml
+```
+
+逐项确认：
+
+- 没有无关文件或格式化改动。
+- 没有提交、推送、分支、worktree、锁文件、CI、格式化规则或依赖变化。
+- 没有新增线程、ROS包、launch入口、并发RealMan连接或常态SDK轮询。
+- mock模式不导入、不依赖厂商SDK。
+- `configure_safety_limits` 默认仍为开启。
+- `move_to_initial_pose_on_connect` 默认仍为关闭。
+- `left_arm_teleop`、`right_arm_teleop` 节点名不变。
+- 工作空间/圆柱限位、TCP与关节限速、XR命令超时和slow-stop逻辑仍保留。
+- 验证期间未连接真机、移动机械臂或操作夹爪。

docs/superpowers/plans/2026-07-29-rm75-qp-convergence.md

@@ -0,0 +1,412 @@
+# RM75 QP 收敛优化实施计划
+
+> **执行要求：** 使用 `superpowers:executing-plans` 逐项执行。用户未授权
+> subagent、独立worktree或本地分支，因此本计划只允许当前会话内联实施。所有
+> 步骤使用复选框跟踪。
+
+**目标：** 将当前每周期单步QP改为有界迭代QP，使低跟随RM75在手柄移动10 cm
+后约1秒内稳定到位，并消除由近距离台阶目标造成的持续轻微晃动。
+
+**实现方式：** 每个正常控制周期仍先用UDP实际关节角同步Placo，然后在一次
+`PlacoIkSolver.solve()`内部最多迭代30次，提前达到1 mm位置误差和0.005 rad
+姿态误差即返回。最终关节解继续经过现有90 Hz关节速度与加速度限幅后，以
+`follow=false`发送；不增加预测状态、线程、连接、依赖或配置参数。
+
+**技术栈：** Python 3.10、ROS2 Humble、Placo 0.9.4、NumPy、pytest、
+ament/colcon。
+
+**设计文档：**
+`docs/superpowers/specs/2026-07-29-rm75-qp-convergence-design.md`
+
+---
+
+## 仓库与安全约束
+
+- 构建、测试和启动命令在 `/home/robot/WS_xr` 执行。
+- Git命令在 `/home/robot/WS_xr/src` 执行。
+- 每次构建、测试或启动前执行 `source /opt/ros/humble/setup.bash`。
+- 不自动提交、推送、创建分支或worktree。
+- 不连接真机，不发送真实CANFD，不移动机械臂，不操作夹爪。
+- 只通过 `arm_debug.launch.py arm:=right use_mock:=true`进行启动验证。
+- 不修改三份机械臂YAML、RealMan适配器、launch、UI、依赖或公开入口。
+- 保留工作空间、圆柱、TCP速度、姿态速度、关节速度、关节加速度、反馈超时、
+  CANFD恢复、Grip重新使能和安全停止逻辑。
+
+## 文件范围
+
+- 修改 `xr_rm_teleop/test/test_placo_transforms.py`
+  - 增加真实Placo 7 cm目标收敛回归测试。
+- 修改 `xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py`
+  - 增加固定上限、提前收敛和逐步安全校验。
+
+不需要修改 `single_arm_velocity_teleop.py`；现有 `_solve_joint_target()` 已负责
+QP异常时打印限频警告并保持上一组安全关节目标，现有
+`_limit_joint_command_step()` 已负责最终90 Hz真实命令限速。
+
+---
+
+## 任务一：用真实Placo复现单步QP不收敛
+
+**修改文件：**
+
+- `xr_rm_teleop/test/test_placo_transforms.py`
+
+- [x] **步骤1：增加测试辅助函数**
+
+在文件顶部增加：
+
+```python
+import math
+```
+
+在现有URDF结构测试之后增加：
+
+```python
+def _rm75_placo_solver() -> tuple[PlacoIkSolver, list[float]]:
+    pytest.importorskip("placo")
+    urdf_path = (
+        Path(__file__).resolve().parents[1]
+        / "models"
+        / "rm75_omnipicker"
+        / "urdf"
+        / "RM75-B_OmniPicker_fixed.urdf"
+    )
+    joints = [
+        math.radians(value)
+        for value in [
+            -90.14,
+            3.76,
+            -86.89,
+            87.89,
+            -96.53,
+            -79.62,
+            -90.04,
+        ]
+    ]
+    return PlacoIkSolver(str(urdf_path), 1.0 / 90.0), joints
+```
+
+`importorskip()`只让没有Placo的普通系统Python跳过真模型用例；下面的RED/GREEN
+命令会显式加入项目现有Placo 0.9.4路径，因此该用例必须实际执行而不能跳过。
+
+- [x] **步骤2：增加7 cm目标收敛测试**
+
+增加：
+
+```python
+def test_qp_solve_converges_to_reachable_tcp_target() -> None:
+    solver, joints = _rm75_placo_solver()
+    start_pose = solver.update_joint_state(joints)
+    target_pose = start_pose.copy()
+    target_pose[0, 3] += 0.07
+
+    result = solver.solve(target_pose)
+    reached_pose = solver.update_joint_state(result)
+    position_error = np.linalg.norm(
+        target_pose[:3, 3] - reached_pose[:3, 3]
+    )
+    rotation_delta = (
+        target_pose[:3, :3] @ reached_pose[:3, :3].T
+    )
+    orientation_error = math.acos(
+        float(
+            np.clip(
+                (np.trace(rotation_delta) - 1.0) * 0.5,
+                -1.0,
+                1.0,
+            )
+        )
+    )
+
+    assert position_error <= 1e-3
+    assert orientation_error <= 5e-3
+```
+
+该测试验证一次公开 `solve()` 调用返回当前TCP目标对应的收敛关节解，而不是验证
+内部迭代次数。
+
+- [x] **步骤3：运行测试并确认RED**
+
+```bash
+source /opt/ros/humble/setup.bash
+export RM75_PLACO_TEST_PATH="/home/robot/WS_xr/src/xr_rm_teleop:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages/cmeel.prefix/lib/python3.10/site-packages"
+PYTHONPATH="${RM75_PLACO_TEST_PATH}:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test/test_placo_transforms.py::test_qp_solve_converges_to_reachable_tcp_target \
+  -v
+```
+
+预期：测试以位置误差约0.063 m大于0.001 m失败，证明当前单步QP确实不能在一次
+调用内给出收敛关节目标。测试不得因导入错误或跳过而结束。
+
+---
+
+## 任务二：实现有界迭代QP
+
+**修改文件：**
+
+- `xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py`
+
+- [x] **步骤1：增加固定收敛常量**
+
+把模块说明改为：
+
+```python
+"""RM75 的 Placo 0.9.4 有界迭代 QP 逆解。"""
+```
+
+在现有常量后增加：
+
+```python
+QP_MAX_ITERATIONS = 30
+QP_POSITION_TOLERANCE_M = 1e-3
+QP_ORIENTATION_TOLERANCE_RAD = 5e-3
+```
+
+这些值是本次已确认的算法边界，不新增ROS参数。
+
+- [x] **步骤2：增加任务误差读取**
+
+在 `solve()` 前增加：
+
+```python
+def _target_errors(self) -> tuple[float, float]:
+    position_task = self._frame_task.position()
+    orientation_task = self._frame_task.orientation()
+    position_task.update()
+    orientation_task.update()
+    return (
+        float(position_task.error_norm()),
+        float(orientation_task.error_norm()),
+    )
+```
+
+Placo在 `solve(True)` 后只更新关节状态；先更新机器人运动学，再显式更新两个任务，
+确保 `error_norm()`对应当前迭代后的状态而不是前一迭代。
+
+- [x] **步骤3：把单步求解改为最多30次且提前收敛**
+
+用以下实现替换现有 `solve()`：
+
+```python
+def solve(self, target_tool_pose: np.ndarray) -> list[float]:
+    if self._actual_joints is None:
+        raise RuntimeError(
+            "joint state must be initialized before QP solve"
+        )
+    self._frame_task.T_world_frame = _validated_transform(
+        target_tool_pose
+    )
+    result = np.asarray(
+        self._robot.state.q[RM75_Q_SLICE],
+        dtype=float,
+    ).copy()
+    position_error, orientation_error = self._target_errors()
+    if (
+        position_error <= QP_POSITION_TOLERANCE_M
+        and orientation_error <= QP_ORIENTATION_TOLERANCE_RAD
+    ):
+        return result.tolist()
+
+    for _ in range(QP_MAX_ITERATIONS):
+        previous = result
+        self._solver.solve(True)
+        self._robot.update_kinematics()
+        result = np.asarray(
+            self._robot.state.q[RM75_Q_SLICE],
+            dtype=float,
+        ).copy()
+        self._validate_result(result, previous)
+        position_error, orientation_error = self._target_errors()
+        if (
+            position_error <= QP_POSITION_TOLERANCE_M
+            and orientation_error <= QP_ORIENTATION_TOLERANCE_RAD
+        ):
+            return result.tolist()
+
+    raise RuntimeError(
+        "QP did not converge after "
+        f"{QP_MAX_ITERATIONS} iterations: "
+        f"position_error={position_error:.6f} m, "
+        f"orientation_error={orientation_error:.6f} rad"
+    )
+```
+
+目标已到达时直接返回当前关节角，避免静止时进行不必要的数值迭代。
+
+- [x] **步骤4：让速度校验针对每次数值迭代**
+
+把 `_validate_result()` 签名改为：
+
+```python
+def _validate_result(
+    self,
+    result: np.ndarray,
+    reference: np.ndarray | None = None,
+) -> None:
+```
+
+保留现有有限值和关节位置检查，把速度检查替换为：
+
+```python
+if reference is None:
+    reference = self._actual_joints
+if reference is None:
+    raise RuntimeError("joint state has not been initialized")
+reference = np.asarray(reference, dtype=float)
+if reference.shape != (7,) or not np.isfinite(reference).all():
+    raise ValueError("QP reference must contain 7 finite values")
+max_step = self._velocity_limits * self._dt + 1e-9
+if np.any(np.abs(result - reference) > max_step):
+    raise ValueError(
+        "QP result violates RM75 one-cycle velocity limits"
+    )
+```
+
+这样每次内部数值迭代继续满足Placo的URDF关节速度边界；最终收敛解仍由节点现有
+`_limit_joint_command_step()`按真实90 Hz周期限制后才发送。
+
+- [x] **步骤5：运行目标测试并确认GREEN**
+
+```bash
+source /opt/ros/humble/setup.bash
+export RM75_PLACO_TEST_PATH="/home/robot/WS_xr/src/xr_rm_teleop:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages/cmeel.prefix/lib/python3.10/site-packages"
+PYTHONPATH="${RM75_PLACO_TEST_PATH}:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test/test_placo_transforms.py::test_qp_solve_converges_to_reachable_tcp_target \
+  src/xr_rm_teleop/test/test_placo_transforms.py::test_qp_result_rejects_nan_position_and_velocity_violations \
+  -v
+```
+
+预期：两个测试通过；真实Placo用例不被跳过。
+
+- [x] **步骤6：运行Placo变换测试文件**
+
+```bash
+source /opt/ros/humble/setup.bash
+export RM75_PLACO_TEST_PATH="/home/robot/WS_xr/src/xr_rm_teleop:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages/cmeel.prefix/lib/python3.10/site-packages"
+PYTHONPATH="${RM75_PLACO_TEST_PATH}:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test/test_placo_transforms.py \
+  -v
+```
+
+预期：全部通过，无失败或跳过。
+
+---
+
+## 任务三：回归、安全和mock验证
+
+**验证范围：**
+
+- `xr_rm_teleop`全部测试；
+- ROS2工作空间构建；
+- 统一launch的右臂mock启动；
+- 最终差异与安全配置审计。
+
+- [x] **步骤1：运行遥操作包全部测试**
+
+```bash
+source /opt/ros/humble/setup.bash
+export RM75_PLACO_TEST_PATH="/home/robot/WS_xr/src/xr_rm_teleop:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages/cmeel.prefix/lib/python3.10/site-packages"
+PYTHONPATH="${RM75_PLACO_TEST_PATH}:${PYTHONPATH:-}" \
+  python3 -m pytest src/xr_rm_teleop/test -v
+```
+
+预期：全部测试通过，真实Placo收敛用例被执行。
+
+- [x] **步骤2：按项目规则单独运行姿态控制测试**
+
+```bash
+source /opt/ros/humble/setup.bash
+python3 -m pytest \
+  src/xr_rm_teleop/test/test_orientation_control.py \
+  -v
+```
+
+预期：全部通过。
+
+- [x] **步骤3：构建ROS2工作空间**
+
+```bash
+source /opt/ros/humble/setup.bash
+colcon build --symlink-install
+```
+
+预期：`xr_rm_interfaces`、`xr_rm_input`、`xr_rm_teleop`和
+`xr_rm_bringup`全部构建成功。
+
+- [x] **步骤4：通过统一入口进行右臂mock启动验证**
+
+```bash
+source /opt/ros/humble/setup.bash
+source install/setup.bash
+if timeout --signal=INT 10s ros2 launch \
+  xr_rm_bringup arm_debug.launch.py \
+  arm:=right use_mock:=true udp_port:=15123
+then
+  true
+else
+  launch_status=$?
+  test "$launch_status" -eq 124
+fi
+```
+
+预期：
+
+- `udp_controller_receiver`和`single_arm_velocity_teleop`正常启动；
+- 节点报告 `dt=0.0111s`、`follow=False`；
+- mock关节初始化成功；
+- 不导入RealMan SDK，不建立真机连接，不发送CANFD；
+- 10秒后仅由 `timeout`结束。
+
+- [x] **步骤5：最终差异和安全审计**
+
+在 `/home/robot/WS_xr/src` 执行：
+
+```bash
+git diff --check
+git status --short
+git diff -- \
+  xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py \
+  xr_rm_teleop/test/test_placo_transforms.py
+rg -n \
+  "control_rate_hz|follow:|configure_safety_limits|move_to_initial_pose_on_connect" \
+  xr_rm_bringup/config/dual_arm_rm75.yaml \
+  xr_rm_bringup/config/left_arm_rm75.yaml \
+  xr_rm_bringup/config/right_arm_rm75.yaml
+```
+
+预期：
+
+- 生产代码只修改Placo求解器；
+- 测试只增加真实模型收敛验证；
+- 三份配置继续使用90 Hz、`follow: false`、
+  `configure_safety_limits: true`和
+  `move_to_initial_pose_on_connect: false`；
+- 不改变此前由用户保留的 `AGENTS.md` 修改；
+- 不自动提交或推送。
+
+---
+
+## 真机交接验收
+
+Codex不执行本节。自动验证全部通过后，由用户在安全工作区使用：
+
+```bash
+ros2 launch xr_rm_bringup arm_debug.launch.py \
+  arm:=right use_mock:=false
+```
+
+验收步骤：
+
+1. 急停可用、Grip松开、工作区无人后启动。
+2. 按住Grip，快速移动手柄约10 cm后保持不动。
+3. 机械臂应在约1秒内稳定到位，无持续肉眼可见晃动。
+4. 连续观察四个5秒 timing 窗口，`total max`均低于11.111 ms。
+5. 不应出现QP未收敛、反馈超时、CANFD错误或故障锁存日志。
+6. 松开Grip后机械臂按现有逻辑安全停止。
+
+若任一窗口 `total max`达到或超过11.111 ms，或机械臂出现明显振荡，立即松开
+Grip并停止测试，把完整timing和错误日志返回后再调整；不得直接提高控制频率、
+关闭限速或改成高跟随。

docs/superpowers/plans/2026-07-30-rm75-joint-command-braking.md

@@ -0,0 +1,309 @@
+# RM75 关节命令提前制动实施计划
+
+> **供智能体执行者：** 必须使用 `superpowers:subagent-driven-development`
+>（推荐）或 `superpowers:executing-plans` 逐项实施；所有步骤使用复选框跟踪。
+
+**目标：** 修复 90 Hz 关节命令在稳定目标附近反复越界的问题，使 RM75 在保留
+现有速度、加速度限制和低跟随模式的前提下提前制动并稳定停止。
+
+**架构：** 保留当前 QP、反馈和故障恢复链路，只替换
+`SingleArmVelocityTeleop._limit_joint_command_step()` 内部的关节命令生成规则。
+每个关节根据离散制动距离决定继续加速或开始减速，最终命令仍由现有
+`_send_joint_target()` 发送。
+
+**技术栈：** Python 3.10、ROS2 Humble、NumPy、pytest、ament/colcon。
+
+---
+
+## 执行约束
+
+- 设计文档：
+  `docs/superpowers/specs/2026-07-30-rm75-joint-command-braking-design.md`。
+- 构建、测试和启动命令在 `/home/robot/WS_xr` 执行，并先运行
+  `source /opt/ros/humble/setup.bash`。
+- 不连接真机，不发送真实 CANFD，不移动机械臂，不操作夹爪。
+- 启动验证只使用
+  `xr_rm_bringup/launch/arm_debug.launch.py arm:=right use_mock:=true`。
+- 不修改 QP、YAML、RealMan 适配器、launch、UI、依赖和公开 API。
+- 保留工作空间、圆柱、TCP 速度、姿态速度、关节速度、关节加速度、超时保持、
+  CANFD 恢复、Grip 重新使能和安全停止逻辑。
+- 用户未要求 Git 提交，因此本计划不执行 `git commit` 或 `git push`。
+- 保留工作区中已有的其他修改，不回退、不覆盖：
+  `placo_ik_solver.py`、`test_placo_transforms.py` 及现有 Superpowers 文档。
+
+## 文件范围
+
+- 修改 `xr_rm_teleop/test/test_joint_control.py`
+  - 增加固定目标提前制动回归测试。
+- 修改 `xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`
+  - 在现有关节限幅入口实现离散制动距离判断。
+- 不创建新的运行时代码文件或配置项。
+
+### 任务一：增加持续振荡回归测试
+
+**文件：**
+
+- 修改：`xr_rm_teleop/test/test_joint_control.py:206`
+- 测试：`xr_rm_teleop/test/test_joint_control.py`
+
+- [x] **步骤 1：在现有首周期加速度测试后增加固定目标测试**
+
+增加以下测试：
+
+```python
+def test_joint_command_step_brakes_before_fixed_target_without_overshoot() -> None:
+    dt = 1.0 / 90.0
+    max_speed = math.radians(180.0)
+    max_acceleration = math.radians(300.0)
+    target = np.radians(
+        [10.0, -10.0, 3.0, -3.0, 1.0, -1.0, 0.1]
+    ).tolist()
+    command = [0.0] * 7
+    velocity = [0.0] * 7
+
+    for _ in range(180):
+        previous_velocity = list(velocity)
+        command, velocity = (
+            SingleArmVelocityTeleop._limit_joint_command_step(
+                target=target,
+                previous_target=command,
+                previous_velocity=velocity,
+                max_speed=max_speed,
+                max_acceleration=max_acceleration,
+                dt=dt,
+            )
+        )
+
+        for index in range(7):
+            assert min(0.0, target[index]) - 1e-12 <= command[index]
+            assert command[index] <= max(0.0, target[index]) + 1e-12
+            assert abs(velocity[index]) <= max_speed + 1e-12
+            assert (
+                abs(velocity[index] - previous_velocity[index])
+                <= max_acceleration * dt + 1e-12
+            )
+
+    assert command == pytest.approx(target, abs=1e-12)
+    assert velocity == pytest.approx([0.0] * 7, abs=1e-12)
+```
+
+该测试同时覆盖正负方向、不同目标距离、最大速度、最大加速度、禁止越过固定目标
+和最终停止。
+
+- [x] **步骤 2：运行新增测试并确认失败**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+PYTHONPATH="/home/robot/WS_xr/src/xr_rm_teleop:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test/test_joint_control.py::test_joint_command_step_brakes_before_fixed_target_without_overshoot \
+  -v
+```
+
+预期：`FAIL`，现有实现会让至少一个关节命令越过固定目标。失败原因必须来自新增
+越界断言，不能是导入或环境错误。
+
+### 任务二：实现离散提前制动
+
+**文件：**
+
+- 修改：
+  `xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py:1232-1263`
+- 测试：`xr_rm_teleop/test/test_joint_control.py`
+
+- [x] **步骤 1：用离散制动逻辑替换现有限幅计算**
+
+保留方法签名和现有长度、参数校验，将
+`desired_velocity = np.clip(...)` 到返回值的部分替换为：
+
+```python
+        values = np.asarray(
+            [target, previous_target, previous_velocity],
+            dtype=float,
+        )
+        if not np.isfinite(values).all():
+            raise ValueError("joint command contains NaN/Inf")
+
+        velocity_step = max_acceleration * dt
+        arrival_distance = velocity_step * dt
+        limited_target = []
+        limited_velocity = []
+        for desired_target, last_target, last_velocity in zip(
+            target,
+            previous_target,
+            previous_velocity,
+        ):
+            error = desired_target - last_target
+            if (
+                abs(last_velocity) <= 1e-12
+                and abs(error) <= arrival_distance
+            ):
+                velocity = error / dt
+                position = desired_target
+            else:
+                direction = (
+                    math.copysign(1.0, error)
+                    if abs(error) > 1e-12
+                    else 0.0
+                )
+                accelerated_speed = min(
+                    abs(last_velocity) + velocity_step,
+                    max_speed,
+                )
+                braking_steps = max(
+                    0,
+                    math.ceil(accelerated_speed / velocity_step) - 1,
+                )
+                braking_distance = accelerated_speed * dt + dt * (
+                    braking_steps * accelerated_speed
+                    - velocity_step
+                    * braking_steps
+                    * (braking_steps + 1)
+                    / 2.0
+                )
+                desired_velocity = direction * max_speed
+                if (
+                    last_velocity * error > 0.0
+                    and abs(error) <= braking_distance
+                ):
+                    desired_velocity = 0.0
+                velocity = _clamp(
+                    desired_velocity,
+                    last_velocity - velocity_step,
+                    last_velocity + velocity_step,
+                )
+                velocity = _clamp(velocity, -max_speed, max_speed)
+                position = last_target + velocity * dt
+
+            limited_target.append(position)
+            limited_velocity.append(velocity)
+
+        if not np.isfinite(limited_target).all():
+            raise ValueError("joint command contains NaN/Inf")
+        return limited_target, limited_velocity
+```
+
+不要增加 ROS 参数或辅助类。制动距离直接使用当前方法已有的
+`max_acceleration`、`max_speed` 和 `dt`。
+
+- [x] **步骤 2：运行新增测试并确认通过**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+PYTHONPATH="/home/robot/WS_xr/src/xr_rm_teleop:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test/test_joint_control.py::test_joint_command_step_brakes_before_fixed_target_without_overshoot \
+  -v
+```
+
+预期：`PASS`。
+
+- [x] **步骤 3：运行关节控制测试文件**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+PYTHONPATH="/home/robot/WS_xr/src/xr_rm_teleop:${PYTHONPATH:-}" \
+  python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py -v
+```
+
+预期：全部通过；现有
+`test_joint_command_step_limits_acceleration_from_rest` 继续通过，证明首周期
+加速度行为没有回归。
+
+### 任务三：完整验证
+
+**文件：**
+
+- 不修改文件。
+
+- [x] **步骤 1：运行 `xr_rm_teleop` 全部测试**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+export RM75_TEST_PYTHONPATH="/home/robot/WS_xr/src/xr_rm_teleop:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages/cmeel.prefix/lib/python3.10/site-packages"
+PYTHONPATH="${RM75_TEST_PYTHONPATH}:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test -v
+```
+
+预期：全部测试通过，无失败；真实 Placo 回归测试必须执行，不能因缺少模块而跳过。
+
+- [x] **步骤 2：单独运行姿态控制测试**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+PYTHONPATH="/home/robot/WS_xr/src/xr_rm_teleop:${PYTHONPATH:-}" \
+  python3 -m pytest \
+  src/xr_rm_teleop/test/test_orientation_control.py -v
+```
+
+预期：全部通过。
+
+- [x] **步骤 3：构建工作空间**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+colcon build --symlink-install
+```
+
+预期：`xr_rm_input`、`xr_rm_interfaces`、`xr_rm_teleop` 和 `xr_rm_bringup`
+全部构建成功。
+
+- [x] **步骤 4：使用 mock 启动右臂统一 launch**
+
+运行：
+
+```bash
+source /opt/ros/humble/setup.bash
+source install/setup.bash
+timeout 10s ros2 launch xr_rm_bringup arm_debug.launch.py \
+  arm:=right use_mock:=true
+```
+
+预期：
+
+- 节点日志显示控制周期约 `dt=0.0111s`；
+- 日志显示 `follow=False`；
+- 不连接厂商 SDK，不发送真实 CANFD；
+- 除 `timeout` 主动结束产生的退出状态外，没有 Python 异常或 ROS 错误。
+
+- [x] **步骤 5：检查最终差异**
+
+运行：
+
+```bash
+git -C /home/robot/WS_xr/src diff --check
+git -C /home/robot/WS_xr/src status --short
+git -C /home/robot/WS_xr/src diff -- \
+  xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py \
+  xr_rm_teleop/test/test_joint_control.py
+```
+
+预期：
+
+- `diff --check` 无输出；
+- 本次运行时代码改动只涉及上述两个文件；
+- 原有工作区修改仍保留；
+- 不存在提交或远程推送。
+
+## 用户真机验证边界
+
+自动验证完成后，只提供手动验证步骤，不由 Codex 操作真机：
+
+1. 保持低跟随，从安全姿态和小于 5 mm 的上下位移开始；
+2. 手柄停止后观察机械臂是否立即减振并稳定；
+3. 确认无持续 QP、UDP、CANFD 或故障锁存错误后，再测试 10 mm；
+4. 若不再振荡但仍有不可接受的整臂大幅构型变化，停止扩大位移，转入独立的奇异点
+   处理设计。

docs/superpowers/plans/2026-07-30-rm75-qp-udp-feedback-fix.md

@@ -0,0 +1,352 @@
+# RM75 QP 与 UDP 反馈周期修复实施计划
+
+> **供代理执行：** 必须使用 `superpowers:subagent-driven-development`（推荐）或
+> `superpowers:executing-plans` 子技能，按任务逐项实施。步骤使用复选框
+>（`- [ ]`）跟踪。
+
+**目标：** 将 QP 位置收敛阈值调整为 2 mm，并把项目配置的毫秒周期正确换算为
+睿尔曼 SDK 的 5 ms 周期单位，从根因上降低误触发 UDP 反馈超时的概率。
+
+**架构：** 保持 ROS 参数和 YAML 中 `realtime_push_cycle_ms` 的毫秒语义，仅在
+`RealManAdapter.connect()` 的 SDK 边界执行单位换算。QP 只调整现有位置收敛常量；
+反馈超时状态机保持不变，只在首次超时日志中增加实际反馈年龄。
+
+**技术栈：** Python 3.10、ROS2 Humble、pytest、ament/colcon、睿尔曼 Python
+API2、Placo 0.9.4。
+
+---
+
+## 文件范围
+
+- 修改 `xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py`：QP 位置收敛阈值。
+- 修改 `xr_rm_teleop/test/test_placo_transforms.py`：1.5 mm 近收敛结果测试。
+- 修改 `xr_rm_teleop/xr_rm_teleop/realman_adapter.py`：毫秒到 SDK 周期单位换算。
+- 修改 `xr_rm_teleop/test/test_initial_joint_pose.py`：5 ms、10 ms 换算测试。
+- 修改 `xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`：超时日志增加年龄。
+- 修改 `xr_rm_teleop/test/test_joint_control.py`：超时年龄日志测试。
+- 不修改 YAML、launch、UI、消息定义或依赖。
+
+## 测试环境
+
+所有命令从工作空间根目录 `/home/robot/WS_xr` 执行：
+
+```bash
+source /opt/ros/humble/setup.bash
+export PYTHONPATH=/home/robot/WS_xr/src/xr_rm_teleop:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages:/home/robot/miniconda3/envs/xr/lib/python3.10/site-packages/cmeel.prefix/lib/python3.10/site-packages:${PYTHONPATH}
+```
+
+### 任务 1：接受 2 mm 内的 QP 位置残差
+
+**文件：**
+
+- 修改：`xr_rm_teleop/test/test_placo_transforms.py`
+- 修改：`xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py:14`
+
+- [ ] **步骤 1：编写 1.5 mm 近收敛结果的失败测试**
+
+在 `test_placo_transforms.py` 的 QP 测试附近增加：
+
+```python
+def test_qp_solve_accepts_position_error_within_two_millimeters() -> None:
+    solver = object.__new__(PlacoIkSolver)
+    solver._actual_joints = np.zeros(7)
+    solver._robot = SimpleNamespace(
+        state=SimpleNamespace(q=np.zeros(14))
+    )
+    solver._frame_task = SimpleNamespace(T_world_frame=None)
+    solver._target_errors = lambda: (1.5e-3, 0.0)
+
+    result = solver.solve(np.eye(4))
+
+    assert result == pytest.approx([0.0] * 7)
+```
+
+并在文件顶部加入现有标准库类型：
+
+```python
+from types import SimpleNamespace
+```
+
+同时从 `placo_ik_solver` 导入现有收敛常量：
+
+```python
+from xr_rm_teleop.placo_ik_solver import (
+    QP_POSITION_TOLERANCE_M,
+    PlacoIkSolver,
+    _validated_transform,
+)
+```
+
+将真实 Placo 可达目标测试的位置断言改为引用同一收敛常量：
+
+```python
+assert position_error <= QP_POSITION_TOLERANCE_M
+```
+
+再增加超过 2 mm 时仍拒绝结果的边界测试：
+
+```python
+def test_qp_solve_rejects_position_error_above_two_millimeters() -> None:
+    solver = object.__new__(PlacoIkSolver)
+    solver._actual_joints = np.zeros(7)
+    solver._robot = SimpleNamespace(
+        state=SimpleNamespace(q=np.zeros(14)),
+        update_kinematics=lambda: None,
+    )
+    solver._frame_task = SimpleNamespace(T_world_frame=None)
+    solver._solver = SimpleNamespace(solve=lambda update: None)
+    solver._validate_result = lambda result, previous: None
+    solver._target_errors = lambda: (2.1e-3, 0.0)
+
+    with pytest.raises(RuntimeError, match="QP did not converge after 30"):
+        solver.solve(np.eye(4))
+```
+
+- [ ] **步骤 2：运行测试并确认当前实现失败**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_placo_transforms.py::test_qp_solve_accepts_position_error_within_two_millimeters -q
+```
+
+预期：失败；当前 1 mm 阈值不会直接接收 1.5 mm 残差，测试对象缺少后续 QP
+求解器。
+
+- [ ] **步骤 3：最小修改 QP 位置阈值**
+
+在 `placo_ik_solver.py` 修改现有常量：
+
+```python
+QP_POSITION_TOLERANCE_M = 2e-3
+```
+
+保留 `QP_MAX_ITERATIONS = 30`、姿态阈值、有限值检查、关节位置和速度限制不变。
+
+- [ ] **步骤 4：运行局部测试并确认通过**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_placo_transforms.py -q
+```
+
+预期：该文件全部测试通过；1.5 mm 残差被接受，2.1 mm 残差仍在 30 次后被拒绝。
+
+- [ ] **步骤 5：提交 QP 修改**
+
+```bash
+git add src/xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py src/xr_rm_teleop/test/test_placo_transforms.py
+git commit -m "fix: 放宽 RM75 QP 位置收敛阈值"
+```
+
+### 任务 2：修正睿尔曼 UDP 实时上报周期单位
+
+**文件：**
+
+- 修改：`xr_rm_teleop/test/test_initial_joint_pose.py`
+- 修改：`xr_rm_teleop/xr_rm_teleop/realman_adapter.py:190-196`
+
+- [ ] **步骤 1：将现有连接测试改为周期换算参数化测试**
+
+将
+`test_connect_configures_udp_feedback_and_waits_for_first_frame` 改为：
+
+```python
+@pytest.mark.parametrize(
+    ("cycle_ms", "sdk_cycle"),
+    [(5, 1), (10, 2)],
+)
+def test_connect_converts_udp_feedback_cycle_to_sdk_units(
+    monkeypatch,
+    cycle_ms,
+    sdk_cycle,
+) -> None:
+    fake_sdk = _install_fake_sdk(monkeypatch)
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "192.168.192.148",
+        8090,
+        realtime_push_cycle_ms=cycle_ms,
+        configure_safety_limits=False,
+    )
+
+    adapter.connect()
+
+    arm = fake_sdk.RoboticArm.instance
+    assert arm is not None
+    assert arm.config.args == (
+        sdk_cycle,
+        True,
+        8090,
+        0,
+        "192.168.192.148",
+    )
+    assert arm.callback is adapter._realtime_callback
+    assert adapter.get_latest_joint_state() is not None
+    assert not hasattr(adapter, "_feedback_thread")
+```
+
+- [ ] **步骤 2：运行参数化测试并确认当前实现失败**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_initial_joint_pose.py::test_connect_converts_udp_feedback_cycle_to_sdk_units -q
+```
+
+预期：两个用例均失败；当前代码把 5 和 10 直接传给 SDK，而预期分别为 1 和 2。
+
+- [ ] **步骤 3：在 SDK 边界执行单位换算**
+
+在 `RealManAdapter.connect()` 创建 `rm_realtime_push_config_t` 时修改第一个参数：
+
+```python
+config = rm_realtime_push_config_t(
+    self._realtime_push_cycle_ms // 5,
+    True,
+    self._realtime_push_port,
+    0,
+    self._realtime_push_host_ip,
+)
+```
+
+构造函数已有“正的 5 ms 整数倍”校验，因此不新增辅助函数或重复校验。启动日志继续
+输出项目参数的真实毫秒值。
+
+- [ ] **步骤 4：运行适配器测试并确认通过**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_initial_joint_pose.py -q
+```
+
+预期：该文件全部测试通过；mock 测试仍不依赖厂商 SDK。
+
+- [ ] **步骤 5：提交周期换算修改**
+
+```bash
+git add src/xr_rm_teleop/xr_rm_teleop/realman_adapter.py src/xr_rm_teleop/test/test_initial_joint_pose.py
+git commit -m "fix: 修正 RM75 UDP 上报周期单位"
+```
+
+### 任务 3：在 UDP 超时日志中打印反馈年龄
+
+**文件：**
+
+- 修改：`xr_rm_teleop/test/test_joint_control.py`
+- 修改：`xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py:1036-1044`
+
+- [ ] **步骤 1：扩展现有短暂超时测试，捕获并断言日志**
+
+在
+`test_short_udp_timeout_repeats_last_limited_target_without_query` 中创建遥操对象后加入：
+
+```python
+warnings = []
+teleop.get_logger = lambda: SimpleNamespace(
+    warn=lambda message: warnings.append(message)
+)
+```
+
+在现有断言末尾加入：
+
+```python
+assert warnings == [
+    "right_rm75 UDP关节反馈超时（age=200.0 ms），保持最后安全目标。"
+]
+```
+
+- [ ] **步骤 2：运行测试并确认当前实现失败**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py::test_short_udp_timeout_repeats_last_limited_target_without_query -q
+```
+
+预期：失败；当前日志中没有 `age=200.0 ms`。
+
+- [ ] **步骤 3：给首次超时警告增加实际反馈年龄**
+
+在 `_handle_stale_joint_feedback()` 中仅修改现有警告：
+
+```python
+self.get_logger().warn(
+    f"{self._arm_name} UDP关节反馈超时"
+    f"（age={age * 1000.0:.1f} ms），保持最后安全目标。"
+)
+```
+
+不改变 `_joint_feedback_ready` 的一次性日志条件、最后安全目标重发、500 ms
+重新同步、Grip 重使能和故障锁存逻辑。
+
+- [ ] **步骤 4：运行关节控制测试并确认通过**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_joint_control.py -q
+```
+
+预期：该文件全部测试通过。
+
+- [ ] **步骤 5：提交诊断日志修改**
+
+```bash
+git add src/xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py src/xr_rm_teleop/test/test_joint_control.py
+git commit -m "fix: 补充 RM75 UDP 超时反馈年龄"
+```
+
+### 任务 4：完整回归与 mock 启动验证
+
+**文件：**
+
+- 不修改生产代码。
+
+- [ ] **步骤 1：运行 `xr_rm_teleop` 全部测试**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test -q
+```
+
+预期：全部通过，无失败或错误。
+
+- [ ] **步骤 2：单独运行姿态控制测试**
+
+```bash
+python3 -m pytest src/xr_rm_teleop/test/test_orientation_control.py -q
+```
+
+预期：全部通过。
+
+- [ ] **步骤 3：构建完整工作空间**
+
+```bash
+colcon build --symlink-install
+```
+
+预期：`xr_rm_interfaces`、`xr_rm_input`、`xr_rm_teleop`、`xr_rm_bringup`
+全部构建成功。
+
+- [ ] **步骤 4：使用右臂 mock 启动统一 launch**
+
+```bash
+timeout 15s ros2 launch xr_rm_bringup arm_debug.launch.py arm:=right use_mock:=true
+```
+
+预期：节点正常启动，无 Python 异常或厂商 SDK 导入错误；因 `timeout` 主动终止，
+命令退出码可以是 124。禁止将 `use_mock` 改为 `false`。
+
+- [ ] **步骤 5：检查最终差异**
+
+```bash
+git status --short
+git diff HEAD~3 --check
+git diff HEAD~3 --stat
+```
+
+预期：除本实施计划文档外，代码差异仅包含计划列出的 6 个代码/测试文件；无空白
+错误，不包含 YAML、launch、UI、依赖或用户其他改动。
+
+## 真机验收
+
+自动验证完成后，由用户在安全条件下启动真机右臂模式并观察：
+
+1. `feedback_interval mean` 从约 25 ms 降到接近 5 ms；
+2. 正常遥操不再频繁触发 `UDP关节反馈超时`；
+3. 反馈确实超过 120 ms 时，日志显示实际 `age`，且仍保持最后安全目标；
+4. 位置残差小于 2 mm 时不再出现 QP 未收敛警告；
+5. 若仍频繁出现超过 120 ms 的中断，保留当前阈值，依据日志继续排查网络、SDK
+   回调或控制器固件。

docs/superpowers/specs/2026-07-29-rm75-feedback-recovery-design.md

@@ -0,0 +1,151 @@
+# RM75 关节反馈与故障恢复设计
+
+## 目标
+
+将当前 RM75 QP 遥操作链路调整为：
+
+- 启动时使用 `rm_get_joint_degree()` 获取实际关节角并初始化 QP。
+- 运行时只把 RealMan UDP `joint_position` 作为连续实际关节反馈。
+- UDP 短暂超时时保持最后一次已限速的安全关节目标，不生成新运动。
+- UDP 持续超时或 CANFD 发送错误时，按明确的同步、停止和人工恢复流程处理。
+- `rm_movej_canfd()` 保持低跟随，控制频率使用 `xr_rm_teleop` 分支的 90 Hz。
+
+不改变现有工作空间/圆柱限位、TCP与关节速度和加速度限制、XR命令超时、安全停止、外设控制及双臂节点名。
+
+## 控制数据源
+
+启动初始化与运行反馈使用不同的数据源：
+
+1. `RealManAdapter` 建立现有唯一厂商连接。
+2. 节点同步调用一次 `rm_get_joint_degree()`。
+3. 查询成功后把7个角度转换为弧度，用于初始化 Placo QP、最后安全目标和关节限速历史。
+4. 查询失败时关闭适配器、打印错误并使节点启动失败，不发送 CANFD。
+5. 正常运行后，QP 的连续实际状态只来自已校验且运动状态正常的 UDP `joint_position`。
+
+同步查询只用于启动、持续反馈超时恢复和 CANFD 错误恢复，不新增连接，不进行常态轮询。
+
+## 状态与转换
+
+### 正常运行
+
+控制定时器以 90 Hz 执行。每个周期读取最新 UDP 关节快照，同步 QP，执行现有目标生成、安全限位、单步 QP、关节速度/加速度限制，然后调用：
+
+```text
+rm_movej_canfd(target_degrees, follow=false, ...)
+```
+
+只有已经通过关节限速并成功发送的目标才能成为“最后安全目标”。
+
+### UDP 短暂超时
+
+UDP 快照年龄超过现有 `command_timeout_sec=0.12` 秒、但未达到 `feedback_resync_timeout_sec=0.5` 秒时：
+
+- 不使用过期反馈同步 QP。
+- 不运行目标生成和 QP。
+- 不更新任何目标、滤波器或限速历史。
+- 若超时前正在遥操作且已有成功发送的安全目标，以 90 Hz 原样重发该目标。
+- 若超时前未在遥操作或没有成功发送的目标，保持停止，不开始 CANFD 输出。
+- 首次进入时打印节流后的警告。
+
+机械臂报警、关节掉使能、关节错误或非有限关节值不是普通超时，仍立即执行安全停止。
+
+### UDP 持续超时
+
+UDP 快照年龄达到 0.5 秒时，每次中断只同步调用一次 `rm_get_joint_degree()`：
+
+- 查询成功：用实际角度重置 QP、最后安全目标和关节限速历史；不生成新运动，继续保持并等待 UDP 恢复。
+- 查询失败：调用 slow-stop，停止 CANFD，进入锁存故障并打印错误。
+
+同一次中断不会反复查询。收到新的有效 UDP 反馈后，查询标志才复位。
+
+### UDP 恢复
+
+UDP 恢复后先持续同步实际关节状态，但不能直接恢复运动：
+
+1. 当前 Grip 必须松开。
+2. 节点清除重新使能要求。
+3. 操作者再次按下 Grip，节点以新的手柄和机械臂实际位姿建立相对控制起点。
+
+### CANFD 错误
+
+`rm_movej_canfd()` 返回错误或抛出异常时：
+
+1. 立即停止后续 CANFD 发送。
+2. 调用 slow-stop。
+3. 打印包含机械臂名称、命令名称和原始错误的终端错误日志。
+4. 调用 `rm_get_joint_degree()` 查询实际关节角。
+5. 查询成功时重置 QP和关节命令历史，但不再发送保持命令；等待有效 UDP 和 Grip 松开后重新按下。
+6. 查询失败时进入锁存故障并打印查询错误。
+
+### 锁存故障
+
+锁存故障只作用于发生错误的机械臂节点：
+
+- 控制定时器不再查询、运行 QP或发送 CANFD。
+- slow-stop 只发送一次。
+- 后续 UDP 恢复或 Grip 操作不能自动解锁。
+- 终端保留明确错误信息，但不在每个周期重复刷屏。
+- 操作者检查后必须重启对应遥操作节点才能恢复。
+
+## 代码边界
+
+### `xr_rm_teleop/xr_rm_teleop/realman_adapter.py`
+
+- 给真实与 mock 适配器增加同步关节角查询能力。
+- 复用现有厂商连接。
+- 校验返回码、数量和有限值，统一返回弧度。
+- 保留 UDP 回调作为运行时快照来源。
+
+### `xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`
+
+- 启动时查询并初始化 QP。
+- 增加 `feedback_resync_timeout_sec` 参数，默认 0.5 秒。
+- 校验 `feedback_resync_timeout_sec > command_timeout_sec > 0`。
+- 在现有控制周期内加入保持、一次性重新同步、等待 Grip 重使能和锁存判断。
+- 复用现有 `_safe_stop()`、Grip 重使能和关节限速逻辑，不新增状态机类。
+
+### 配置
+
+以下配置的 `control_rate_hz` 从 125 Hz 改为 90 Hz，并增加相同的 0.5 秒持续超时参数：
+
+- `xr_rm_bringup/config/dual_arm_rm75.yaml`
+- `xr_rm_bringup/config/left_arm_rm75.yaml`
+- `xr_rm_bringup/config/right_arm_rm75.yaml`
+
+三份配置继续使用 `follow: false`，双臂节点名保持 `left_arm_teleop` 和 `right_arm_teleop`。
+
+## 错误日志
+
+以下转换必须写入 ROS2 终端日志：
+
+- 启动关节查询失败：`error`。
+- 首次进入 UDP 短暂超时：`warn`。
+- 持续超时查询开始及成功：`warn`/`info`。
+- 持续超时查询失败并锁存：`error`。
+- CANFD 发送失败：`error`。
+- CANFD 后关节查询失败并锁存：`error`。
+- UDP 恢复并等待 Grip 人工重使能：`info`。
+
+日志包含机械臂名称和失败阶段；周期性路径使用状态转换或节流避免刷屏。
+
+## 测试与验证
+
+使用现有 mock 和单元测试完成，不连接真机：
+
+1. 适配器正确解析 `rm_get_joint_degree()` 成功结果，并拒绝错误码、错误数量和 NaN/Inf。
+2. 启动查询结果初始化 QP 和安全目标；查询失败时节点不能进入控制。
+3. 0.12～0.5 秒反馈超时期间不调用 QP，只重发同一安全目标。
+4. 0.5 秒持续超时只查询一次；成功后等待 UDP 与 Grip，失败后锁存。
+5. CANFD 错误后停止发送并查询；查询成功要求 Grip 重使能，查询失败锁存。
+6. 机械臂报警或掉使能仍立即停止，不能进入保持路径。
+7. 三份配置均使用 90 Hz、0.5 秒持续超时和低跟随。
+
+在工作空间根目录 `/home/robot/WS_xr` 执行：
+
+```bash
+source /opt/ros/humble/setup.bash
+pytest src/xr_rm_teleop/test/test_initial_joint_pose.py
+pytest src/xr_rm_teleop/test/test_joint_control.py
+pytest src/xr_rm_teleop/test/test_orientation_control.py
+colcon build --symlink-install
+```

docs/superpowers/specs/2026-07-29-rm75-qp-convergence-design.md

@@ -0,0 +1,192 @@
+# RM75 QP 收敛与低跟随稳定性优化设计
+
+## 背景
+
+右臂真机以90 Hz、`follow: false`运行时，用户观察到：
+
+- 手柄移动约10 cm后，`target_pose`很快稳定；
+- `current_pose`仍需约3秒缓慢追赶；
+- 运动过程中机械臂存在肉眼可见的轻微晃动。
+
+现场 timing 日志同时表明：
+
+- 控制周期约11.111 ms；
+- 控制回调平均约2.6 ms，最大约6.0 ms；
+- QP平均约0.39 ms；
+- CANFD发送平均约0.21 ms；
+- UDP实际关节反馈平均约25 ms一帧，即约40 Hz。
+
+因此，控制线程、QP单次计算和CANFD调用本身没有耗尽90 Hz周期；慢速发生在
+`target_pose`生成之后。
+
+## 根因
+
+当前 `PlacoIkSolver.solve()` 每次只调用一次：
+
+```python
+self._solver.solve(True)
+```
+
+该调用把一次QP增量应用为 `q + Δq`。与此同时，90 Hz控制循环每次都会先用
+最新实际关节反馈重置Placo模型。由于实际反馈约40 Hz，同一帧反馈通常会被重复
+使用两到三次。
+
+结果是每次下发的关节目标只位于实际关节角前方一小步，而不是当前TCP目标对应的
+收敛关节解。低跟随控制器持续追逐这个短距离移动点，表现为：
+
+- 对稳定TCP目标呈缓慢的渐近追赶；
+- 实际反馈每约25 ms更新一次时，关节目标随反馈发生台阶式修正；
+- 低跟随内部平滑与台阶式关节目标叠加，形成轻微晃动。
+
+本地RM75模型对照结果支持该判断：从右臂初始姿态求解7 cm平移目标时，单次QP
+只产生约7.6 mm TCP位移；在同一次逆解中连续迭代30次后，目标误差可降至接近
+零，计算耗时约3.56 ms。
+
+## 目标
+
+保持现有安全基线并实现：
+
+- 手柄移动10 cm后，机械臂约1秒内稳定到位；
+- 运动和到位后无持续肉眼可见晃动；
+- 控制频率保持90 Hz；
+- `rm_movej_canfd()`保持低跟随；
+- 运行时仍以UDP `joint_position`作为实际关节反馈；
+- 保留工作空间、圆柱、TCP速度、姿态速度、关节速度与关节加速度限制；
+- 保留反馈超时、CANFD错误恢复、Grip重新使能和安全停止逻辑。
+
+## 不在本次范围
+
+- 不启用高跟随；
+- 不提高TCP或关节安全上限；
+- 不修改XR手柄滤波和坐标映射；
+- 不修改UDP反馈周期或增加反馈预测器；
+- 不新增线程、RealMan连接、依赖或状态机；
+- 不处理双臂碰撞检测。
+
+## 方案比较
+
+### 方案一：有限次数迭代QP
+
+每个控制周期仍从实际关节角开始，但在一次 `solve()` 调用内部迭代QP，直到TCP
+目标收敛或达到固定迭代上限。得到的完整关节目标继续经过现有关节速度与加速度
+限幅后才发送。
+
+优点：
+
+- 直接修复单步QP只生成近距离移动点的根因；
+- 不需要预测状态，不会在反馈中断时继续外推；
+- 复用现有限速、错误回退和CANFD发送路径；
+- 本地测量表明计算量可放入90 Hz周期。
+
+缺点：
+
+- 单周期QP耗时会高于当前单步求解；
+- 不可达目标需要明确的未收敛处理。
+
+### 方案二：反馈帧之间维护预测关节状态
+
+仅在新UDP反馈到达时校正模型，其余90 Hz周期从上一条关节命令继续积分QP。
+
+优点：
+
+- 每周期仍只求解一次QP；
+- 可避免同一反馈帧反复重置模型。
+
+缺点：
+
+- 引入预测状态、反馈校正和漂移处理；
+- 反馈与预测偏差可能在校正时产生新的关节跳动；
+- 超时与恢复逻辑需要同时管理实际状态和预测状态。
+
+### 方案三：只调整滤波、速度或高跟随参数
+
+`target_pose`已经快速稳定，继续提高 `max_linear_speed` 或减小目标滤波不能解决
+下游渐近追赶。启用高跟随则违反本次低跟随约束。
+
+## 决策
+
+采用方案一。它在不引入预测状态的情况下直接修复根因，改动范围只涉及Placo
+求解器及其测试。
+
+## 控制数据流
+
+正常运行时的数据流调整为：
+
+```text
+UDP实际关节反馈
+→ 更新Placo实际关节状态和current_pose
+→ 现有XR相对位姿、工作空间、圆柱、滤波和TCP限速
+→ 有限次数迭代QP，得到收敛关节目标
+→ 现有关节速度与加速度限幅
+→ rm_movej_canfd(..., follow=false)
+```
+
+反馈短暂超时仍只以90 Hz重发最后一次已通过限速的关节目标，不运行QP。反馈持续
+超时和CANFD错误仍沿用现有同步、停止与故障锁存逻辑。
+
+## QP迭代规则
+
+`PlacoIkSolver.solve()`按以下规则执行：
+
+1. 校验目标变换。
+2. 记录本次内部迭代前的关节状态。
+3. 调用一次 `self._solver.solve(True)`。
+4. 更新Placo运动学。
+5. 校验本次候选关节状态：
+   - 7个有限数值；
+   - 不违反RM75关节位置限制；
+   - 本次数值迭代步长不超过Placo按 `dt=1/90` 应用的URDF关节速度限制。
+6. 使用位置任务与姿态任务的 `error_norm()`检查收敛：
+   - 位置误差不超过1 mm；
+   - 姿态误差不超过0.005 rad。
+7. 未收敛则继续迭代，最多30次。
+
+30次后仍未收敛，或任一迭代产生非法结果时，抛出异常。节点复用现有
+`_solve_joint_target()`错误路径，在终端限频打印QP失败原因，并保持上一组安全
+关节目标。
+
+内部迭代得到的是逆解目标，不会直接绕过发送限速。最终下发仍必须经过
+`_limit_joint_command_step()`，因此每个90 Hz真实命令继续满足现有
+`joint_max_speed` 和 `joint_max_acc`。
+
+## 晃动抑制
+
+本次不再叠加新的低通滤波器。晃动通过两层现有机制抑制：
+
+1. QP先收敛到当前TCP目标对应的关节解，避免关节目标随40 Hz反馈只前进一小步；
+2. 最终关节目标由现有关节速度与加速度限幅器生成连续90 Hz命令。
+
+若真机仍存在晃动，再根据“目标关节角与实际关节角误差”追加诊断；本次不预先
+引入预测器或额外滤波。
+
+## 性能与安全验收
+
+自动验证：
+
+- 7 cm可达TCP平移目标在一次 `solve()` 后位置误差不超过1 mm；
+- 姿态误差满足0.005 rad阈值；
+- 非法结果和未收敛目标继续触发现有安全回退；
+- 关节命令速度与加速度限幅测试继续通过；
+- `xr_rm_teleop`全部pytest通过；
+- `colcon build --symlink-install`通过；
+- `arm_debug.launch.py arm:=right use_mock:=true`正常启动。
+
+真机由用户验证：
+
+- 手柄快速移动10 cm并保持不动，机械臂约1秒内稳定；
+- 无持续肉眼可见晃动；
+- 连续四个5秒 timing 窗口中 `total` 最大值低于11.111 ms；
+- 无QP失败、反馈超时、CANFD错误或故障锁存日志；
+- 松开Grip后仍立即退出遥操作并执行安全停止。
+
+若单周期 `total` 达到或超过11.111 ms，停止真机运动并降低最大QP迭代次数，
+不得通过提高控制频率或关闭安全检查规避计算超时。
+
+## 文件范围
+
+- 修改 `xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py`。
+- 修改 `xr_rm_teleop/test/test_placo_transforms.py`。
+- 如现有QP失败测试需要补充未收敛原因断言，只精确修改
+  `xr_rm_teleop/test/test_joint_control.py`。
+
+不修改三份机械臂YAML、RealMan适配器、launch、UI、依赖或公开入口。

docs/superpowers/specs/2026-07-30-rm75-joint-command-braking-design.md

@@ -0,0 +1,99 @@
+# RM75 关节命令提前制动设计
+
+## 背景
+
+右臂真机保持 90 Hz 和 `follow: false`。有界迭代 QP 提高跟随速度后，手柄上下
+移动约 10 mm 时出现整臂剧烈晃动，并且到达目标后仍持续振荡。
+
+现场日志表明控制计算未超时，但存在少量 QP 未收敛警告。离线检查确认：
+
+- 手柄上下移动按现有映射对应机器人 X 方向；
+- 当前初始姿态的关节雅可比条件数约为 116，该方向的逆解对关节运动较敏感；
+- 机器人 X 方向 10 mm 的收敛逆解可能包含最大约 17° 的关节变化；
+- 现有关节命令限幅器只限制速度和加速度，没有根据剩余距离提前制动。
+
+固定 10° 关节目标的离线复现中，现有限幅器运行 3 秒后仍处于约 10.56°、
+-20°/s，证明稳定目标本身也会被反复越过。这与真机“到位后继续晃动”的现象
+一致。
+
+## 目标
+
+- 关节目标稳定后，90 Hz 关节命令提前减速并停止在目标上；
+- 不再因命令限幅器反复越过目标而持续振荡；
+- 保留当前较快的有界迭代 QP；
+- 保留现有关节最大速度和最大加速度限制；
+- 保持 `rm_movej_canfd(..., follow=false)`；
+- 不改变 UDP 反馈、超时保持、CANFD 错误恢复和安全停止行为。
+
+## 不在本次范围
+
+- 不修改 QP 迭代次数、收敛阈值或失败回退；
+- 不修改初始姿态、XR 坐标映射或姿态控制；
+- 不增加奇异点阻尼、预测器、新线程、新依赖或新 ROS 参数；
+- 不修改三份机械臂 YAML、RealMan 适配器、launch 或 UI；
+- 不连接真机，不由 Codex 发送运动命令。
+
+若修复制动后机械臂运动已经平稳，但上下运动仍伴随不可接受的整臂大幅构型变化，
+再单独设计奇异点处理；本次不把两个问题混在同一改动中。
+
+## 方案
+
+只修改 `SingleArmVelocityTeleop._limit_joint_command_step()`。
+
+现有逻辑在目标仍位于运动方向前方时持续加速，只有到达或越过目标后才开始反向
+减速。新逻辑对每个关节使用同一组现有状态：
+
+- 上一次已发送的关节目标；
+- 上一次关节命令速度；
+- 当前 QP 关节目标；
+- 现有关节最大速度、最大加速度和控制周期。
+
+每周期按以下规则生成命令：
+
+1. 计算关节剩余距离和单周期最大速度变化
+   `velocity_step = max_acceleration * dt`。
+2. 根据 90 Hz 离散积分规则，计算“本周期再加速一次、随后以最大允许减速度制动”
+   所需的总距离。
+3. 若关节正在朝目标运动，并且剩余距离已经不大于该制动距离，则本周期开始减速；
+   否则继续朝目标加速，但不超过现有最大速度。
+4. 使用 `velocity_step` 限制本周期速度变化，保持现有加速度上限。
+5. 使用新速度积分得到本周期关节目标。
+6. 当关节已经停下且剩余距离不超过一个最大加速度位移
+   `max_acceleration * dt²` 时，在不违反单周期加速度限制的前提下精确落到目标，
+   避免离散步长形成极小往复振荡。
+
+该逻辑只负责命令轨迹制动，不改变 QP 输出。对于持续移动的目标，若目标突然越过
+当前命令位置，控制器仍优先遵守加速度限制，以最大允许减速度反向；不会为禁止
+瞬时越界而跳变速度。
+
+## 安全行为
+
+- 每周期命令速度绝对值不超过 `joint_max_speed`；
+- 相邻周期间速度变化不超过 `joint_max_acc * dt`；
+- 输出继续校验 NaN 和 Inf；
+- QP 异常仍保持上一组安全目标并在终端限频打印警告；
+- UDP 短暂超时仍以 90 Hz 重发最后一次已限速目标，不运行 QP；
+- UDP 持续超时、CANFD 错误、Grip 重新使能和安全停止逻辑保持不变。
+
+## 测试
+
+先增加失败测试，再修改实现：
+
+1. 对固定 10° 关节目标连续运行限幅器，验证命令不越过目标并最终停止；
+2. 在整个序列中验证速度不超过现有上限；
+3. 验证相邻命令速度变化不超过现有加速度上限；
+4. 保留现有“从静止开始限制首周期加速度”测试；
+5. 运行 `xr_rm_teleop` 全部测试；
+6. 按项目规则运行 `colcon build --symlink-install`；
+7. 使用 `arm_debug.launch.py arm:=right use_mock:=true`验证 90 Hz、低跟随和启动路径。
+
+真机只由用户分阶段验证：先小位移、低风险姿态，确认不再到位后持续振荡，再逐步
+增加位移。若仍有明显整臂构型变化但不再振荡，应停止扩大位移并转入奇异点处理，
+不得通过提高速度、加速度或启用高跟随规避。
+
+## 文件范围
+
+- 修改 `xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`；
+- 修改 `xr_rm_teleop/test/test_joint_control.py`；
+- 新增本中文设计文档；
+- 后续新增一份中文实施计划。

docs/superpowers/specs/2026-07-30-rm75-qp-udp-feedback-fix-design.md

@@ -0,0 +1,176 @@
+# RM75 QP 阈值与 UDP 反馈周期修复设计
+
+## 背景
+
+右臂真机在 90 Hz、`follow: false` 遥操过程中频繁出现两类警告：
+
+```text
+QP did not converge after 30 iterations:
+position_error=0.001245~0.001609 m
+
+UDP关节反馈超时，保持最后安全目标。
+```
+
+现场 timing 日志同时表明：
+
+- 控制回调最大约 5.1 ms，没有耗尽 11.111 ms 周期；
+- UDP 反馈间隔均值约 25 ms，即实际约 40 Hz；
+- UDP 间隔存在 37~71 ms 的明显抖动；
+- UDP 短暂超时后约 44 ms 收到新帧，但现有安全状态机要求先松开 Grip。
+
+因此，控制线程计算量不是这两类警告的原因。
+
+## 根因
+
+### QP 近阈值失败
+
+当前 QP 最多迭代 30 次，并要求：
+
+```text
+位置误差 <= 1 mm
+姿态误差 <= 0.005 rad
+```
+
+现场失败时姿态误差约 0.0007 rad，已经满足要求；位置误差仅比 1 mm 高
+0.245~0.609 mm。30 次迭代中的每一步已经通过关节有限值、关节位置限制和单步
+速度限制校验，但最终结果仍因严格的 1 mm 判定被整体丢弃。
+
+用户已明确确认 2 mm 位置残差可接受；该数值也与现有 1 mm 手柄位置死区处于
+同一量级。将位置收敛阈值改为 2 mm，可以接收现场这类安全的近收敛
+结果，同时继续拒绝此前出现过的 7.5 mm 等明显未收敛结果。
+
+### UDP 周期单位错误
+
+项目参数 `realtime_push_cycle_ms` 的单位是毫秒，三份机械臂配置均填写 `5`。
+当前适配器把这个值原样传给：
+
+```python
+rm_realtime_push_config_t(cycle, ...)
+```
+
+但睿尔曼 SDK 的 `cycle` 单位不是毫秒，而是 5 ms 的倍数。因此：
+
+```text
+当前传入 cycle=5
+实际周期 = 5 × 5 ms = 25 ms
+实际频率 = 40 Hz
+```
+
+这与现场 `feedback_interval mean≈25 ms` 完全一致。期望 5 ms 上报时，SDK
+参数应为 `cycle=1`。
+
+在 120 ms 反馈超时窗口内，25 ms 上报只有约 5 次发送机会；修正为 5 ms 后有
+约 24 次发送机会，能显著提高对偶发丢包和调度抖动的容忍度。若网络或 SDK
+回调整体停顿超过 120 ms，仍应触发现有安全超时。
+
+## 目标
+
+- QP 位置收敛阈值由 1 mm 调整为 2 mm；
+- `realtime_push_cycle_ms: 5` 实际配置成 SDK `cycle=1`，恢复 5 ms 上报；
+- UDP 超时警告打印触发时的实际反馈年龄；
+- 保持控制频率 90 Hz 和 `follow: false`；
+- 保持 UDP 短超时保持、持续超时重同步、Grip 重使能、CANFD 恢复和故障锁存；
+- 保留工作空间、圆柱、TCP、姿态和关节安全限制。
+
+## 不在本次范围
+
+- 不增加 QP 最大迭代次数；
+- 不修改姿态收敛阈值；
+- 不放宽 `command_timeout_sec=0.12`；
+- 不修改 `feedback_resync_timeout_sec=0.5`；
+- 不允许短超时后自动恢复 Grip；
+- 不修改三份机械臂 YAML；
+- 不修改 UDP 目标 IP、端口、线程模式或增加新连接；
+- 不处理网卡、交换机或控制器固件问题；
+- 不修改用户当前未提交的 `right_arm_rm75.yaml` 参数调整。
+
+## 修改方案
+
+### QP 收敛
+
+在 `placo_ik_solver.py` 中只修改：
+
+```python
+QP_POSITION_TOLERANCE_M = 2e-3
+```
+
+30 次迭代、逐步关节安全校验、姿态阈值和未收敛异常格式保持不变。
+
+### UDP 周期换算
+
+保留公开参数 `realtime_push_cycle_ms` 的毫秒语义和“正数且为 5 ms 倍数”的现有
+校验。构造 SDK 配置时执行：
+
+```python
+sdk_cycle = self._realtime_push_cycle_ms // 5
+```
+
+示例：
+
+| 项目参数 | SDK `cycle` | 实际周期 |
+|---:|---:|---:|
+| 5 ms | 1 | 5 ms |
+| 10 ms | 2 | 10 ms |
+| 25 ms | 5 | 25 ms |
+
+启动日志仍打印毫秒值，避免把 SDK 内部单位暴露为用户配置。
+
+### UDP 超时日志
+
+首次进入短暂超时时打印：
+
+```text
+right_rm75 UDP关节反馈超时（age=xxx.x ms），保持最后安全目标。
+```
+
+日志只增加诊断值，不改变节流、保持目标、QP 停止和 Grip 重使能行为。
+
+## 数据流与安全
+
+正常路径：
+
+```text
+YAML 5 ms
+→ 适配器换算 SDK cycle=1
+→ 控制器约每 5 ms UDP 上报
+→ 回调校验并缓存 joint_position
+→ 90 Hz 控制读取最新实际反馈
+→ 最多 30 次 QP，位置阈值 2 mm
+→ 现有关节提前制动限幅
+→ rm_movej_canfd(..., follow=false)
+```
+
+反馈年龄超过 120 ms 时仍停止生成新目标和 QP，以 90 Hz 重发最后安全目标，并
+要求 Grip 松开后重新使能。达到 500 ms 时仍只尝试一次
+`rm_get_joint_degree()`；失败时仍停止并锁存故障。
+
+## 测试与验证
+
+自动测试：
+
+1. 增加 QP 1.5 mm 位置残差的行为测试，验证其被 2 mm 阈值接受；
+2. 保留明显未收敛结果抛出异常的行为；
+3. 修改 SDK 配置测试，验证项目 5 ms 参数传入 SDK 时为 `cycle=1`；
+4. 增加 10 ms 到 `cycle=2` 的换算覆盖；
+5. 验证 UDP 超时日志包含实际反馈年龄；
+6. 运行 `xr_rm_teleop` 全部测试和姿态控制测试；
+7. 运行 `colcon build --symlink-install`；
+8. 使用 `arm_debug.launch.py arm:=right use_mock:=true`验证启动路径。
+
+真机由用户验证：
+
+- 连续 timing 窗口中 `feedback_interval mean` 从约 25 ms 降到接近 5 ms；
+- 正常遥操不再频繁出现 UDP 超时；
+- 位置残差小于 2 mm 时不再出现 QP 未收敛警告；
+- 真正超过 120 ms 的反馈中断仍打印带 `age` 的警告并执行现有安全保持；
+- 若修正后仍频繁出现超过 120 ms 的中断，再依据 `age` 和间隔数据排查网络、
+  SDK 回调或控制器固件，不继续盲目放宽超时。
+
+## 文件范围
+
+- 修改 `xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py`；
+- 修改 `xr_rm_teleop/xr_rm_teleop/realman_adapter.py`；
+- 修改 `xr_rm_teleop/xr_rm_teleop/single_arm_velocity_teleop.py`；
+- 修改相关现有测试；
+- 新增本中文设计文档和后续中文实施计划；
+- 不修改 YAML、launch、UI 或依赖。

openspec/config.yaml

@@ -1,20 +0,0 @@
-schema: spec-driven
-
-# Project context (optional)
-# This is shown to AI when creating artifacts.
-# Add your tech stack, conventions, style guides, domain knowledge, etc.
-# Example:
-#   context: |
-#     Tech stack: TypeScript, React, Node.js
-#     We use conventional commits
-#     Domain: e-commerce platform
-
-# Per-artifact rules (optional)
-# Add custom rules for specific artifacts.
-# Example:
-#   rules:
-#     proposal:
-#       - Keep proposals under 500 words
-#       - Always include a "Non-goals" section
-#     tasks:
-#       - Break tasks into chunks of max 2 hours

xr_rm_bringup/config/dual_arm_rm75.yaml

@@ -1,10 +1,10 @@
 # 阶段一：PICO 遥操作双 RM75 平台配置。
 #
-# 当前控制方式是“相对位姿透传”：
+# 当前控制方式是“相对 TCP + 单步 QP”：
 # 按下 grip 时锁定当前手柄位姿和 TCP 位姿，之后将手柄相对位移和相对旋转
 # 映射为目标 TCP 位姿，经过工作空间限幅、目标低通、姿态低通和单帧步长
-# 限制后，通过 rm_movep_canfd 下发。cmd_vel 仅作为目标位姿变化率调试话题，
-# 不是机械臂执行命令。
+# 限制后，通过 Placo 单步 QP 和 rm_movej_canfd 下发 7 个关节目标。
+# cmd_vel 仅作为目标位姿变化率调试话题，不是机械臂执行命令。
 # 末端外设由 peripherals_rm75.yaml 配置，真机连接阶段初始化后由遥操作节点复用。
 
 left_arm_teleop:
@@ -13,6 +13,7 @@ left_arm_teleop:
     controller_topic: /xr/left_controller
     control_rate_hz: 90.0
     command_timeout_sec: 0.12
+    feedback_resync_timeout_sec: 0.5
 
     # 位姿目标生成与平滑参数。
     scale: 0.75
@@ -26,9 +27,7 @@ left_arm_teleop:
     enable_orientation_axes: [true, true, true]
     orientation_deadband_rad: 0.005
     orientation_filter_alpha: 0.65
-    max_orientation_speed: 0.6
-    current_pose_poll_hz: 10.0
-
+    max_orientation_speed: 0.5
     workspace_min: [-0.70, -0.60, 0.10]
     workspace_max: [0.70, 0.40, 0.70]
     cyl_radius_limit: [0.20, 0.60]
@@ -41,16 +40,20 @@ left_arm_teleop:
                          0.0, 0.0, 1.0,
                          -1.0, 0.0, 0.0]
 
-    use_mock: false
-    mock_initial_pose: [-0.2562, -0.2765, 0.1489, -3.0190, -0.1010, 3.1400]
     robot_ip: 192.168.192.18
     robot_port: 8080
+    realtime_push_host_ip: 192.168.192.148
+    realtime_push_port: 8089
+    realtime_push_cycle_ms: 5
     avoid_singularity: 0
-    frame_type: 1
     follow: false
     canfd_trajectory_mode: 2
     canfd_radio: 0
     configure_safety_limits: true
+    enable_tool_control: true
+    enable_trigger_gripper_control: true
+    trigger_close_threshold: 0.95
+    configure_peripheral_on_connect: true
     max_line_speed: 1.0
     max_angular_speed: 1.5
     max_line_acc: 1.0
@@ -68,6 +71,7 @@ right_arm_teleop:
     controller_topic: /xr/right_controller
     control_rate_hz: 90.0
     command_timeout_sec: 0.12
+    feedback_resync_timeout_sec: 0.5
 
     scale: 0.75
     deadband_m: 0.001
@@ -80,9 +84,7 @@ right_arm_teleop:
     enable_orientation_axes: [true, true, true]
     orientation_deadband_rad: 0.005
     orientation_filter_alpha: 0.65
-    max_orientation_speed: 0.6
-    current_pose_poll_hz: 10.0
-
+    max_orientation_speed: 0.5
     workspace_min: [-0.70, -0.60, 0.10]
     workspace_max: [0.70, 0.40, 0.70]
     cyl_radius_limit: [0.20, 0.60]
@@ -95,16 +97,20 @@ right_arm_teleop:
                          0.0, 0.0, 1.0,
                          1.0, 0.0, 0.0]
 
-    use_mock: false
-    mock_initial_pose: [0.2663, -0.2606, 0.1027, 3.0330, 0.0000, 1.0910]
     robot_ip: 192.168.192.19
     robot_port: 8080
+    realtime_push_host_ip: 192.168.192.148
+    realtime_push_port: 8090
+    realtime_push_cycle_ms: 5
     avoid_singularity: 1
-    frame_type: 1
     follow: false
     canfd_trajectory_mode: 2
     canfd_radio: 0
     configure_safety_limits: true
+    enable_tool_control: true
+    enable_trigger_gripper_control: true
+    trigger_close_threshold: 0.95
+    configure_peripheral_on_connect: true
     max_line_speed: 1.0
     max_angular_speed: 1.5
     max_line_acc: 1.0

xr_rm_bringup/config/left_arm_rm75.yaml

@@ -1,4 +1,4 @@
-# 左臂单独调试配置：XR 相对位姿透传控制 RM75 TCP。
+# 左臂单独调试配置：XR TCP 目标经 Placo QP 转换为 RM75 关节目标。
 # 末端外设由 peripherals_rm75.yaml 配置，真机连接阶段初始化后由遥操作节点复用。
 
 single_arm_velocity_teleop:
@@ -7,6 +7,7 @@ single_arm_velocity_teleop:
     controller_topic: /xr/left_controller
     control_rate_hz: 90.0
     command_timeout_sec: 0.12
+    feedback_resync_timeout_sec: 0.5
 
     # 手柄相对位姿 -> 目标 TCP 位姿；随后做目标低通、姿态低通和单帧步长限制。
     scale: 1.0
@@ -20,9 +21,7 @@ single_arm_velocity_teleop:
     enable_orientation_axes: [true, true, true]
     orientation_deadband_rad: 0.005
     orientation_filter_alpha: 0.65
-    max_orientation_speed: 0.6
-    current_pose_poll_hz: 10.0
-
+    max_orientation_speed: 0.5
     workspace_min: [-0.70, -0.60, 0.10]
     workspace_max: [0.70, 0.40, 0.70]
     cyl_radius_limit: [0.20, 0.60]
@@ -34,23 +33,27 @@ single_arm_velocity_teleop:
                          0.0, 0.0, 1.0,
                          -1.0, 0.0, 0.0]
 
-    use_mock: false
-    mock_initial_pose: [-0.2562, -0.2765, 0.1489, -3.0190, -0.1010, 3.1400]
     robot_ip: 192.168.192.18
     robot_port: 8080
+    realtime_push_host_ip: 192.168.192.148
+    realtime_push_port: 8089
+    realtime_push_cycle_ms: 5
     avoid_singularity: 0
-    frame_type: 1
     follow: false
     canfd_trajectory_mode: 2
     canfd_radio: 0
     configure_safety_limits: true
+    enable_tool_control: true
+    enable_trigger_gripper_control: true
+    trigger_close_threshold: 0.95
+    configure_peripheral_on_connect: true
     max_line_speed: 1.0
     max_angular_speed: 1.5
     max_line_acc: 1.0
     max_angular_acc: 2.0
     joint_max_speed: 180.0
     joint_max_acc: 180.0
-    move_to_initial_pose_on_connect: true
+    move_to_initial_pose_on_connect: false
     initial_joint_pose: [-79.55, -9.99, 71.01, 101.45, 95.07, -84.47, -74.52]
     init_move_speed: 20
     debug_topic_prefix: /xr_rm

xr_rm_bringup/config/right_arm_rm75.yaml

@@ -1,4 +1,4 @@
-# 右臂单独调试配置：XR 相对位姿透传控制 RM75 TCP。
+# 右臂单独调试配置：XR TCP 目标经 Placo QP 转换为 RM75 关节目标。
 # 末端外设由 peripherals_rm75.yaml 配置，真机连接阶段初始化后由遥操作节点复用。
 
 single_arm_velocity_teleop:
@@ -7,6 +7,7 @@ single_arm_velocity_teleop:
     controller_topic: /xr/right_controller
     control_rate_hz: 90.0
     command_timeout_sec: 0.12
+    feedback_resync_timeout_sec: 0.5
 
     scale: 0.7
     deadband_m: 0.001
@@ -20,11 +21,9 @@ single_arm_velocity_teleop:
     orientation_deadband_rad: 0.005
     orientation_filter_alpha: 0.65
     max_orientation_speed: 0.5
-    current_pose_poll_hz: 10.0
-
-    workspace_min: [-0.60, -0.60, 0.10]
-    workspace_max: [0.60, 0.70, 0.55]
-    cyl_radius_limit: [0.10, 0.70]
+    workspace_min: [-0.70, -0.70, 0.10]
+    workspace_max: [0.70, 0.70, 0.75]
+    cyl_radius_limit: [0.10, 0.80]
     low_z_threshold: 0.1
     low_z_min_radius: 0.1
 
@@ -33,23 +32,28 @@ single_arm_velocity_teleop:
                          0.0, 0.0, 1.0,
                          1.0, 0.0, 0.0]
 
-    use_mock: false
-    mock_initial_pose: [0.2663, -0.2606, 0.1027, 3.0330, 0.0000, 1.0910]
     robot_ip: 192.168.192.19
     robot_port: 8080
+    realtime_push_host_ip: 192.168.192.148
+    realtime_push_port: 8090
+    realtime_push_cycle_ms: 5
     avoid_singularity: 1
-    frame_type: 1
+    # 厂商 MovejCANFD 示例默认低跟随；高跟随仅在验证规划轨迹后单独开启。
     follow: false
     canfd_trajectory_mode: 2
     canfd_radio: 0
     configure_safety_limits: true
-    max_line_speed: 1.0
-    max_angular_speed: 1.5
-    max_line_acc: 1.0
-    max_angular_acc: 2.0
+    enable_tool_control: true
+    enable_trigger_gripper_control: true
+    trigger_close_threshold: 0.95
+    configure_peripheral_on_connect: true
+    max_line_speed: 0.25
+    max_angular_speed: 0.6
+    max_line_acc: 1.3
+    max_angular_acc: 3.0
     joint_max_speed: 180.0
-    joint_max_acc: 180.0
-    move_to_initial_pose_on_connect: true
+    joint_max_acc: 300.0
+    move_to_initial_pose_on_connect: false
     initial_joint_pose: [-90.14, 3.76, -86.89, 87.89, -96.53, -79.62, -90.04]
     init_move_speed: 20
     debug_topic_prefix: /xr_rm

xr_rm_bringup/launch/arm_debug.launch.py

@@ -5,6 +5,8 @@
 手柄接收节点，再根据 `arm:=left|right|both` 选择对应的遥操作节点。
 """
 
+from pathlib import Path
+
 from launch import LaunchDescription
 from launch.actions import DeclareLaunchArgument, OpaqueFunction
 from launch.substitutions import LaunchConfiguration, PathJoinSubstitution
@@ -12,6 +14,9 @@ from launch_ros.actions import Node
 from launch_ros.substitutions import FindPackageShare
 
 
+XR_PYTHON = "/home/robot/miniconda3/envs/xr/bin/python"
+
+
 def _as_bool(value: str) -> bool:
     """把 launch 字符串参数转换成 Python bool，便于在 OpaqueFunction 中分支。"""
     return value.strip().lower() in ("1", "true", "yes", "on")
@@ -26,8 +31,14 @@ def _config_file(name: str) -> PathJoinSubstitution:
     ])
 
 
-def _initial_pose_override(value: str) -> dict[str, bool]:
-    return {} if value == "auto" else {"move_to_initial_pose_on_connect": _as_bool(value)}
+def _rm75_urdf() -> PathJoinSubstitution:
+    return PathJoinSubstitution([
+        FindPackageShare("xr_rm_teleop"),
+        "models",
+        "rm75_omnipicker",
+        "urdf",
+        "RM75-B_OmniPicker_fixed.urdf",
+    ])
 
 
 def _udp_receiver_node() -> Node:
@@ -50,42 +61,21 @@ def _udp_receiver_node() -> Node:
 def _single_arm_node(
     arm: str,
     use_mock: bool,
-    move_to_initial_pose: str,
-    avoid_singularity: int,
-    frame_type: int,
-    control_rate_hz: float,
-    follow: bool,
-    configure_safety_limits: bool,
-    enable_tool_control: bool,
-    enable_trigger_gripper_control: bool,
-    trigger_close_threshold: float,
-    configure_peripheral_on_connect: bool,
 ) -> Node:
     """创建单臂调试节点；左/右臂分别使用独立 YAML，节点名保持单臂默认名。"""
     config_name = "left_arm_rm75.yaml" if arm == "left" else "right_arm_rm75.yaml"
-    robot_ip = LaunchConfiguration("left_robot_ip" if arm == "left" else "right_robot_ip")
     arm_name = _arm_name(arm)
     return Node(
         package="xr_rm_teleop",
         executable="single_arm_velocity_teleop",
         name="single_arm_velocity_teleop",
         output="screen",
+        prefix=[XR_PYTHON],
         parameters=[
             _config_file(config_name),
             {
                 "use_mock": use_mock,
-                "robot_ip": robot_ip,
-                "robot_port": LaunchConfiguration("robot_port"),
-                "avoid_singularity": avoid_singularity,
-                "frame_type": frame_type,
-                "control_rate_hz": control_rate_hz,
-                "follow": follow,
-                "configure_safety_limits": configure_safety_limits,
-                **_initial_pose_override(move_to_initial_pose),
-                "enable_tool_control": enable_tool_control,
-                "enable_trigger_gripper_control": enable_trigger_gripper_control,
-                "trigger_close_threshold": trigger_close_threshold,
-                "configure_peripheral_on_connect": configure_peripheral_on_connect,
+                "robot_urdf_path": _rm75_urdf(),
                 "peripheral_config_file": _config_file("peripherals_rm75.yaml"),
                 "peripheral_arm": arm,
                 "tool_command_topic": f"/xr_rm/{arm_name}/tool_enable",
@@ -98,20 +88,7 @@ def _arm_name(arm: str) -> str:
     return "left_rm75" if arm == "left" else "right_rm75"
 
 
-def _dual_arm_nodes(
-    use_mock: bool,
-    move_to_initial_pose: str,
-    left_avoid_singularity: int,
-    right_avoid_singularity: int,
-    frame_type: int,
-    control_rate_hz: float,
-    follow: bool,
-    configure_safety_limits: bool,
-    enable_tool_control: bool,
-    enable_trigger_gripper_control: bool,
-    trigger_close_threshold: float,
-    configure_peripheral_on_connect: bool,
-) -> list[Node]:
+def _dual_arm_nodes(use_mock: bool) -> list[Node]:
     """创建双臂节点；两个节点共用双臂 YAML，但节点名区分左右臂参数命名空间。"""
     config_file = _config_file("dual_arm_rm75.yaml")
     return [
@@ -120,22 +97,12 @@ def _dual_arm_nodes(
             executable="single_arm_velocity_teleop",
             name="left_arm_teleop",
             output="screen",
+            prefix=[XR_PYTHON],
             parameters=[
                 config_file,
                 {
                     "use_mock": use_mock,
-                    "robot_ip": LaunchConfiguration("left_robot_ip"),
-                    "robot_port": LaunchConfiguration("robot_port"),
-                    "avoid_singularity": left_avoid_singularity,
-                    "frame_type": frame_type,
-                    "control_rate_hz": control_rate_hz,
-                    "follow": follow,
-                    "configure_safety_limits": configure_safety_limits,
-                    **_initial_pose_override(move_to_initial_pose),
-                    "enable_tool_control": enable_tool_control,
-                    "enable_trigger_gripper_control": enable_trigger_gripper_control,
-                    "trigger_close_threshold": trigger_close_threshold,
-                    "configure_peripheral_on_connect": configure_peripheral_on_connect,
+                    "robot_urdf_path": _rm75_urdf(),
                     "peripheral_config_file": _config_file("peripherals_rm75.yaml"),
                     "peripheral_arm": "left",
                     "tool_command_topic": "/xr_rm/left_rm75/tool_enable",
@@ -147,22 +114,12 @@ def _dual_arm_nodes(
             executable="single_arm_velocity_teleop",
             name="right_arm_teleop",
             output="screen",
+            prefix=[XR_PYTHON],
             parameters=[
                 config_file,
                 {
                     "use_mock": use_mock,
-                    "robot_ip": LaunchConfiguration("right_robot_ip"),
-                    "robot_port": LaunchConfiguration("robot_port"),
-                    "avoid_singularity": right_avoid_singularity,
-                    "frame_type": frame_type,
-                    "control_rate_hz": control_rate_hz,
-                    "follow": follow,
-                    "configure_safety_limits": configure_safety_limits,
-                    **_initial_pose_override(move_to_initial_pose),
-                    "enable_tool_control": enable_tool_control,
-                    "enable_trigger_gripper_control": enable_trigger_gripper_control,
-                    "trigger_close_threshold": trigger_close_threshold,
-                    "configure_peripheral_on_connect": configure_peripheral_on_connect,
+                    "robot_urdf_path": _rm75_urdf(),
                     "peripheral_config_file": _config_file("peripherals_rm75.yaml"),
                     "peripheral_arm": "right",
                     "tool_command_topic": "/xr_rm/right_rm75/tool_enable",
@@ -175,76 +132,22 @@ def _dual_arm_nodes(
 def _launch_setup(context, *args, **kwargs):
     """运行时读取 launch 参数，决定启动单臂还是双臂。"""
     del args, kwargs
+    if not Path(XR_PYTHON).is_file():
+        raise RuntimeError(
+            f"XR Python not found: {XR_PYTHON}; "
+            "Placo 0.9.4 must not be installed globally"
+        )
     arm = LaunchConfiguration("arm").perform(context).strip().lower()
     use_mock = _as_bool(LaunchConfiguration("use_mock").perform(context))
-    move_to_initial_pose = LaunchConfiguration(
-        "move_to_initial_pose_on_connect"
-    ).perform(context).strip().lower()
-    avoid_override = LaunchConfiguration("avoid_singularity").perform(context).strip()
-    left_avoid_singularity = int(
-        avoid_override or LaunchConfiguration("left_avoid_singularity").perform(context)
-    )
-    right_avoid_singularity = int(
-        avoid_override or LaunchConfiguration("right_avoid_singularity").perform(context)
-    )
-    frame_type = int(LaunchConfiguration("frame_type").perform(context))
-    control_rate_hz = float(LaunchConfiguration("control_rate_hz").perform(context))
-    follow = _as_bool(LaunchConfiguration("follow").perform(context))
-    configure_safety_limits = _as_bool(
-        LaunchConfiguration("configure_safety_limits").perform(context)
-    )
-    configure_peripheral_on_connect = _as_bool(
-        LaunchConfiguration("configure_peripheral_on_connect").perform(context)
-    )
-    enable_tool_control = _as_bool(
-        LaunchConfiguration("enable_tool_control").perform(context)
-    )
-    enable_trigger_gripper_control = _as_bool(
-        LaunchConfiguration("enable_trigger_gripper_control").perform(context)
-    )
-    trigger_close_threshold = float(
-        LaunchConfiguration("trigger_close_threshold").perform(context)
-    )
 
     if arm not in ("left", "right", "both"):
         raise ValueError("arm must be one of: left, right, both")
 
     nodes = [_udp_receiver_node()]
     if arm == "both":
-        nodes.extend(
-            _dual_arm_nodes(
-                use_mock,
-                move_to_initial_pose,
-                left_avoid_singularity,
-                right_avoid_singularity,
-                frame_type,
-                control_rate_hz,
-                follow,
-                configure_safety_limits,
-                enable_tool_control,
-                enable_trigger_gripper_control,
-                trigger_close_threshold,
-                configure_peripheral_on_connect,
-            )
-        )
+        nodes.extend(_dual_arm_nodes(use_mock))
     else:
-        avoid_singularity = left_avoid_singularity if arm == "left" else right_avoid_singularity
-        nodes.append(
-            _single_arm_node(
-                arm,
-                use_mock,
-                move_to_initial_pose,
-                avoid_singularity,
-                frame_type,
-                control_rate_hz,
-                follow,
-                configure_safety_limits,
-                enable_tool_control,
-                enable_trigger_gripper_control,
-                trigger_close_threshold,
-                configure_peripheral_on_connect,
-            )
-        )
+        nodes.append(_single_arm_node(arm, use_mock))
     return nodes
 
 
@@ -259,30 +162,6 @@ def generate_launch_description() -> LaunchDescription:
         DeclareLaunchArgument("udp_port", default_value="15000"),
         # UDP receiver 轮询频率高于 PICO 发送频率，减少 socket 中等待时间。
         DeclareLaunchArgument("udp_timer_hz", default_value="200.0"),
-        # 左右 RM75 默认 IP，可在命令行中按现场网络覆盖。
-        DeclareLaunchArgument("left_robot_ip", default_value="192.168.192.18"),
-        DeclareLaunchArgument("right_robot_ip", default_value="192.168.192.19"),
-        DeclareLaunchArgument("robot_port", default_value="8080"),
-        # 真机位姿透传与安全配置参数。
-        DeclareLaunchArgument("left_avoid_singularity", default_value="0"),
-        DeclareLaunchArgument("right_avoid_singularity", default_value="1"),
-        # 非空时作为左右臂全局覆盖，例如 avoid_singularity:=0。
-        DeclareLaunchArgument("avoid_singularity", default_value=""),
-        DeclareLaunchArgument("frame_type", default_value="1"),
-        # 现场调参入口：默认按 PICO 90Hz 输入节奏发送 rm_movep_canfd。
-        DeclareLaunchArgument("control_rate_hz", default_value="90.0"),
-        # 默认低跟随；高跟随请确认控制器和网络能稳定满足厂商周期要求后再打开。
-        DeclareLaunchArgument("follow", default_value="false"),
-        DeclareLaunchArgument("configure_safety_limits", default_value="true"),
-        # 工具控制通过遥操作节点复用同一个 RealMan 连接，避免两个进程抢同一机械臂连接。
-        DeclareLaunchArgument("enable_tool_control", default_value="true"),
-        # trigger 上升沿切换夹爪开/关；grip 仍只控制机械臂运动。
-        DeclareLaunchArgument("enable_trigger_gripper_control", default_value="true"),
-        DeclareLaunchArgument("trigger_close_threshold", default_value="0.95"),
-        # 连接成功后是否配置外设；关闭后仅订阅开合话题，但开合前需要另行完成外设配置。
-        DeclareLaunchArgument("configure_peripheral_on_connect", default_value="true"),
-        # auto 时由单/双臂 YAML 决定；也可显式传 true/false 覆盖。
-        DeclareLaunchArgument("move_to_initial_pose_on_connect", default_value="auto"),
         # OpaqueFunction 允许根据 arm/use_mock 等运行时参数动态生成节点。
         OpaqueFunction(function=_launch_setup),
     ])

xr_rm_bringup/test/test_launcher_ui_cleanup.py

@@ -0,0 +1,94 @@
+import importlib.util
+import signal
+import subprocess
+import unittest
+from pathlib import Path
+from unittest import mock
+
+
+MODULE_PATH = Path(__file__).parents[1] / "tools" / "launcher_ui.py"
+SPEC = importlib.util.spec_from_file_location("launcher_ui", MODULE_PATH)
+launcher_ui = importlib.util.module_from_spec(SPEC)
+assert SPEC.loader is not None
+SPEC.loader.exec_module(launcher_ui)
+
+
+class LauncherCleanupTest(unittest.TestCase):
+    def test_xrobotoolkit_cleanup_policy_only_stops_service_on_window_close(self) -> None:
+        stop_all_patterns = set(
+            launcher_ui._xrobotoolkit_cleanup_patterns(stop_pc_service=False)
+        )
+        window_close_patterns = set(
+            launcher_ui._xrobotoolkit_cleanup_patterns(stop_pc_service=True)
+        )
+
+        self.assertLessEqual(
+            {"RobotLinuxDemo.x86_64", "PXREAClientUnity"},
+            stop_all_patterns,
+        )
+        self.assertNotIn("RoboticsServiceProcess", stop_all_patterns)
+        self.assertIn("RoboticsServiceProcess", window_close_patterns)
+
+    def test_close_and_stop_all_select_different_pc_service_policies(self) -> None:
+        app = object.__new__(launcher_ui.LauncherApp)
+        calls = []
+
+        class Root:
+            destroyed = False
+
+            def destroy(self) -> None:
+                self.destroyed = True
+
+        app.root = Root()
+        app.stop_launched_processes = lambda **kwargs: calls.append(kwargs) or True
+
+        app.kill_launched_processes()
+        app.on_close_requested()
+
+        self.assertEqual(
+            calls,
+            [
+                {"confirm": True, "notify": True, "stop_pc_service": False},
+                {"confirm": True, "notify": False, "stop_pc_service": True},
+            ],
+        )
+        self.assertTrue(app.root.destroyed)
+
+    def test_stop_all_keeps_oldest_pc_service_and_stops_duplicates(self) -> None:
+        app = object.__new__(launcher_ui.LauncherApp)
+        app.status = mock.Mock()
+        app.close_related_terminal_windows = lambda: 0
+
+        def fake_check_output(command, **_kwargs):
+            if command == ["pgrep", "-o", "-f", "RoboticsServiceProcess"]:
+                return "101\n"
+            if command == ["pgrep", "-f", "RoboticsServiceProcess"]:
+                return "101\n202\n303\n"
+            raise subprocess.CalledProcessError(1, command)
+
+        with (
+            mock.patch.object(
+                launcher_ui.subprocess,
+                "check_output",
+                side_effect=fake_check_output,
+            ),
+            mock.patch.object(launcher_ui.os, "kill") as kill,
+            mock.patch.object(launcher_ui.time, "sleep"),
+        ):
+            app.stop_launched_processes(
+                confirm=False,
+                notify=False,
+                stop_pc_service=False,
+            )
+
+        self.assertEqual(
+            kill.call_args_list,
+            [
+                mock.call(202, signal.SIGTERM),
+                mock.call(303, signal.SIGTERM),
+            ],
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

xr_rm_bringup/tools/launcher_ui.py

@@ -121,6 +121,17 @@ def _xrobotoolkit_bridge_command() -> str:
     )
 
 
+def _xrobotoolkit_cleanup_patterns(*, stop_pc_service: bool) -> tuple[str, ...]:
+    patterns = (
+        XROBOTOOLKIT_BRIDGE_EXECUTABLE,
+        "RobotLinuxDemo.x86_64",
+        "PXREAClientUnity",
+    )
+    if stop_pc_service:
+        return (*patterns, "RoboticsServiceProcess")
+    return patterns
+
+
 def _tool_command(arm: str, open_tool: bool) -> str:
     arm_name = "left_rm75" if arm == "left" else "right_rm75"
     value = "true" if open_tool else "false"
@@ -289,8 +300,7 @@ def build_commands_by_mode(mode: str) -> list[tuple[str, str]]:
             ("Ping Left RM75", f"ping -c 4 {DEFAULT_LEFT_IP}"),
             (
                 "Left Arm RealMan Launch",
-                "ros2 launch xr_rm_bringup arm_debug.launch.py arm:=left use_mock:=false "
-                f"left_robot_ip:={DEFAULT_LEFT_IP}",
+                "ros2 launch xr_rm_bringup arm_debug.launch.py arm:=left use_mock:=false",
             ),
             ("XRobotoolkit UDP Bridge (90 Hz)", _xrobotoolkit_bridge_command()),
             ("Left Tool Open", _tool_command("left", True)),
@@ -306,8 +316,7 @@ def build_commands_by_mode(mode: str) -> list[tuple[str, str]]:
             ("Ping Right RM75", f"ping -c 4 {DEFAULT_RIGHT_IP}"),
             (
                 "Right Arm RealMan Launch",
-                "ros2 launch xr_rm_bringup arm_debug.launch.py arm:=right use_mock:=false "
-                f"right_robot_ip:={DEFAULT_RIGHT_IP}",
+                "ros2 launch xr_rm_bringup arm_debug.launch.py arm:=right use_mock:=false",
             ),
             ("XRobotoolkit UDP Bridge (90 Hz)", _xrobotoolkit_bridge_command()),
             ("Right Tool Open", _tool_command("right", True)),
@@ -324,9 +333,7 @@ def build_commands_by_mode(mode: str) -> list[tuple[str, str]]:
             ("Ping Right RM75", f"ping -c 4 {DEFAULT_RIGHT_IP}"),
             (
                 "Dual Arm RealMan Launch",
-                "ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false "
-                f"left_robot_ip:={DEFAULT_LEFT_IP} right_robot_ip:={DEFAULT_RIGHT_IP} "
-                "move_to_initial_pose_on_connect:=false",
+                "ros2 launch xr_rm_bringup arm_debug.launch.py arm:=both use_mock:=false",
             ),
             ("XRobotoolkit UDP Bridge (90 Hz)", _xrobotoolkit_bridge_command()),
             (
@@ -1283,13 +1290,27 @@ class LauncherApp:
         return closed
 
     def on_close_requested(self) -> None:
-        if self.stop_launched_processes(confirm=True, notify=False):
+        if self.stop_launched_processes(
+            confirm=True,
+            notify=False,
+            stop_pc_service=True,
+        ):
             self.root.destroy()
 
     def kill_launched_processes(self) -> None:
-        self.stop_launched_processes(confirm=True, notify=True)
+        self.stop_launched_processes(
+            confirm=True,
+            notify=True,
+            stop_pc_service=False,
+        )
 
-    def stop_launched_processes(self, *, confirm: bool, notify: bool) -> bool:
+    def stop_launched_processes(
+        self,
+        *,
+        confirm: bool,
+        notify: bool,
+        stop_pc_service: bool,
+    ) -> bool:
         if confirm and not messagebox.askyesno(
             "Confirm Stop",
             "Stop XR-RM launcher terminals, topic monitors, ROS nodes, and bridge processes started from this workspace?",
@@ -1302,7 +1323,7 @@ class LauncherApp:
             "ros2 run xr_rm_input",
             "ros2 run xr_rm_teleop",
             "XR_RM_LAUNCHER_SESSION=1",
-            XROBOTOOLKIT_BRIDGE_EXECUTABLE,
+            *_xrobotoolkit_cleanup_patterns(stop_pc_service=stop_pc_service),
             TERMINAL_TITLE_PREFIX,
             TOPIC_MONITOR_TITLE,
             CMD_VEL_MONITOR_TITLE,
@@ -1326,6 +1347,20 @@ class LauncherApp:
             "ros2 topic list",
             "ros2 node list",
         ]
+        pc_service_keep_pid: int | None = None
+        if not stop_pc_service:
+            try:
+                output = subprocess.check_output(
+                    ["pgrep", "-o", "-f", "RoboticsServiceProcess"],
+                    text=True,
+                )
+                pc_service_keep_pid = int(output.strip())
+                patterns.append("RoboticsServiceProcess")
+            except (subprocess.CalledProcessError, ValueError):
+                pass
+            except Exception as exc:
+                print(f"Failed to find the oldest RoboticsServiceProcess: {exc}")
+
         protected = {os.getpid(), os.getppid()}
         killed: set[int] = set()
 
@@ -1345,6 +1380,8 @@ class LauncherApp:
                     continue
                 if pid in protected or pid in killed:
                     continue
+                if pattern == "RoboticsServiceProcess" and pid == pc_service_keep_pid:
+                    continue
                 try:
                     os.kill(pid, signal.SIGTERM)
                     killed.add(pid)

xr_rm_teleop/models/rm75/RM75-B.urdf

@@ -0,0 +1,453 @@
+<?xml version="1.0" encoding="utf-8"?>
+<!-- This URDF was automatically created by SolidWorks to URDF Exporter! Originally created by Stephen Brawner (brawner@gmail.com) 
+     Commit Version: 1.6.0-1-g15f4949  Build Version: 1.6.7594.29634
+     For more information, please see http://wiki.ros.org/sw_urdf_exporter -->
+<robot
+  name="RM75-B">
+  <link
+    name="base_link">
+    <inertial>
+      <origin
+        xyz="0.00049987 5.2709E-05 0.060019"
+        rpy="0 0 0" />
+      <mass
+        value="1.862" />
+      <inertia
+        ixx="0.0017232"
+        ixy="-3.1058E-06"
+        ixz="-3.7924E-05"
+        iyy="0.0017051"
+        iyz="1.3691E-06"
+        izz="0.00090158" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/base_link.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/base_link.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <link
+    name="link_1">
+    <inertial>
+      <origin
+        xyz="0.000241 -0.013273 -0.00995"
+        rpy="0 0 0" />
+      <mass
+        value="1.574" />
+      <inertia
+        ixx="0.002487573"
+        ixy="0.000009663"
+        ixz="-0.000007909"
+        iyy="0.002321038"
+        iyz="0.000179393"
+        izz="0.001450554" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_1.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_1.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_1"
+    type="revolute">
+    <origin
+      xyz="0 0 0.2405"
+      rpy="0 0 0" />
+    <parent
+      link="base_link" />
+    <child
+      link="link_1" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-3.106"
+      upper="3.106"
+      effort="60"
+      velocity="3.14" />
+  </joint>
+  <link
+    name="link_2">
+    <inertial>
+      <origin
+        xyz="-0.000357 -0.106789 0.005329"
+        rpy="0 0 0" />
+      <mass
+        value="1.217" />
+      <inertia
+        ixx="0.003494121"
+        ixy="0.000002921"
+        ixz="-0.000005613"
+        iyy="0.000892721"
+        iyz="-0.000583884"
+        izz="0.003444080" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_2.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_2.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_2"
+    type="revolute">
+    <origin
+      xyz="0 0 0"
+      rpy="-1.5708 0 0" />
+    <parent
+      link="link_1" />
+    <child
+      link="link_2" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-2.2689"
+      upper="2.2689"
+      effort="60"
+      velocity="3.14" />
+  </joint>
+  <link
+    name="link_3">
+    <inertial>
+      <origin
+        xyz="0.000003 -0.01398 -0.011324"
+        rpy="0 0 0" />
+      <mass
+        value="1.11" />
+      <inertia
+        ixx="0.001836663"
+        ixy="0.000002259"
+        ixz="-0.000004216"
+        iyy="0.001498875"
+        iyz="0.000037167"
+        izz="0.001062545" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_3.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_3.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_3"
+    type="revolute">
+    <origin
+      xyz="0 -0.256 0"
+      rpy="1.5708 0 0" />
+    <parent
+      link="link_2" />
+    <child
+      link="link_3" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-3.106"
+      upper="3.106"
+      effort="30"
+      velocity="3.14" />
+  </joint>
+  <link
+    name="link_4">
+    <inertial>
+      <origin
+        xyz="-0.000005 -0.084658 0.004747"
+        rpy="0 0 0" />
+      <mass
+        value="0.685" />
+      <inertia
+        ixx="0.001282444"
+        ixy="-0.000000551"
+        ixz="-0.000000630"
+        iyy="0.000373013"
+        iyz="-0.000232084"
+        izz="0.001256177" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_4.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_4.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_4"
+    type="revolute">
+    <origin
+      xyz="0 0 0"
+      rpy="-1.5708 0 0" />
+    <parent
+      link="link_3" />
+    <child
+      link="link_4" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-2.356"
+      upper="2.356"
+      effort="30"
+      velocity="3.14" />
+  </joint>
+  <link
+    name="link_5">
+    <inertial>
+      <origin
+        xyz="0.000078 -0.012937 -0.008781"
+        rpy="0 0 0" />
+      <mass
+        value="0.619" />
+      <inertia
+        ixx="0.000627336"
+        ixy="0.000001636"
+        ixz="-0.000001345"
+        iyy="0.000542455"
+        iyz="0.000034970"
+        izz="0.000370291" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_5.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_5.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_5"
+    type="revolute">
+    <origin
+      xyz="0 -0.21 0"
+      rpy="1.5708 0 0" />
+    <parent
+      link="link_4" />
+    <child
+      link="link_5" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-3.106"
+      upper="3.106"
+      effort="10"
+      velocity="3.14" />
+  </joint>
+  <link
+    name="link_6">
+    <inertial>
+      <origin
+        xyz="-0.000014 -0.078524 0.002819"
+        rpy="0 0 0" />
+      <mass
+        value="0.602" />
+      <inertia
+        ixx="0.000780774"
+        ixy="-0.000000121"
+        ixz="-0.000000469"
+        iyy="0.000289973"
+        iyz="-0.000120513"
+        izz="0.000763955" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_6.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_6.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_6"
+    type="revolute">
+    <origin
+      xyz="0 0 0"
+      rpy="-1.5708 0 0" />
+    <parent
+      link="link_5" />
+    <child
+      link="link_6" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-2.234"
+      upper="2.234"
+      effort="10"
+      velocity="3.14" />
+  </joint>
+  <link
+    name="link_7">
+    <inertial>
+      <origin
+        xyz="0.001094 -0.000077 -0.010119"
+        rpy="0 0 0" />
+      <mass
+        value="0.107" />
+      <inertia
+        ixx="0.000044123"
+        ixy="-0.000000064"
+        ixz="0.0000003"
+        iyy="0.000035078"
+        iyz="-0.000000029"
+        izz="0.000065445" />
+    </inertial>
+    <visual>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_7.STL" />
+      </geometry>
+      <material
+        name="">
+        <color
+          rgba="1 1 1 1" />
+      </material>
+    </visual>
+    <collision>
+      <origin
+        xyz="0 0 0"
+        rpy="0 0 0" />
+      <geometry>
+        <mesh
+          filename="meshes/link_7.STL" />
+      </geometry>
+    </collision>
+  </link>
+  <joint
+    name="joint_7"
+    type="revolute">
+    <origin
+      xyz="0 -0.144 0"
+      rpy="1.5708 0 0" />
+    <parent
+      link="link_6" />
+    <child
+      link="link_7" />
+    <axis
+      xyz="0 0 1" />
+    <limit
+      lower="-6.28"
+      upper="6.28"
+      effort="10"
+      velocity="3.14" />
+  </joint>
+</robot>

xr_rm_teleop/models/rm75_omnipicker/urdf/RM75-B_OmniPicker_fixed.urdf

@@ -0,0 +1,503 @@
+<?xml version='1.0' encoding='UTF-8'?>
+<robot name="RM75_B_OmniPicker_fixed">
+  <link name="base_link">
+    <inertial>
+      <origin xyz="0.00049987 5.2709E-05 0.060019" rpy="0 0 0"/>
+      <mass value="1.862"/>
+      <inertia ixx="0.0017232" ixy="-3.1058E-06" ixz="-3.7924E-05" iyy="0.0017051" iyz="1.3691E-06" izz="0.00090158"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/base_link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/base_link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <link name="link_1">
+    <inertial>
+      <origin xyz="0.000241 -0.013273 -0.00995" rpy="0 0 0"/>
+      <mass value="1.574"/>
+      <inertia ixx="0.002487573" ixy="0.000009663" ixz="-0.000007909" iyy="0.002321038" iyz="0.000179393" izz="0.001450554"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_1.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_1.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_1" type="revolute">
+    <origin xyz="0 0 0.2405" rpy="0 0 0"/>
+    <parent link="base_link"/>
+    <child link="link_1"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-3.106" upper="3.106" effort="60" velocity="3.14"/>
+  </joint>
+  <link name="link_2">
+    <inertial>
+      <origin xyz="-0.000357 -0.106789 0.005329" rpy="0 0 0"/>
+      <mass value="1.217"/>
+      <inertia ixx="0.003494121" ixy="0.000002921" ixz="-0.000005613" iyy="0.000892721" iyz="-0.000583884" izz="0.003444080"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_2.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_2.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_2" type="revolute">
+    <origin xyz="0 0 0" rpy="-1.5708 0 0"/>
+    <parent link="link_1"/>
+    <child link="link_2"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-2.2689" upper="2.2689" effort="60" velocity="3.14"/>
+  </joint>
+  <link name="link_3">
+    <inertial>
+      <origin xyz="0.000003 -0.01398 -0.011324" rpy="0 0 0"/>
+      <mass value="1.11"/>
+      <inertia ixx="0.001836663" ixy="0.000002259" ixz="-0.000004216" iyy="0.001498875" iyz="0.000037167" izz="0.001062545"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_3.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_3.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_3" type="revolute">
+    <origin xyz="0 -0.256 0" rpy="1.5708 0 0"/>
+    <parent link="link_2"/>
+    <child link="link_3"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-3.106" upper="3.106" effort="30" velocity="3.14"/>
+  </joint>
+  <link name="link_4">
+    <inertial>
+      <origin xyz="-0.000005 -0.084658 0.004747" rpy="0 0 0"/>
+      <mass value="0.685"/>
+      <inertia ixx="0.001282444" ixy="-0.000000551" ixz="-0.000000630" iyy="0.000373013" iyz="-0.000232084" izz="0.001256177"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_4.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_4.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_4" type="revolute">
+    <origin xyz="0 0 0" rpy="-1.5708 0 0"/>
+    <parent link="link_3"/>
+    <child link="link_4"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-2.356" upper="2.356" effort="30" velocity="3.14"/>
+  </joint>
+  <link name="link_5">
+    <inertial>
+      <origin xyz="0.000078 -0.012937 -0.008781" rpy="0 0 0"/>
+      <mass value="0.619"/>
+      <inertia ixx="0.000627336" ixy="0.000001636" ixz="-0.000001345" iyy="0.000542455" iyz="0.000034970" izz="0.000370291"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_5.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_5.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_5" type="revolute">
+    <origin xyz="0 -0.21 0" rpy="1.5708 0 0"/>
+    <parent link="link_4"/>
+    <child link="link_5"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-3.106" upper="3.106" effort="10" velocity="3.14"/>
+  </joint>
+  <link name="link_6">
+    <inertial>
+      <origin xyz="-0.000014 -0.078524 0.002819" rpy="0 0 0"/>
+      <mass value="0.602"/>
+      <inertia ixx="0.000780774" ixy="-0.000000121" ixz="-0.000000469" iyy="0.000289973" iyz="-0.000120513" izz="0.000763955"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_6.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_6.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_6" type="revolute">
+    <origin xyz="0 0 0" rpy="-1.5708 0 0"/>
+    <parent link="link_5"/>
+    <child link="link_6"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-2.234" upper="2.234" effort="10" velocity="3.14"/>
+  </joint>
+  <link name="link_7">
+    <inertial>
+      <origin xyz="0.001094 -0.000077 -0.010119" rpy="0 0 0"/>
+      <mass value="0.107"/>
+      <inertia ixx="0.000044123" ixy="-0.000000064" ixz="0.0000003" iyy="0.000035078" iyz="-0.000000029" izz="0.000065445"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_7.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="1 1 1 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/rm75/link_7.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="joint_7" type="revolute">
+    <origin xyz="0 -0.144 0" rpy="1.5708 0 0"/>
+    <parent link="link_6"/>
+    <child link="link_7"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-6.28" upper="6.28" effort="10" velocity="3.14"/>
+  </joint>
+  <!-- RM75 end-flange alias. link_7 is treated as the tool mounting frame. -->
+  <link name="rm75_flange"/>
+  <joint name="rm75_link7_to_flange" type="fixed">
+    <origin xyz="0 0 0" rpy="0 0 0"/>
+    <parent link="link_7"/>
+    <child link="rm75_flange"/>
+  </joint>
+  <!-- OmniPicker mounting transform. Adjust xyz/rpy here if an adapter plate or different clocking is used. -->
+  <joint name="rm75_flange_to_omnipicker" type="fixed">
+    <origin xyz="0 0 0" rpy="0 0 0"/>
+    <parent link="rm75_flange"/>
+    <child link="omnipicker_base_link"/>
+  </joint>
+  <link name="omnipicker_base_link">
+    <inertial>
+      <origin xyz="-0.00005520 1.2341E-05 0.03296193" rpy="0 0 0"/>
+      <mass value="0.25641368"/>
+      <inertia ixx="4.6351E-04" ixy="-1.0E-08" ixz="-1.04E-06" iyy="4.4525E-04" iyz="-5.00E-08" izz="1.0438E-04"/>
+    </inertial>
+    <visual>
+      <origin rpy="0 0 0" xyz="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/base_link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.89804 0.91765 0.92941 1"/>
+      </material>
+    </visual>
+  </link>
+  <link name="omnipicker_hand_narrow1_Link">
+    <inertial>
+      <origin xyz="0.0094685 0.0068806 6.5437E-05" rpy="0 0 0"/>
+      <mass value="0.025428"/>
+      <inertia ixx="1.9574E-06" ixy="-4.2911E-07" ixz="-3.7111E-11" iyy="2.3919E-06" iyz="-3.008E-10" izz="1.5501E-06"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow1_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow1_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_narrow1_joint" type="fixed">
+    <origin xyz="0 -0.0195 0.0565" rpy="-2.9951 -1.5708 -0.15964"/>
+    <parent link="omnipicker_base_link"/>
+    <child link="omnipicker_hand_narrow1_Link"/>
+  </joint>
+  <link name="omnipicker_hand_narrow2_Link">
+    <inertial>
+      <origin xyz="0.0088027 -0.007035 1.6424E-05" rpy="0 0 0"/>
+      <mass value="0.0040132"/>
+      <inertia ixx="1.0307E-07" ixy="5.7851E-08" ixz="9.5801E-11" iyy="1.1385E-07" iyz="-2.81E-11" izz="1.7009E-07"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow2_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow2_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_narrow2_joint" type="fixed">
+    <origin xyz="0.030852 0.018551 0" rpy="0 0 0"/>
+    <parent link="omnipicker_hand_narrow1_Link"/>
+    <child link="omnipicker_hand_narrow2_Link"/>
+  </joint>
+  <link name="omnipicker_hand_narrow3_Link">
+    <inertial>
+      <origin xyz="0.012508 -0.0079729 9.4339E-05" rpy="0 0 0"/>
+      <mass value="0.018029"/>
+      <inertia ixx="1.1403E-06" ixy="5.5159E-07" ixz="-4.0096E-13" iyy="2.5704E-06" iyz="1.0951E-12" izz="2.3252E-06"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow3_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow3_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_narrow3_joint" type="fixed">
+    <origin xyz="0.018118 -0.01574 0" rpy="0 0 0"/>
+    <parent link="omnipicker_hand_narrow2_Link"/>
+    <child link="omnipicker_hand_narrow3_Link"/>
+  </joint>
+  <link name="omnipicker_hand_narrow4_Link">
+  </link>
+  <joint name="omnipicker_hand_narrow4_joint" type="fixed">
+    <origin xyz="0 -0.0104 0" rpy="0 0 0"/>
+    <parent link="omnipicker_hand_narrow3_Link"/>
+    <child link="omnipicker_hand_narrow4_Link"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-3.14" upper="3.14" effort="0" velocity="0"/>
+  </joint>
+  <link name="omnipicker_hand_narrow_loop_Link">
+    <inertial>
+      <origin xyz="0.014869 -0.0036066 0.00029307" rpy="0 0 0"/>
+      <mass value="0.022591"/>
+      <inertia ixx="4.3916E-06" ixy="1.114E-07" ixz="-4.9655E-12" iyy="4.737E-06" iyz="1.7121E-11" izz="6.0445E-07"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow_loop_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/narrow_loop_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_narrow_loop_joint" type="fixed">
+    <origin xyz="0 -0.021633 0.07387" rpy="-2.9951 -1.5708 -0.15964"/>
+    <parent link="omnipicker_base_link"/>
+    <child link="omnipicker_hand_narrow_loop_Link"/>
+  </joint>
+  <link name="omnipicker_hand_wide1_Link">
+    <inertial>
+      <origin xyz="0.0095051 -0.0068479 6.8268E-05" rpy="0 0 0"/>
+      <mass value="0.025428"/>
+      <inertia ixx="1.9565E-06" ixy="4.2798E-07" ixz="2.3844E-10" iyy="2.3928E-06" iyz="-1.4454E-10" izz="1.5501E-06"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide1_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide1_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_wide1_joint" type="fixed">
+    <origin xyz="0 0.0195 0.0565" rpy="-2.9951 -1.5708 -0.15964"/>
+    <parent link="omnipicker_base_link"/>
+    <child link="omnipicker_hand_wide1_Link"/>
+  </joint>
+  <link name="omnipicker_hand_wide2_Link">
+    <inertial>
+      <origin xyz="0.0088027 0.007035 -1.6424E-05" rpy="0 0 0"/>
+      <mass value="0.0040132"/>
+      <inertia ixx="1.0307E-07" ixy="-5.7851E-08" ixz="-9.58E-11" iyy="1.1385E-07" iyz="-2.81E-11" izz="1.7009E-07"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide2_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide2_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_wide2_joint" type="fixed">
+    <origin xyz="0.030852 -0.018551 0" rpy="0 0 0"/>
+    <parent link="omnipicker_hand_wide1_Link"/>
+    <child link="omnipicker_hand_wide2_Link"/>
+  </joint>
+  <link name="omnipicker_hand_wide3_Link">
+    <inertial>
+      <origin xyz="0.016206 0.0094593 4.7668E-05" rpy="0 0 0"/>
+      <mass value="0.035835"/>
+      <inertia ixx="8.5056E-06" ixy="-1.1363E-06" ixz="-4.2908E-11" iyy="1.1235E-05" iyz="-2.9251E-11" izz="4.6309E-06"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide3_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide3_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_wide3_joint" type="fixed">
+    <origin xyz="0.018118 0.01574 0" rpy="0 0 0"/>
+    <parent link="omnipicker_hand_wide2_Link"/>
+    <child link="omnipicker_hand_wide3_Link"/>
+  </joint>
+  <link name="omnipicker_hand_wide4_Link">
+  </link>
+  <joint name="omnipicker_hand_wide4_joint" type="fixed">
+    <origin xyz="0 0.0104 0" rpy="0 0 0"/>
+    <parent link="omnipicker_hand_wide3_Link"/>
+    <child link="omnipicker_hand_wide4_Link"/>
+    <axis xyz="0 0 1"/>
+    <limit lower="-3.14" upper="3.14" effort="0" velocity="0"/>
+  </joint>
+  <link name="omnipicker_hand_wide_loop_Link">
+    <inertial>
+      <origin xyz="0.016268 0.0040555 0.00030323" rpy="0 0 0"/>
+      <mass value="0.025142"/>
+      <inertia ixx="5.887E-06" ixy="-1.1234E-07" ixz="2.1954E-11" iyy="6.236E-06" iyz="-9.473E-12" izz="6.2389E-07"/>
+    </inertial>
+    <visual>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide_loop_Link.STL"/>
+      </geometry>
+      <material name="">
+        <color rgba="0.75294 0.75294 0.75294 1"/>
+      </material>
+    </visual>
+    <collision>
+      <origin xyz="0 0 0" rpy="0 0 0"/>
+      <geometry>
+        <mesh filename="package://xr_rm_teleop/models/rm75_omnipicker/meshes/omnipicker/wide_loop_Link.STL"/>
+      </geometry>
+    </collision>
+  </link>
+  <joint name="omnipicker_hand_wide_loop_joint" type="fixed">
+    <origin xyz="0 0.021633 0.07387" rpy="-2.9951 -1.5708 -0.15964"/>
+    <parent link="omnipicker_base_link"/>
+    <child link="omnipicker_hand_wide_loop_Link"/>
+  </joint>
+  <link name="omnipicker_mount_frame"/>
+  <joint name="omnipicker_base_to_mount_frame" type="fixed">
+    <origin xyz="0 0 0" rpy="0 0 0"/>
+    <parent link="omnipicker_base_link"/>
+    <child link="omnipicker_mount_frame"/>
+  </joint>
+  <link name="omnipicker_tcp"/>
+  <joint name="omnipicker_tcp_joint" type="fixed">
+    <parent link="omnipicker_base_link"/>
+    <child link="omnipicker_tcp"/>
+    <origin xyz="0 0 0.16" rpy="0 0 0"/>
+  </joint>
+</robot>

xr_rm_teleop/setup.py

@@ -1,8 +1,10 @@
 """xr_rm_teleop 包安装配置。
 
-该包提供基于 XR 相对位姿的 RM75 笛卡尔位姿透传遥操作节点。
+该包提供基于 XR 相对位姿和 Placo QP 的 RM75 遥操作节点。
 """
 
+from glob import glob
+
 from setuptools import setup
 
 package_name = "xr_rm_teleop"
@@ -14,6 +16,26 @@ setup(
     data_files=[
         ("share/ament_index/resource_index/packages", [f"resource/{package_name}"]),
         (f"share/{package_name}", ["package.xml"]),
+        (
+            f"share/{package_name}/models/rm75",
+            ["models/rm75/RM75-B.urdf"],
+        ),
+        (
+            f"share/{package_name}/models/rm75/meshes",
+            glob("models/rm75/meshes/*.STL"),
+        ),
+        (
+            f"share/{package_name}/models/rm75_omnipicker/urdf",
+            glob("models/rm75_omnipicker/urdf/*.urdf"),
+        ),
+        (
+            f"share/{package_name}/models/rm75_omnipicker/meshes/rm75",
+            glob("models/rm75_omnipicker/meshes/rm75/*.STL"),
+        ),
+        (
+            f"share/{package_name}/models/rm75_omnipicker/meshes/omnipicker",
+            glob("models/rm75_omnipicker/meshes/omnipicker/*.STL"),
+        ),
     ],
     install_requires=["setuptools"],
     zip_safe=True,

xr_rm_teleop/test/placo_ik_smoke.py

@@ -0,0 +1,94 @@
+from __future__ import annotations
+
+import math
+import sys
+import time
+from pathlib import Path
+
+import numpy as np
+
+from xr_rm_teleop.placo_ik_solver import PlacoIkSolver
+
+
+CASES = {
+    "left": [-79.55, -9.99, 71.01, 101.45, 95.07, -84.47, -74.52],
+    "right": [-90.14, 3.76, -86.89, 87.89, -96.53, -79.62, -90.04],
+}
+
+
+def rotation_z(angle: float) -> np.ndarray:
+    cosine = math.cos(angle)
+    sine = math.sin(angle)
+    return np.asarray(
+        [
+            [cosine, -sine, 0.0],
+            [sine, cosine, 0.0],
+            [0.0, 0.0, 1.0],
+        ]
+    )
+
+
+def angle_error(actual: np.ndarray, target: np.ndarray) -> float:
+    cosine = np.clip((np.trace(target @ actual.T) - 1.0) * 0.5, -1.0, 1.0)
+    return float(math.acos(cosine))
+
+
+def main() -> None:
+    urdf_path = Path(sys.argv[1]).resolve()
+    for arm, joint_degrees in CASES.items():
+        initial_joints = np.deg2rad(joint_degrees)
+        drift_solver = PlacoIkSolver(str(urdf_path), 1.0 / 125.0)
+        joints = initial_joints.tolist()
+        stationary_target = drift_solver.update_joint_state(joints)
+        flange = drift_solver._robot.get_T_world_frame("link_7")
+        flange_to_tcp = np.linalg.inv(flange) @ stationary_target
+        assert np.allclose(flange_to_tcp[:3, 3], [0.0, 0.0, 0.16])
+        assert np.allclose(flange_to_tcp[:3, :3], np.eye(3))
+        for _ in range(250):
+            drift_solver.update_joint_state(joints)
+            joints = drift_solver.solve(stationary_target)
+        drift_degrees = float(
+            np.max(np.abs(np.rad2deg(np.asarray(joints) - initial_joints)))
+        )
+        assert drift_degrees <= 0.05, (
+            f"{arm} stationary target drifted {drift_degrees:.3f}deg"
+        )
+
+        solver = PlacoIkSolver(str(urdf_path), 1.0 / 125.0)
+        joints = initial_joints.tolist()
+        current = solver.update_joint_state(joints)
+        assert current.shape == (4, 4)
+        target = current.copy()
+        target[0, 3] += 0.01
+        target[:3, :3] = rotation_z(0.05) @ target[:3, :3]
+
+        solve_durations = []
+        for _ in range(250):
+            solver.update_joint_state(joints)
+            started_at = time.perf_counter()
+            joints = solver.solve(target)
+            solve_durations.append(time.perf_counter() - started_at)
+
+        actual = solver.update_joint_state(joints)
+        position_error = np.linalg.norm(actual[:3, 3] - target[:3, 3])
+        orientation_error = angle_error(actual[:3, :3], target[:3, :3])
+        assert len(joints) == 7
+        assert np.isfinite(joints).all()
+        assert np.allclose(
+            solver.base_configuration,
+            [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
+        )
+        assert position_error <= 0.005
+        assert orientation_error <= math.radians(2.0)
+        print(
+            f"{arm}: position_error={position_error:.6f}m, "
+            f"orientation_error={math.degrees(orientation_error):.3f}deg, "
+            f"stationary_drift={drift_degrees:.3f}deg, "
+            f"solve_avg={1000.0 * np.mean(solve_durations):.3f}ms, "
+            f"solve_max={1000.0 * max(solve_durations):.3f}ms, "
+            f"solve_overruns={sum(value > 1.0 / 125.0 for value in solve_durations)}"
+        )
+
+
+if __name__ == "__main__":
+    main()

xr_rm_teleop/test/test_initial_joint_pose.py

@@ -1,4 +1,13 @@
+import math
+import sys
+from types import ModuleType, SimpleNamespace
+
+import pytest
+
+from xr_rm_teleop import realman_adapter
 from xr_rm_teleop.realman_adapter import RealManAdapter
+from xr_rm_teleop.realman_adapter import MockRealManAdapter
+from xr_rm_teleop.fun_peripheral import PeripheralConfig, _configure_tool_frame
 
 
 def test_initial_pose_uses_joint_move_only() -> None:
@@ -11,9 +20,509 @@ def test_initial_pose_uses_joint_move_only() -> None:
             return 0
 
     joints = [-167.21, 28.48, 28.21, 61.35, -14.40, 84.49, -124.51]
-    adapter = RealManAdapter("127.0.0.1", 8080, 0, 1, initial_joint_pose=joints)
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+        initial_joint_pose=joints,
+    )
     adapter._arm = FakeArm()
 
     adapter._move_to_initial_pose()
 
     assert adapter._arm.calls == [(joints, 20, 0, 0, 1)]
+
+
+def test_peripheral_config_exposes_selected_tool() -> None:
+    config = PeripheralConfig(
+        scissorgripper=1,
+        tools_in_ee={
+            "first": [[0.0] * 7, [0.0] * 7],
+            "second": [[0.0, 0.0, 0.16, 0.0, 0.0, 0.0, 1.0], [0.0] * 7],
+        },
+    )
+
+    assert config.tool_name == "second"
+    assert config.tool_pose == [0.0, 0.0, 0.16, 0.0, 0.0, 0.0, 1.0]
+
+
+@pytest.mark.parametrize(
+    ("existing", "expected_operation"),
+    [(False, "create"), (True, "update")],
+)
+def test_tool_frame_is_created_or_updated(existing, expected_operation) -> None:
+    class FakeArm:
+        def __init__(self) -> None:
+            self.calls = []
+
+        def rm_get_total_tool_frame(self):
+            self.calls.append(("get",))
+            names = ["omnipic"] if existing else []
+            return {"return_code": 0, "tool_names": names}
+
+        def rm_set_manual_tool_frame(self, *, frame):
+            self.calls.append(("create", frame))
+            return 0
+
+        def rm_update_tool_frame(self, *, frame):
+            self.calls.append(("update", frame))
+            return 0
+
+        def rm_change_tool_frame(self, tool_name):
+            self.calls.append(("change", tool_name))
+            return 0
+
+    arm = FakeArm()
+    frame = object()
+
+    _configure_tool_frame(arm, frame, "omnipic")
+
+    assert arm.calls == [
+        ("get",),
+        (expected_operation, frame),
+        ("change", "omnipic"),
+    ]
+
+
+@pytest.mark.parametrize(
+    ("existing", "failure", "operation"),
+    [
+        (False, "query", "rm_get_total_tool_frame"),
+        (False, "create", "rm_set_manual_tool_frame"),
+        (True, "update", "rm_update_tool_frame"),
+        (False, "change", "rm_change_tool_frame"),
+    ],
+)
+def test_tool_frame_sdk_failures_are_reported(existing, failure, operation) -> None:
+    class FakeArm:
+        def rm_get_total_tool_frame(self):
+            names = ["omnipic"] if existing else []
+            return {
+                "return_code": 1 if failure == "query" else 0,
+                "tool_names": names,
+            }
+
+        def rm_set_manual_tool_frame(self, *, frame):
+            del frame
+            return 1 if failure == "create" else 0
+
+        def rm_update_tool_frame(self, *, frame):
+            del frame
+            return 1 if failure == "update" else 0
+
+        def rm_change_tool_frame(self, tool_name):
+            del tool_name
+            return 1 if failure == "change" else 0
+
+    with pytest.raises(RuntimeError, match=operation):
+        _configure_tool_frame(FakeArm(), object(), "omnipic")
+
+
+def _udp_state(
+    *,
+    robot_ip: str = "127.0.0.1",
+    joints=None,
+    error_code: int = 0,
+    joint_enabled=None,
+    joint_error_codes=None,
+    arm_error_codes=None,
+    arm_current_status: int = 0,
+):
+    if joints is None:
+        joints = [0.0, 10.0, -20.0, 30.0, -40.0, 50.0, -60.0]
+    if joint_enabled is None:
+        joint_enabled = [True] * 7
+    if joint_error_codes is None:
+        joint_error_codes = [0] * 7
+    if arm_error_codes is None:
+        arm_error_codes = []
+    return SimpleNamespace(
+        errCode=error_code,
+        arm_ip=robot_ip.encode(),
+        joint_status=SimpleNamespace(
+            joint_position=joints,
+            joint_en_flag=joint_enabled,
+            joint_err_code=joint_error_codes,
+        ),
+        err=SimpleNamespace(
+            err_len=len(arm_error_codes),
+            err=list(arm_error_codes),
+        ),
+        arm_current_status=arm_current_status,
+    )
+
+
+def _install_fake_sdk(monkeypatch, *, push_return=0, send_feedback=True):
+    class FakeThreadMode:
+        RM_TRIPLE_MODE_E = 2
+
+    class FakePushConfig:
+        def __init__(self, *args):
+            self.args = args
+
+    class FakeArm:
+        instance = None
+
+        def __init__(self, mode):
+            self.mode = mode
+            self.callback = None
+            self.config = None
+            self.delete_calls = 0
+            FakeArm.instance = self
+
+        def rm_create_robot_arm(self, robot_ip, robot_port):
+            self.robot_ip = robot_ip
+            self.robot_port = robot_port
+            return SimpleNamespace(id=1)
+
+        def rm_realtime_arm_state_call_back(self, callback):
+            self.callback = callback
+
+        def rm_set_realtime_push(self, config):
+            self.config = config
+            if push_return == 0 and send_feedback:
+                self.callback(_udp_state())
+            return push_return
+
+        def rm_delete_robot_arm(self):
+            self.delete_calls += 1
+            return 0
+
+    sdk = ModuleType("Robotic_Arm.rm_robot_interface")
+    sdk.RoboticArm = FakeArm
+    sdk.rm_thread_mode_e = FakeThreadMode
+    sdk.rm_realtime_push_config_t = FakePushConfig
+    sdk.rm_realtime_arm_state_callback_ptr = lambda callback: callback
+    package = ModuleType("Robotic_Arm")
+    package.rm_robot_interface = sdk
+    monkeypatch.setitem(sys.modules, "Robotic_Arm", package)
+    monkeypatch.setitem(sys.modules, "Robotic_Arm.rm_robot_interface", sdk)
+    return SimpleNamespace(RoboticArm=FakeArm)
+
+
+def test_joint_degree_query_returns_validated_radians() -> None:
+    class FakeArm:
+        def rm_get_joint_degree(self):
+            return 0, [0.0, 10.0, -20.0, 30.0, -40.0, 50.0, -60.0]
+
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._arm = FakeArm()
+
+    snapshot = adapter.read_joint_state()
+
+    assert snapshot.positions == pytest.approx(
+        [math.radians(value) for value in [0, 10, -20, 30, -40, 50, -60]]
+    )
+    assert snapshot.read_duration_ms is not None
+
+
+@pytest.mark.parametrize(
+    "result",
+    [
+        (7, [0.0] * 7),
+        (0, [0.0] * 6),
+        (0, [0.0, 0.0, 0.0, math.nan, 0.0, 0.0, 0.0]),
+    ],
+)
+def test_joint_degree_query_rejects_sdk_errors_and_invalid_values(result) -> None:
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._arm = SimpleNamespace(rm_get_joint_degree=lambda: result)
+
+    with pytest.raises((RuntimeError, ValueError)):
+        adapter.read_joint_state()
+
+
+def test_mock_joint_query_uses_current_mock_positions() -> None:
+    adapter = MockRealManAdapter([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0])
+
+    snapshot = adapter.read_joint_state()
+
+    assert snapshot.positions == pytest.approx(
+        [math.radians(value) for value in [1, 2, 3, 4, 5, 6, 7]]
+    )
+
+
+def test_udp_feedback_is_cached_in_radians(monkeypatch) -> None:
+    monotonic = iter([10.0, 10.005])
+    monkeypatch.setattr(realman_adapter.time, "monotonic", lambda: next(monotonic))
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._accept_realtime_feedback = True
+
+    adapter._on_realtime_arm_state(_udp_state())
+    first = adapter.get_latest_joint_state()
+    adapter._on_realtime_arm_state(_udp_state())
+    second = adapter.get_latest_joint_state()
+
+    assert first is not None
+    assert first.positions == pytest.approx(
+        [math.radians(value) for value in [0, 10, -20, 30, -40, 50, -60]]
+    )
+    assert first.read_duration_ms is None
+    assert first.update_interval_ms is None
+    assert second is not None
+    assert second.read_duration_ms is None
+    assert second.update_interval_ms == pytest.approx(5.0)
+
+
+@pytest.mark.parametrize(
+    "state",
+    [
+        _udp_state(error_code=-3),
+        _udp_state(robot_ip="192.168.192.18"),
+        _udp_state(joints=[0.0] * 6),
+        _udp_state(joints=[0.0, 0.0, 0.0, math.nan, 0.0, 0.0, 0.0]),
+    ],
+)
+def test_invalid_udp_feedback_does_not_replace_snapshot(state) -> None:
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._accept_realtime_feedback = True
+    adapter._on_realtime_arm_state(_udp_state(joints=[1.0] * 7))
+    before = adapter.get_latest_joint_state()
+
+    adapter._on_realtime_arm_state(state)
+
+    after = adapter.get_latest_joint_state()
+    assert after is not None
+    assert before is not None
+    assert after.positions == before.positions
+    assert after.received_at == before.received_at
+    assert after.motion_ready is False
+
+
+def test_udp_joint_fault_marks_snapshot_unready() -> None:
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._accept_realtime_feedback = True
+
+    adapter._on_realtime_arm_state(
+        _udp_state(
+            joint_enabled=[True, True, False, True, True, True, True],
+            joint_error_codes=[0, 0, 17, 0, 0, 0, 0],
+            arm_error_codes=[42],
+            arm_current_status=9,
+        )
+    )
+
+    snapshot = adapter.get_latest_joint_state()
+    assert snapshot is not None
+    assert snapshot.motion_ready is False
+
+
+def test_udp_stop_status_marks_snapshot_unready_without_joint_error() -> None:
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._accept_realtime_feedback = True
+
+    adapter._on_realtime_arm_state(_udp_state(arm_current_status=9))
+
+    snapshot = adapter.get_latest_joint_state()
+    assert snapshot is not None
+    assert snapshot.motion_ready is False
+
+
+def test_udp_zero_arm_error_code_is_motion_ready() -> None:
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._accept_realtime_feedback = True
+
+    adapter._on_realtime_arm_state(_udp_state(arm_error_codes=[0]))
+
+    snapshot = adapter.get_latest_joint_state()
+    assert snapshot is not None
+    assert snapshot.motion_ready is True
+
+
+def test_udp_fault_and_recovery_are_logged_once_per_transition() -> None:
+    class FakeLogger:
+        def __init__(self) -> None:
+            self.infos = []
+            self.warnings = []
+
+        def info(self, message):
+            self.infos.append(message)
+
+        def warn(self, message):
+            self.warnings.append(message)
+
+    logger = FakeLogger()
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+        logger=logger,
+    )
+    adapter._accept_realtime_feedback = True
+    adapter._on_realtime_arm_state(_udp_state())
+    fault = _udp_state(
+        joint_enabled=[False] * 7,
+        joint_error_codes=[17, 0, 0, 0, 0, 0, 0],
+        arm_error_codes=[42],
+        arm_current_status=9,
+    )
+
+    adapter._on_realtime_arm_state(fault)
+    adapter._on_realtime_arm_state(fault)
+    adapter._on_realtime_arm_state(_udp_state())
+
+    assert len(logger.warnings) == 1
+    assert "joint_errors=[17, 0, 0, 0, 0, 0, 0]" in logger.warnings[0]
+    assert len([message for message in logger.infos if "恢复正常" in message]) == 1
+
+
+@pytest.mark.parametrize(
+    ("cycle_ms", "sdk_cycle"),
+    [(5, 1), (10, 2)],
+)
+def test_connect_converts_udp_feedback_cycle_to_sdk_units(
+    monkeypatch,
+    cycle_ms,
+    sdk_cycle,
+) -> None:
+    fake_sdk = _install_fake_sdk(monkeypatch)
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "192.168.192.148",
+        8090,
+        realtime_push_cycle_ms=cycle_ms,
+        configure_safety_limits=False,
+    )
+
+    adapter.connect()
+
+    arm = fake_sdk.RoboticArm.instance
+    assert arm is not None
+    assert arm.config.args == (
+        sdk_cycle,
+        True,
+        8090,
+        0,
+        "192.168.192.148",
+    )
+    assert arm.callback is adapter._realtime_callback
+    assert adapter.get_latest_joint_state() is not None
+    assert not hasattr(adapter, "_feedback_thread")
+
+
+def test_udp_configuration_failure_cleans_up_robot_handle(monkeypatch) -> None:
+    fake_sdk = _install_fake_sdk(monkeypatch, push_return=1)
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "192.168.192.148",
+        8090,
+        configure_safety_limits=False,
+    )
+
+    with pytest.raises(RuntimeError, match="rm_set_realtime_push"):
+        adapter.connect()
+
+    assert fake_sdk.RoboticArm.instance.delete_calls == 1
+    assert adapter._arm is None
+
+
+def test_udp_first_frame_timeout_cleans_up_robot_handle(monkeypatch) -> None:
+    fake_sdk = _install_fake_sdk(monkeypatch, send_feedback=False)
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "192.168.192.148",
+        8090,
+        configure_safety_limits=False,
+    )
+    adapter._feedback_ready = SimpleNamespace(
+        clear=lambda: None,
+        set=lambda: None,
+        wait=lambda timeout: False,
+    )
+
+    with pytest.raises(RuntimeError, match="within 2 seconds"):
+        adapter.connect()
+
+    assert fake_sdk.RoboticArm.instance.delete_calls == 1
+    assert adapter._arm is None
+
+
+def test_joint_target_uses_movej_canfd_in_degrees() -> None:
+    class FakeArm:
+        def __init__(self) -> None:
+            self.calls = []
+
+        def rm_movej_canfd(self, *args):
+            self.calls.append(args)
+            return 0
+
+    adapter = RealManAdapter(
+        "127.0.0.1",
+        8080,
+        0,
+        "127.0.0.1",
+        8090,
+    )
+    adapter._arm = FakeArm()
+    target = [math.radians(value) for value in [1, 2, 3, 4, 5, 6, 7]]
+
+    adapter.send_joint_target(target, follow=False)
+
+    assert len(adapter._arm.calls) == 1
+    degrees, follow, expand, trajectory_mode, radio = adapter._arm.calls[0]
+    assert degrees == pytest.approx([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0])
+    assert (follow, expand, trajectory_mode, radio) == (False, 0, 2, 0)
+
+
+def test_mock_joint_feedback_is_available_without_vendor_sdk() -> None:
+    adapter = MockRealManAdapter([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0])
+
+    adapter.connect()
+    snapshot = adapter.get_latest_joint_state()
+
+    assert snapshot is not None
+    assert snapshot.positions == pytest.approx(
+        [math.radians(value) for value in [1, 2, 3, 4, 5, 6, 7]]
+    )

xr_rm_teleop/test/test_joint_control.py

@@ -0,0 +1,601 @@
+import math
+import time
+from types import SimpleNamespace
+
+import numpy as np
+import pytest
+
+from xr_rm_teleop.realman_adapter import JointStateSnapshot
+from xr_rm_teleop.single_arm_velocity_teleop import (
+    SingleArmVelocityTeleop,
+    _make_transform,
+    _so3_exp,
+)
+
+
+class FakeLogger:
+    def info(self, *args, **kwargs):
+        del args, kwargs
+
+    def warn(self, *args, **kwargs):
+        del args, kwargs
+
+    def error(self, *args, **kwargs):
+        del args, kwargs
+
+
+class FakeTime:
+    def __sub__(self, other):
+        del other
+        return SimpleNamespace(nanoseconds=0)
+
+
+def test_startup_joint_query_initializes_qp_and_command_history() -> None:
+    positions = [0.1] * 7
+    pose = np.eye(4)
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._arm_name = "right_rm75"
+    teleop._adapter = SimpleNamespace(
+        read_joint_state=lambda: JointStateSnapshot(
+            positions,
+            time.monotonic(),
+        )
+    )
+    teleop._ik_solver = SimpleNamespace(
+        update_joint_state=lambda joints: pose
+    )
+    teleop.get_logger = lambda: FakeLogger()
+
+    teleop._initialize_joint_state()
+
+    assert teleop._latest_joint_positions == positions
+    assert teleop._last_valid_joint_target == positions
+    assert teleop._last_joint_command_target == positions
+    assert teleop._last_joint_command_velocity == [0.0] * 7
+    assert teleop._last_current_pose is pose
+
+
+def test_startup_joint_query_failure_closes_adapter() -> None:
+    class FailingAdapter:
+        def __init__(self):
+            self.close_calls = 0
+
+        def read_joint_state(self):
+            raise RuntimeError("rm_get_joint_degree failed with code 7")
+
+        def close(self):
+            self.close_calls += 1
+
+    errors = []
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._arm_name = "left_rm75"
+    teleop._adapter = FailingAdapter()
+    teleop.get_logger = lambda: SimpleNamespace(
+        error=lambda message: errors.append(message)
+    )
+
+    with pytest.raises(RuntimeError, match="code 7"):
+        teleop._initialize_joint_state()
+
+    assert teleop._adapter.close_calls == 1
+    assert "left_rm75" in errors[0]
+    assert "启动关节同步失败" in errors[0]
+
+
+def _timeout_teleop(adapter) -> SingleArmVelocityTeleop:
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._adapter = adapter
+    teleop._arm_name = "right_rm75"
+    teleop._follow = False
+    teleop._active = True
+    teleop._joint_feedback_ready = True
+    teleop._grip_rearm_required = False
+    teleop._feedback_resync_attempted = False
+    teleop._control_fault_latched = False
+    teleop._last_joint_command_target = [0.1] * 7
+    teleop._last_joint_command_velocity = [0.0] * 7
+    teleop._latest_joint_positions = [0.1] * 7
+    teleop._last_valid_joint_target = [0.1] * 7
+    teleop._last_current_pose = np.eye(4)
+    teleop._controller_start = None
+    teleop._controller_orientation_start = None
+    teleop._robot_start_transform = None
+    teleop._filtered_target = None
+    teleop._filtered_orientation_target = None
+    teleop._last_sent_target = None
+    teleop._last_sent_orientation = None
+    teleop._last_command_time = None
+    teleop._ik_solver = SimpleNamespace(
+        update_joint_state=lambda joints: np.eye(4)
+    )
+    teleop._stop_sent = False
+    teleop._feedback_resync_timeout_sec = 0.5
+    teleop._publish_stop_debug = lambda: None
+    teleop.get_logger = lambda: FakeLogger()
+    return teleop
+
+
+def test_missing_or_disabled_joint_snapshot_is_not_motion_ready() -> None:
+    assert not SingleArmVelocityTeleop._joint_snapshot_is_motion_ready(None)
+    assert not SingleArmVelocityTeleop._joint_snapshot_is_motion_ready(
+        JointStateSnapshot(
+            [0.0] * 7,
+            time.monotonic(),
+            motion_ready=False,
+        )
+    )
+
+
+def test_short_udp_timeout_repeats_last_limited_target_without_query() -> None:
+    sends = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: sends.append(
+            (list(joints), follow)
+        ),
+        read_joint_state=lambda: pytest.fail("query must not run"),
+        stop=lambda: pytest.fail("stop must not run"),
+    )
+    teleop = _timeout_teleop(adapter)
+    warnings = []
+    teleop.get_logger = lambda: SimpleNamespace(
+        warn=lambda message: warnings.append(message)
+    )
+
+    teleop._handle_stale_joint_feedback(0.2)
+
+    assert sends == [([0.1] * 7, False)]
+    assert teleop._last_joint_command_target == [0.1] * 7
+    assert teleop._grip_rearm_required
+    assert warnings == [
+        "right_rm75 UDP关节反馈超时（age=200.0 ms），保持最后安全目标。"
+    ]
+
+
+def test_short_udp_timeout_without_active_target_stays_stopped() -> None:
+    stop_calls = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: pytest.fail(
+            "inactive control must not start CANFD output"
+        ),
+        read_joint_state=lambda: pytest.fail("query must not run"),
+        stop=lambda: stop_calls.append(True),
+    )
+    teleop = _timeout_teleop(adapter)
+    teleop._active = False
+
+    teleop._handle_stale_joint_feedback(0.2)
+
+    assert len(stop_calls) == 1
+
+
+def test_persistent_udp_timeout_queries_once_and_holds_actual_position() -> None:
+    sends = []
+    query_calls = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: sends.append(list(joints)),
+        read_joint_state=lambda: (
+            query_calls.append(True)
+            or JointStateSnapshot([0.2] * 7, time.monotonic())
+        ),
+        stop=lambda: None,
+    )
+    teleop = _timeout_teleop(adapter)
+
+    teleop._handle_stale_joint_feedback(0.5)
+    teleop._handle_stale_joint_feedback(0.6)
+
+    assert len(query_calls) == 1
+    assert sends == [[0.2] * 7, [0.2] * 7]
+    assert teleop._last_valid_joint_target == [0.2] * 7
+    assert teleop._last_joint_command_velocity == [0.0] * 7
+
+
+def test_persistent_udp_timeout_query_failure_latches_control() -> None:
+    stop_calls = []
+    adapter = SimpleNamespace(
+        send_joint_target=lambda joints, follow: pytest.fail(
+            "CANFD must stop after query failure"
+        ),
+        read_joint_state=lambda: (_ for _ in ()).throw(
+            RuntimeError("rm_get_joint_degree failed with code 7")
+        ),
+        stop=lambda: stop_calls.append(True),
+    )
+    teleop = _timeout_teleop(adapter)
+
+    teleop._handle_stale_joint_feedback(0.5)
+    teleop._handle_stale_joint_feedback(0.6)
+
+    assert teleop._control_fault_latched
+    assert len(stop_calls) == 1
+
+
+def test_joint_command_step_limits_acceleration_from_rest() -> None:
+    dt = 1.0 / 125.0
+    target, velocity = SingleArmVelocityTeleop._limit_joint_command_step(
+        target=[0.2] * 7,
+        previous_target=[0.0] * 7,
+        previous_velocity=[0.0] * 7,
+        max_speed=math.radians(180.0),
+        max_acceleration=math.radians(300.0),
+        dt=dt,
+    )
+
+    assert velocity == pytest.approx([math.radians(2.4)] * 7)
+    assert target == pytest.approx([math.radians(0.0192)] * 7)
+
+
+def test_joint_command_step_rejects_non_finite_limits() -> None:
+    for max_speed, max_acceleration, dt in (
+        (math.inf, 1.0, 0.1),
+        (1.0, math.inf, 0.1),
+        (1.0, 1.0, math.inf),
+    ):
+        with pytest.raises(ValueError, match="finite and positive"):
+            SingleArmVelocityTeleop._limit_joint_command_step(
+                target=[0.5] * 7,
+                previous_target=[0.0] * 7,
+                previous_velocity=[0.0] * 7,
+                max_speed=max_speed,
+                max_acceleration=max_acceleration,
+                dt=dt,
+            )
+
+
+def test_joint_command_step_arrival_respects_max_speed() -> None:
+    target, velocity = SingleArmVelocityTeleop._limit_joint_command_step(
+        target=[0.5] * 7,
+        previous_target=[0.0] * 7,
+        previous_velocity=[0.0] * 7,
+        max_speed=1.0,
+        max_acceleration=100.0,
+        dt=0.1,
+    )
+
+    assert velocity == pytest.approx([1.0] * 7)
+    assert target == pytest.approx([0.1] * 7)
+
+
+def test_joint_command_step_reverses_with_acceleration_limit() -> None:
+    command = [0.0] * 7
+    velocity = [0.0] * 7
+    for _ in range(5):
+        command, velocity = SingleArmVelocityTeleop._limit_joint_command_step(
+            target=[1.0] * 7,
+            previous_target=command,
+            previous_velocity=velocity,
+            max_speed=1.0,
+            max_acceleration=1.0,
+            dt=0.1,
+        )
+
+    previous_command = list(command)
+    previous_velocity = list(velocity)
+    command, velocity = SingleArmVelocityTeleop._limit_joint_command_step(
+        target=[-1.0] * 7,
+        previous_target=command,
+        previous_velocity=velocity,
+        max_speed=1.0,
+        max_acceleration=1.0,
+        dt=0.1,
+    )
+
+    assert previous_velocity == pytest.approx([0.5] * 7)
+    assert velocity == pytest.approx([0.4] * 7)
+    assert [
+        current - previous
+        for current, previous in zip(command, previous_command)
+    ] == pytest.approx([value * 0.1 for value in velocity])
+    assert all(
+        current > previous
+        for current, previous in zip(command, previous_command)
+    )
+
+
+def test_joint_command_step_brakes_before_fixed_target_without_overshoot() -> None:
+    dt = 1.0 / 90.0
+    max_speed = math.radians(180.0)
+    max_acceleration = math.radians(300.0)
+    target = np.radians(
+        [10.0, -10.0, 3.0, -3.0, 1.0, -1.0, 0.1]
+    ).tolist()
+    command = [0.0] * 7
+    velocity = [0.0] * 7
+
+    for _ in range(180):
+        previous_command = list(command)
+        previous_velocity = list(velocity)
+        command, velocity = (
+            SingleArmVelocityTeleop._limit_joint_command_step(
+                target=target,
+                previous_target=command,
+                previous_velocity=velocity,
+                max_speed=max_speed,
+                max_acceleration=max_acceleration,
+                dt=dt,
+            )
+        )
+
+        for index in range(7):
+            assert min(0.0, target[index]) - 1e-12 <= command[index]
+            assert command[index] <= max(0.0, target[index]) + 1e-12
+            assert abs(velocity[index]) <= max_speed + 1e-12
+            assert (
+                abs(velocity[index] - previous_velocity[index])
+                <= max_acceleration * dt + 1e-12
+            )
+            assert command[index] - previous_command[index] == pytest.approx(
+                velocity[index] * dt,
+                abs=1e-12,
+            )
+
+    assert command == pytest.approx(target, abs=1e-12)
+    assert velocity == pytest.approx([0.0] * 7, abs=1e-12)
+
+
+def test_feedback_fault_blocks_grip_until_release() -> None:
+    class FakeClock:
+        def now(self):
+            return FakeTime()
+
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._adapter = SimpleNamespace(
+        get_latest_joint_state=lambda: JointStateSnapshot(
+            [0.1] * 7,
+            time.monotonic(),
+        )
+    )
+    teleop._command_timeout_sec = 0.12
+    teleop._joint_feedback_ready = True
+    teleop._arm_name = "right_rm75"
+    teleop._last_msg = SimpleNamespace(
+        grip=True,
+        pose=SimpleNamespace(
+            position=SimpleNamespace(x=0.0, y=0.0, z=0.0),
+            orientation=SimpleNamespace(x=0.0, y=0.0, z=0.0, w=1.0),
+        ),
+    )
+    teleop._last_msg_time = FakeTime()
+    teleop._active = False
+    teleop._enable_orientation_control = False
+    teleop._last_valid_joint_target = None
+    teleop._last_current_pose = None
+    teleop._ik_solver = SimpleNamespace(
+        update_joint_state=lambda joints: np.eye(4)
+    )
+    teleop._grip_rearm_required = True
+    teleop._control_fault_latched = False
+    teleop._feedback_resync_attempted = False
+    teleop.get_clock = lambda: FakeClock()
+    teleop.get_logger = lambda: FakeLogger()
+    stopped = []
+    entered = []
+    teleop._safe_stop = lambda reset_active: stopped.append(reset_active)
+    teleop._enter_active_control = lambda *args: entered.append(args)
+
+    teleop._control_tick()
+    assert entered == []
+
+    teleop._last_msg.grip = False
+    teleop._control_tick()
+    assert teleop._grip_rearm_required is False
+
+    teleop._last_msg.grip = True
+    teleop._control_tick()
+    assert len(entered) == 1
+
+
+def test_first_feedback_initializes_last_valid_target_without_solving() -> None:
+    class FakeSolver:
+        def __init__(self) -> None:
+            self.solve_calls = 0
+
+        def update_joint_state(self, joints):
+            assert joints == [0.1] * 7
+            transform = np.eye(4)
+            transform[:3, 3] = [0.3, 0.0, 0.2]
+            return transform
+
+        def solve(self, target):
+            del target
+            self.solve_calls += 1
+            return [0.2] * 7
+
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._ik_solver = FakeSolver()
+    teleop._active = False
+    teleop._last_valid_joint_target = None
+    teleop._last_current_pose = None
+
+    pose = teleop._sync_joint_feedback(
+        JointStateSnapshot([0.1] * 7, time.monotonic())
+    )
+
+    assert pose == pytest.approx(
+        _make_transform([0.3, 0.0, 0.2], np.eye(3))
+    )
+    assert teleop._last_valid_joint_target == [0.1] * 7
+    assert teleop._ik_solver.solve_calls == 0
+
+
+def test_qp_failure_returns_last_known_good_target() -> None:
+    class FailingSolver:
+        def solve(self, target):
+            del target
+            raise RuntimeError("NaN in QP solution")
+
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._ik_solver = FailingSolver()
+    teleop._last_valid_joint_target = [0.1] * 7
+    teleop._arm_name = "right_rm75"
+    teleop.get_logger = lambda: FakeLogger()
+
+    target = teleop._solve_joint_target(np.eye(4))
+
+    assert target == pytest.approx([0.1] * 7)
+    assert teleop._last_valid_joint_target == pytest.approx([0.1] * 7)
+
+
+def test_qp_success_updates_last_known_good_target() -> None:
+    class SuccessfulSolver:
+        def solve(self, target):
+            del target
+            return [0.2] * 7
+
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._ik_solver = SuccessfulSolver()
+    teleop._last_valid_joint_target = [0.1] * 7
+    teleop._arm_name = "left_rm75"
+    teleop.get_logger = lambda: FakeLogger()
+
+    target = teleop._solve_joint_target(np.eye(4))
+
+    assert target == pytest.approx([0.2] * 7)
+    assert teleop._last_valid_joint_target == pytest.approx([0.2] * 7)
+
+
+def test_enter_active_control_initializes_se3_orientation_state() -> None:
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    transform = _make_transform(
+        [0.3, -0.1, 0.2],
+        _so3_exp(np.asarray([0.1, -0.2, 0.3])),
+    )
+    published = []
+    teleop._arm_name = "right_rm75"
+    teleop.get_logger = lambda: FakeLogger()
+    teleop._publish_debug = lambda *args: published.append(args)
+
+    teleop._enter_active_control(
+        [0.0, 0.0, 0.0],
+        (0.0, 0.0, 0.0, 1.0),
+        transform,
+        FakeTime(),
+    )
+
+    assert teleop._robot_start_transform == pytest.approx(transform)
+    assert teleop._filtered_target == pytest.approx(transform[:3, 3])
+    assert teleop._filtered_orientation_target == pytest.approx(transform[:3, :3])
+    assert teleop._last_sent_orientation == pytest.approx(transform[:3, :3])
+    assert len(published) == 1
+
+
+def test_command_angular_velocity_uses_so3_rotation_vector() -> None:
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._dt = 0.1
+    teleop._last_sent_target = [0.0, 0.0, 0.0]
+    teleop._last_sent_orientation = np.eye(3)
+    teleop._last_command_time = None
+
+    velocity = teleop._estimate_command_velocity(
+        [0.0, 0.0, 0.0],
+        _so3_exp(np.asarray([0.0, 0.0, 0.1])),
+        FakeTime(),
+    )
+
+    assert velocity == pytest.approx([0.0, 0.0, 0.0, 0.0, 0.0, 1.0])
+
+
+def test_timing_stats_logs_summary_and_clears_window() -> None:
+    messages = []
+    teleop = object.__new__(SingleArmVelocityTeleop)
+    teleop._arm_name = "right_rm75"
+    teleop._dt = 0.008
+    teleop._timing_stats_window = 3
+    teleop._timing_samples = {
+        name: []
+        for name in (
+            "period",
+            "total",
+            "qp",
+            "send",
+            "feedback_age",
+            "feedback_read",
+            "feedback_interval",
+        )
+    }
+    teleop._last_timing_feedback_received_at = None
+    teleop.get_logger = lambda: SimpleNamespace(
+        info=lambda message: messages.append(message)
+    )
+    first_feedback = JointStateSnapshot([0.0] * 7, 10.0, 2.0, None)
+    second_feedback = JointStateSnapshot([0.0] * 7, 10.011, 3.0, 11.0)
+
+    teleop._record_timing_sample(7.0, 6.0, 1.0, 0.5, 3.0, first_feedback)
+    teleop._record_timing_sample(8.0, 8.0, 1.5, 0.6, 3.5, first_feedback)
+    assert messages == []
+
+    teleop._record_timing_sample(9.0, 10.0, 2.0, 0.7, 4.0, second_feedback)
+
+    assert len(messages) == 1
+    assert "right_rm75 timing n=3 deadline=8.000 ms" in messages[0]
+    assert (
+        "period[n=3 mean=8.000 p95=8.900 p99=8.980 "
+        "max=9.000 ms overruns=1]"
+    ) in messages[0]
+    assert (
+        "total[n=3 mean=8.000 p95=9.800 p99=9.960 "
+        "max=10.000 ms overruns=1]"
+    ) in messages[0]
+    assert "qp[n=3" in messages[0]
+    assert "send[n=3" in messages[0]
+    assert "feedback_age[n=3" in messages[0]
+    assert "feedback_read[n=2" in messages[0]
+    assert "feedback_interval[n=1" in messages[0]
+    assert all(not samples for samples in teleop._timing_samples.values())
+
+
+def test_canfd_error_stops_queries_and_requires_grip_rearm() -> None:
+    class RecoveringAdapter:
+        def __init__(self):
+            self.stop_calls = 0
+            self.read_calls = 0
+
+        def send_joint_target(self, joints, follow):
+            del joints, follow
+            raise RuntimeError("rm_movej_canfd failed with code 9")
+
+        def stop(self):
+            self.stop_calls += 1
+
+        def read_joint_state(self):
+            self.read_calls += 1
+            return JointStateSnapshot([0.2] * 7, time.monotonic())
+
+    teleop = _timeout_teleop(RecoveringAdapter())
+    teleop._joint_command_max_speed = math.radians(180.0)
+    teleop._joint_command_max_acceleration = math.radians(300.0)
+    teleop._dt = 1.0 / 90.0
+
+    sent = teleop._send_joint_target([0.3] * 7)
+
+    assert not sent
+    assert teleop._adapter.stop_calls == 1
+    assert teleop._adapter.read_calls == 1
+    assert not teleop._control_fault_latched
+    assert teleop._grip_rearm_required
+    assert teleop._last_joint_command_target == [0.2] * 7
+
+
+def test_canfd_error_latches_when_joint_query_also_fails() -> None:
+    class FailingAdapter:
+        def __init__(self):
+            self.stop_calls = 0
+
+        def send_joint_target(self, joints, follow):
+            del joints, follow
+            raise RuntimeError("rm_movej_canfd failed with code 9")
+
+        def stop(self):
+            self.stop_calls += 1
+
+        def read_joint_state(self):
+            raise RuntimeError("rm_get_joint_degree failed with code 7")
+
+    teleop = _timeout_teleop(FailingAdapter())
+    teleop._joint_command_max_speed = math.radians(180.0)
+    teleop._joint_command_max_acceleration = math.radians(300.0)
+    teleop._dt = 1.0 / 90.0
+
+    assert not teleop._send_joint_target([0.3] * 7)
+    assert teleop._control_fault_latched
+    assert teleop._adapter.stop_calls == 1

xr_rm_teleop/test/test_orientation_control.py

@@ -1,14 +1,20 @@
 import math
+import time
 from types import SimpleNamespace
 
+import numpy as np
 import pytest
 
-from xr_rm_teleop.realman_adapter import ArmPose, MockRealManAdapter
+from xr_rm_teleop.realman_adapter import JointStateSnapshot
 from xr_rm_teleop.single_arm_velocity_teleop import (
     SingleArmVelocityTeleop,
-    _euler_to_quaternion,
+    _make_transform,
+    _matrix_to_quaternion,
     _normalize_quaternion,
-    _quaternion_to_euler,
+    _project_rotation,
+    _quaternion_to_matrix,
+    _so3_exp,
+    _so3_log,
 )
 
 
@@ -17,7 +23,10 @@ def _make_teleop_for_orientation() -> SingleArmVelocityTeleop:
     teleop._enable_orientation_control = True
     teleop._enable_orientation_axes = [True, True, True]
     teleop._controller_orientation_start = (0.0, 0.0, 0.0, 1.0)
-    teleop._robot_start_pose = ArmPose(0.3, 0.0, 0.2, 0.1, -0.2, 0.3)
+    teleop._robot_start_transform = _make_transform(
+        [0.3, 0.0, 0.2],
+        _so3_exp(np.asarray([0.1, -0.2, 0.3])),
+    )
     teleop._xr_to_robot_matrix = [
         0.0, 1.0, 0.0,
         0.0, 0.0, 1.0,
@@ -26,47 +35,111 @@ def _make_teleop_for_orientation() -> SingleArmVelocityTeleop:
     return teleop
 
 
-def assert_angles_close(actual: list[float] | tuple[float, ...], expected: list[float]) -> None:
-    assert len(actual) == len(expected)
-    for actual_value, expected_value in zip(actual, expected):
-        assert math.atan2(math.sin(actual_value - expected_value), math.cos(actual_value - expected_value)) == pytest.approx(0.0)
-
-
 def test_identity_controller_orientation_keeps_tcp_orientation() -> None:
     teleop = _make_teleop_for_orientation()
 
     target = teleop._raw_orientation_from_controller((0.0, 0.0, 0.0, 1.0))
 
-    assert_angles_close(target, teleop._robot_start_pose.rpy())
+    assert target == pytest.approx(teleop._robot_start_transform[:3, :3])
 
 
 def test_xr_relative_rotation_maps_through_xr_to_robot_matrix() -> None:
     teleop = _make_teleop_for_orientation()
-    teleop._robot_start_pose = ArmPose(0.3, 0.0, 0.2, 0.0, 0.0, 0.0)
-    xr_roll = _euler_to_quaternion(0.2, 0.0, 0.0)
+    teleop._robot_start_transform = np.eye(4)
+    xr_roll = _matrix_to_quaternion(_so3_exp(np.asarray([0.2, 0.0, 0.0])))
 
     target = teleop._raw_orientation_from_controller(xr_roll)
 
-    assert_angles_close(target, [0.0, 0.0, 0.2])
+    assert _so3_log(target) == pytest.approx([0.0, 0.0, 0.2])
 
 
-def test_orientation_deadband_filter_and_speed_limit() -> None:
+def test_quaternion_sign_does_not_change_rotation() -> None:
+    quaternion = _normalize_quaternion((0.2, -0.3, 0.1, 0.9))
+
+    assert _quaternion_to_matrix(quaternion) == pytest.approx(
+        _quaternion_to_matrix(tuple(-value for value in quaternion))
+    )
+
+
+@pytest.mark.parametrize("pitch", [math.pi / 2.0 - 1e-5, -math.pi / 2.0 + 1e-5])
+def test_small_rotation_near_gimbal_lock_stays_small(pitch: float) -> None:
+    teleop = _make_teleop_for_orientation()
+    start_rotation = _so3_exp(np.asarray([0.0, pitch, 0.0]))
+    teleop._robot_start_transform = _make_transform([0.3, 0.0, 0.2], start_rotation)
+    teleop._xr_to_robot_matrix = np.eye(3).reshape(-1).tolist()
+    controller = _matrix_to_quaternion(_so3_exp(np.asarray([0.01, 0.0, 0.0])))
+
+    target = teleop._raw_orientation_from_controller(controller)
+
+    error = _so3_log(target @ start_rotation.T)
+    assert np.linalg.norm(error) == pytest.approx(0.01)
+
+
+def test_crossing_old_rpy_branch_uses_shortest_rotation() -> None:
+    teleop = _make_teleop_for_orientation()
+    start_rotation = _so3_exp(np.asarray([0.0, math.pi / 2.0 - 0.001, 0.0]))
+    teleop._robot_start_transform = _make_transform([0.3, 0.0, 0.2], start_rotation)
+    teleop._xr_to_robot_matrix = np.eye(3).reshape(-1).tolist()
+    controller = _matrix_to_quaternion(_so3_exp(np.asarray([0.0, 0.002, 0.0])))
+
+    target = teleop._raw_orientation_from_controller(controller)
+
+    assert _so3_log(target @ start_rotation.T) == pytest.approx(
+        [0.0, 0.002, 0.0],
+        abs=1e-9,
+    )
+
+
+def test_disabled_orientation_axis_zeros_robot_rotation_vector_component() -> None:
+    teleop = _make_teleop_for_orientation()
+    teleop._robot_start_transform = np.eye(4)
+    teleop._xr_to_robot_matrix = np.eye(3).reshape(-1).tolist()
+    teleop._enable_orientation_axes = [True, False, True]
+    controller = _matrix_to_quaternion(_so3_exp(np.asarray([0.1, 0.2, 0.3])))
+
+    target = teleop._raw_orientation_from_controller(controller)
+
+    assert _so3_log(target) == pytest.approx([0.1, 0.0, 0.3])
+
+
+def test_orientation_deadband_filter_and_speed_limit_use_so3_angle() -> None:
     teleop = object.__new__(SingleArmVelocityTeleop)
     teleop._orientation_deadband_rad = 0.01
     teleop._orientation_filter_alpha = 0.5
     teleop._max_orientation_speed = 0.5
-    teleop._dt = 0.1
-    teleop._last_sent_orientation = [0.0, 0.0, 0.0]
-    teleop._filtered_orientation_target = [0.0, 0.0, 0.0]
+    teleop._dt = 1.0 / 125.0
+    teleop._last_sent_orientation = np.eye(3)
+    teleop._filtered_orientation_target = np.eye(3)
 
-    assert teleop._apply_orientation_deadband([0.001, 0.0, 0.0]) == [0.0, 0.0, 0.0]
+    inside_deadband = _so3_exp(np.asarray([0.006, 0.006, 0.0]))
+    assert teleop._apply_orientation_deadband(inside_deadband) == pytest.approx(np.eye(3))
 
-    filtered = teleop._filter_orientation_target([0.2, 0.0, 0.0])
-    assert_angles_close(filtered, [0.1, 0.0, 0.0])
+    target = _so3_exp(np.asarray([0.2, 0.0, 0.0]))
+    filtered = teleop._filter_orientation_target(target)
+    assert _so3_log(filtered) == pytest.approx([0.1, 0.0, 0.0])
 
-    limited, was_limited = teleop._limit_orientation_step([0.2, 0.0, 0.0])
+    limited, was_limited = teleop._limit_orientation_step(target)
     assert was_limited
-    assert_angles_close(limited, [0.05, 0.0, 0.0])
+    assert np.linalg.norm(_so3_log(limited)) == pytest.approx(0.5 / 125.0)
+
+
+def test_rotation_matrix_to_debug_quaternion_is_normalized() -> None:
+    quaternion = _matrix_to_quaternion(_so3_exp(np.asarray([0.2, -0.1, 0.3])))
+
+    assert np.isfinite(quaternion).all()
+    assert np.linalg.norm(quaternion) == pytest.approx(1.0)
+
+
+def test_rotation_projection_accepts_small_error_and_rejects_invalid_matrix() -> None:
+    near_rotation = np.eye(3)
+    near_rotation[0, 1] = 1e-5
+
+    projected = _project_rotation(near_rotation)
+
+    assert projected.T @ projected == pytest.approx(np.eye(3))
+    assert np.linalg.det(projected) == pytest.approx(1.0)
+    with pytest.raises(ValueError):
+        _project_rotation(np.diag([2.0, 1.0, 1.0]))
 
 
 def test_invalid_controller_quaternion_stops_current_tick() -> None:
@@ -95,6 +168,19 @@ def test_invalid_controller_quaternion_stops_current_tick() -> None:
     teleop._arm_name = "test_rm75"
     teleop._command_timeout_sec = 0.12
     teleop._enable_orientation_control = True
+    teleop._adapter = SimpleNamespace(
+        get_latest_joint_state=lambda: JointStateSnapshot(
+            [0.1] * 7,
+            time.monotonic(),
+        )
+    )
+    teleop._ik_solver = SimpleNamespace(update_joint_state=lambda joints: np.eye(4))
+    teleop._active = False
+    teleop._last_valid_joint_target = None
+    teleop._last_current_pose = None
+    teleop._joint_feedback_ready = True
+    teleop._control_fault_latched = False
+    teleop._feedback_resync_attempted = False
     stopped = []
     teleop.get_clock = lambda: FakeClock()
     teleop.get_logger = lambda: FakeLogger()
@@ -105,20 +191,6 @@ def test_invalid_controller_quaternion_stops_current_tick() -> None:
     assert stopped == [True]
 
 
-def test_quaternion_roundtrip_for_small_rpy() -> None:
-    quat = _normalize_quaternion(_euler_to_quaternion(0.2, -0.1, 0.3))
-    assert_angles_close(_quaternion_to_euler(quat), [0.2, -0.1, 0.3])
-
-
 def test_zero_quaternion_is_invalid() -> None:
     with pytest.raises(ValueError):
         _normalize_quaternion([0.0, 0.0, 0.0, 0.0])
-
-
-def test_mock_adapter_uses_shortest_angular_velocity() -> None:
-    adapter = MockRealManAdapter([0.0, 0.0, 0.0, 3.13, 0.0, -3.13], 0.1)
-
-    adapter.send_cartesian_target(ArmPose(0.0, 0.0, 0.0, -3.13, 0.0, 3.13), False)
-
-    assert abs(adapter.last_velocity[3]) < 1.0
-    assert abs(adapter.last_velocity[5]) < 1.0

xr_rm_teleop/test/test_placo_transforms.py

@@ -0,0 +1,169 @@
+import math
+from pathlib import Path
+from types import SimpleNamespace
+from xml.etree import ElementTree
+
+import numpy as np
+import pytest
+
+from xr_rm_teleop.placo_ik_solver import (
+    QP_POSITION_TOLERANCE_M,
+    PlacoIkSolver,
+    _validated_transform,
+)
+
+
+def test_fixed_urdf_has_seven_moving_joints_and_omnipicker_tcp() -> None:
+    urdf_path = (
+        Path(__file__).resolve().parents[1]
+        / "models"
+        / "rm75_omnipicker"
+        / "urdf"
+        / "RM75-B_OmniPicker_fixed.urdf"
+    )
+    root = ElementTree.parse(urdf_path).getroot()
+    moving_joint_names = [
+        joint.attrib["name"]
+        for joint in root.findall("joint")
+        if joint.attrib["type"] != "fixed"
+    ]
+    tcp_joint = root.find("joint[@name='omnipicker_tcp_joint']")
+    mesh_filenames = [
+        mesh.attrib["filename"]
+        for mesh in root.findall(".//mesh")
+    ]
+
+    assert moving_joint_names == [f"joint_{index}" for index in range(1, 8)]
+    assert all(
+        filename.startswith(
+            "package://xr_rm_teleop/models/rm75_omnipicker/meshes/"
+        )
+        for filename in mesh_filenames
+    )
+    assert tcp_joint is not None
+    assert tcp_joint.attrib["type"] == "fixed"
+    assert tcp_joint.find("parent").attrib["link"] == "omnipicker_base_link"
+    assert tcp_joint.find("child").attrib["link"] == "omnipicker_tcp"
+    assert tcp_joint.find("origin").attrib["xyz"] == "0 0 0.16"
+    assert tcp_joint.find("origin").attrib["rpy"] == "0 0 0"
+
+
+def _rm75_placo_solver() -> tuple[PlacoIkSolver, list[float]]:
+    pytest.importorskip("placo")
+    urdf_path = (
+        Path(__file__).resolve().parents[1]
+        / "models"
+        / "rm75_omnipicker"
+        / "urdf"
+        / "RM75-B_OmniPicker_fixed.urdf"
+    )
+    joints = [
+        math.radians(value)
+        for value in [
+            -90.14,
+            3.76,
+            -86.89,
+            87.89,
+            -96.53,
+            -79.62,
+            -90.04,
+        ]
+    ]
+    return PlacoIkSolver(str(urdf_path), 1.0 / 90.0), joints
+
+
+def test_qp_solve_converges_to_reachable_tcp_target() -> None:
+    solver, joints = _rm75_placo_solver()
+    start_pose = solver.update_joint_state(joints)
+    target_pose = start_pose.copy()
+    target_pose[0, 3] += 0.07
+
+    result = solver.solve(target_pose)
+    reached_pose = solver.update_joint_state(result)
+    position_error = np.linalg.norm(
+        target_pose[:3, 3] - reached_pose[:3, 3]
+    )
+    rotation_delta = (
+        target_pose[:3, :3] @ reached_pose[:3, :3].T
+    )
+    orientation_error = math.acos(
+        float(
+            np.clip(
+                (np.trace(rotation_delta) - 1.0) * 0.5,
+                -1.0,
+                1.0,
+            )
+        )
+    )
+
+    assert position_error <= QP_POSITION_TOLERANCE_M
+    assert orientation_error <= 5e-3
+
+
+def test_qp_solve_accepts_position_error_within_two_millimeters() -> None:
+    solver = object.__new__(PlacoIkSolver)
+    solver._actual_joints = np.zeros(7)
+    solver._robot = SimpleNamespace(
+        state=SimpleNamespace(q=np.zeros(14))
+    )
+    solver._frame_task = SimpleNamespace(T_world_frame=None)
+    solver._target_errors = lambda: (1.5e-3, 0.0)
+
+    result = solver.solve(np.eye(4))
+
+    assert result == pytest.approx([0.0] * 7)
+
+
+def test_qp_solve_rejects_position_error_above_two_millimeters() -> None:
+    solver = object.__new__(PlacoIkSolver)
+    solver._actual_joints = np.zeros(7)
+    solver._robot = SimpleNamespace(
+        state=SimpleNamespace(q=np.zeros(14)),
+        update_kinematics=lambda: None,
+    )
+    solver._frame_task = SimpleNamespace(T_world_frame=None)
+    solver._solver = SimpleNamespace(solve=lambda update: None)
+    solver._validate_result = lambda result, previous: None
+    solver._target_errors = lambda: (2.1e-3, 0.0)
+
+    with pytest.raises(RuntimeError, match="QP did not converge after 30"):
+        solver.solve(np.eye(4))
+
+
+def test_validated_transform_accepts_finite_se3_and_returns_a_copy() -> None:
+    transform = np.eye(4)
+    transform[:3, 3] = [0.3, -0.1, 0.2]
+
+    actual = _validated_transform(transform)
+
+    assert actual == pytest.approx(transform)
+    assert actual is not transform
+
+
+@pytest.mark.parametrize(
+    "transform",
+    [
+        np.eye(3),
+        np.full((4, 4), np.nan),
+        np.vstack([np.eye(3, 4), [0.0, 0.0, 0.0, 2.0]]),
+        np.diag([2.0, 1.0, 1.0, 1.0]),
+    ],
+)
+def test_validated_transform_rejects_invalid_se3(transform: np.ndarray) -> None:
+    with pytest.raises(ValueError):
+        _validated_transform(transform)
+
+
+def test_qp_result_rejects_nan_position_and_velocity_violations() -> None:
+    solver = object.__new__(PlacoIkSolver)
+    solver._joint_limits = np.asarray([[-1.0, 1.0]] * 7)
+    solver._velocity_limits = np.ones(7)
+    solver._dt = 0.1
+    solver._actual_joints = np.zeros(7)
+
+    with pytest.raises(ValueError, match="finite"):
+        solver._validate_result(np.full(7, np.nan))
+    with pytest.raises(ValueError, match="position"):
+        solver._validate_result(np.full(7, 2.0))
+    with pytest.raises(ValueError, match="velocity"):
+        solver._validate_result(np.full(7, 0.2))

xr_rm_teleop/xr_rm_teleop/fun_peripheral.py

@@ -25,6 +25,14 @@ class PeripheralConfig:
     tools_in_ee: dict[str, list[list[float]]]
     set_initial_tool_state: bool = False
 
+    @property
+    def tool_name(self) -> str:
+        return list(self.tools_in_ee)[self.scissorgripper]
+
+    @property
+    def tool_pose(self) -> list[float]:
+        return list(self.tools_in_ee[self.tool_name][0])
+
 
 def load_peripheral_config(config_file: str, arm: str) -> PeripheralConfig:
     """从 bringup YAML 读取指定左右臂的外设配置。"""
@@ -94,6 +102,40 @@ def _tool_name_for_index(tools_in_ee: dict[str, list[list[float]]], scissorgripp
     return list(tools_in_ee.keys())[scissorgripper]
 
 
+def _check_sdk_return(result: Any, operation: str) -> None:
+    if result != 0:
+        raise RuntimeError(f"{operation} failed with code {result}: {result!r}")
+
+
+def _configure_tool_frame(robot, tool_frame, tool_name: str) -> None:
+    frames = robot.rm_get_total_tool_frame()
+    if not isinstance(frames, dict):
+        raise RuntimeError(
+            f"rm_get_total_tool_frame returned invalid data: {frames!r}"
+        )
+    _check_sdk_return(
+        frames.get("return_code"),
+        "rm_get_total_tool_frame",
+    )
+    tool_names = frames.get("tool_names")
+    if not isinstance(tool_names, (list, tuple)):
+        raise RuntimeError(
+            f"rm_get_total_tool_frame returned invalid tool_names: {tool_names!r}"
+        )
+
+    if tool_name in tool_names:
+        operation = "rm_update_tool_frame"
+        result = robot.rm_update_tool_frame(frame=tool_frame)
+    else:
+        operation = "rm_set_manual_tool_frame"
+        result = robot.rm_set_manual_tool_frame(frame=tool_frame)
+    _check_sdk_return(result, operation)
+    _check_sdk_return(
+        robot.rm_change_tool_frame(tool_name),
+        "rm_change_tool_frame",
+    )
+
+
 def cal_tool_frame(handle, scissorgripper, tools_in_ee):
     """根据末端工具配置生成 RealMan 工具坐标系。"""
     from Robotic_Arm.rm_robot_interface import rm_frame_t
@@ -152,8 +194,7 @@ def peripheral_cfg(
 
     time.sleep(0.2)
     tool_frame, tool_name = cal_tool_frame(robot, scissorgripper, tools_in_ee)
-    robot.rm_set_manual_tool_frame(frame=tool_frame)
-    robot.rm_change_tool_frame(tool_name)
+    _configure_tool_frame(robot, tool_frame, tool_name)
 
     if scissorgripper == 0:
         # 剪刀夹爪通过工具板数字输出控制。

xr_rm_teleop/xr_rm_teleop/placo_ik_solver.py

@@ -0,0 +1,194 @@
+"""RM75 的 Placo 0.9.4 有界迭代 QP 逆解。"""
+
+from __future__ import annotations
+
+from importlib.metadata import PackageNotFoundError, version
+from pathlib import Path
+
+import numpy as np
+
+EXPECTED_PLACO_VERSION = "0.9.4"
+RM75_JOINT_NAMES = [f"joint_{index}" for index in range(1, 8)]
+RM75_Q_SLICE = slice(7, 14)
+QP_MAX_ITERATIONS = 30
+QP_POSITION_TOLERANCE_M = 2e-3
+QP_ORIENTATION_TOLERANCE_RAD = 5e-3
+
+
+def _validated_transform(transform: np.ndarray) -> np.ndarray:
+    values = np.asarray(transform, dtype=float)
+    if values.shape != (4, 4) or not np.isfinite(values).all():
+        raise ValueError("target transform must be a finite 4x4 matrix")
+    if not np.allclose(values[3], [0.0, 0.0, 0.0, 1.0], atol=1e-9):
+        raise ValueError("target transform must have a valid homogeneous row")
+
+    rotation = values[:3, :3]
+    if (
+        np.linalg.norm(rotation.T @ rotation - np.eye(3)) > 1e-3
+        or np.linalg.det(rotation) <= 0.0
+    ):
+        raise ValueError("target transform must contain a valid rotation")
+    u, _, vt = np.linalg.svd(rotation)
+    projected = u @ vt
+    if np.linalg.det(projected) <= 0.0:
+        raise ValueError("target transform must contain a proper rotation")
+
+    result = values.copy()
+    result[:3, :3] = projected
+    return result
+
+
+class PlacoIkSolver:
+    def __init__(
+        self,
+        urdf_path: str,
+        dt: float,
+    ) -> None:
+        if dt <= 0.0:
+            raise ValueError("dt must be positive")
+        try:
+            installed_version = version("placo")
+            import placo
+        except (ImportError, PackageNotFoundError) as exc:
+            raise RuntimeError(
+                "Placo 0.9.4 must come from "
+                "/home/robot/miniconda3/envs/xr"
+            ) from exc
+        if installed_version != EXPECTED_PLACO_VERSION:
+            raise RuntimeError(
+                f"Placo {EXPECTED_PLACO_VERSION} is required, got {installed_version}"
+            )
+
+        model_path = Path(urdf_path).expanduser().resolve()
+        if not model_path.is_file():
+            raise FileNotFoundError(f"RM75 URDF not found: {model_path}")
+
+        self._dt = dt
+        self._robot = placo.RobotWrapper(str(model_path))
+        if self._robot.state.q.shape != (14,):
+            raise RuntimeError(
+                f"expected Placo q shape (14,), got {self._robot.state.q.shape}"
+            )
+        if list(self._robot.joint_names()) != RM75_JOINT_NAMES:
+            raise RuntimeError(
+                f"unexpected RM75 joint order: {list(self._robot.joint_names())}"
+            )
+        offsets = [
+            self._robot.get_joint_offset(name) for name in RM75_JOINT_NAMES
+        ]
+        if offsets != list(range(7, 14)):
+            raise RuntimeError(f"unexpected RM75 q offsets: {offsets}")
+
+        self._joint_limits = np.asarray(
+            [self._robot.get_joint_limits(name) for name in RM75_JOINT_NAMES]
+        )
+        velocity_offsets = [
+            self._robot.get_joint_v_offset(name) for name in RM75_JOINT_NAMES
+        ]
+        self._velocity_limits = np.asarray(
+            [
+                self._robot.model.velocityLimit[index]
+                for index in velocity_offsets
+            ]
+        )
+        self._actual_joints: np.ndarray | None = None
+
+        self._solver = placo.KinematicsSolver(self._robot)
+        self._solver.dt = dt
+        self._solver.mask_fbase(True)
+        self._solver.enable_velocity_limits(True)
+        self._frame_task = self._solver.add_frame_task(
+            "omnipicker_tcp",
+            np.eye(4),
+        )
+        self._frame_task.configure("rm75_frame", "soft", 1.0)
+        self._solver.add_kinetic_energy_regularization_task(1e-6)
+
+    @property
+    def base_configuration(self) -> list[float]:
+        return self._robot.state.q[:7].tolist()
+
+    def update_joint_state(self, joints: list[float]) -> np.ndarray:
+        values = np.asarray(joints, dtype=float)
+        if values.shape != (7,) or not np.isfinite(values).all():
+            raise ValueError("joint state must contain 7 finite values")
+        is_first_feedback = self._actual_joints is None
+        self._actual_joints = values.copy()
+        self._robot.state.q[RM75_Q_SLICE] = values
+        self._robot.update_kinematics()
+        base_to_tool = self._robot.get_T_world_frame("omnipicker_tcp")
+        if is_first_feedback:
+            self._frame_task.T_world_frame = base_to_tool.copy()
+        return base_to_tool.copy()
+
+    def _target_errors(self) -> tuple[float, float]:
+        position_task = self._frame_task.position()
+        orientation_task = self._frame_task.orientation()
+        position_task.update()
+        orientation_task.update()
+        return (
+            float(position_task.error_norm()),
+            float(orientation_task.error_norm()),
+        )
+
+    def solve(self, target_tool_pose: np.ndarray) -> list[float]:
+        if self._actual_joints is None:
+            raise RuntimeError("joint state must be initialized before QP solve")
+        self._frame_task.T_world_frame = _validated_transform(
+            target_tool_pose
+        )
+        result = np.asarray(
+            self._robot.state.q[RM75_Q_SLICE],
+            dtype=float,
+        ).copy()
+        position_error, orientation_error = self._target_errors()
+        if (
+            position_error <= QP_POSITION_TOLERANCE_M
+            and orientation_error <= QP_ORIENTATION_TOLERANCE_RAD
+        ):
+            return result.tolist()
+
+        for _ in range(QP_MAX_ITERATIONS):
+            previous = result
+            self._solver.solve(True)
+            self._robot.update_kinematics()
+            result = np.asarray(
+                self._robot.state.q[RM75_Q_SLICE],
+                dtype=float,
+            ).copy()
+            self._validate_result(result, previous)
+            position_error, orientation_error = self._target_errors()
+            if (
+                position_error <= QP_POSITION_TOLERANCE_M
+                and orientation_error <= QP_ORIENTATION_TOLERANCE_RAD
+            ):
+                return result.tolist()
+
+        raise RuntimeError(
+            "QP did not converge after "
+            f"{QP_MAX_ITERATIONS} iterations: "
+            f"position_error={position_error:.6f} m, "
+            f"orientation_error={orientation_error:.6f} rad"
+        )
+
+    def _validate_result(
+        self,
+        result: np.ndarray,
+        reference: np.ndarray | None = None,
+    ) -> None:
+        if result.shape != (7,) or not np.isfinite(result).all():
+            raise ValueError("QP result must contain 7 finite values")
+        lower = self._joint_limits[:, 0]
+        upper = self._joint_limits[:, 1]
+        if np.any(result < lower - 1e-9) or np.any(result > upper + 1e-9):
+            raise ValueError("QP result violates RM75 joint position limits")
+        if reference is None:
+            reference = self._actual_joints
+        if reference is None:
+            raise RuntimeError("joint state has not been initialized")
+        reference = np.asarray(reference, dtype=float)
+        if reference.shape != (7,) or not np.isfinite(reference).all():
+            raise ValueError("QP reference must contain 7 finite values")
+        max_step = self._velocity_limits * self._dt + 1e-9
+        if np.any(np.abs(result - reference) > max_step):
+            raise ValueError("QP result violates RM75 one-cycle velocity limits")

xr_rm_teleop/xr_rm_teleop/realman_adapter.py

@@ -1,21 +1,16 @@
-"""RM75 机械臂适配层。
-
-对上提供统一的当前位姿读取、笛卡尔位姿目标发送和停止接口；对下根据配置
-选择 mock 积分模拟器或睿尔曼 Python API2 真机通信。
-"""
+"""RM75 机械臂关节反馈、关节透传和停止适配层。"""
 
 from __future__ import annotations
 
+import ipaddress
 import math
+import threading
+import time
 from dataclasses import dataclass
 from numbers import Number
 from typing import Any
 
 
-def _angle_delta(target: float, current: float) -> float:
-    return math.atan2(math.sin(target - current), math.cos(target - current))
-
-
 @dataclass
 class ArmPose:
     x: float
@@ -32,55 +27,72 @@ class ArmPose:
         return [self.rx, self.ry, self.rz]
 
 
-class MockRealManAdapter:
-    """无机械臂时使用的运动学模拟器，用于验证 ROS2 遥操链路。"""
+@dataclass(frozen=True)
+class JointStateSnapshot:
+    positions: list[float]
+    received_at: float
+    read_duration_ms: float | None = None
+    update_interval_ms: float | None = None
+    motion_ready: bool = True
 
-    def __init__(self, initial_pose: list[float], dt: float) -> None:
-        self._pose = ArmPose(*initial_pose[:6])
-        self._dt = dt
-        self.last_velocity = [0.0] * 6
+
+class MockRealManAdapter:
+    """不导入厂商 SDK 的关节状态 mock。"""
+
+    def __init__(self, initial_joint_degrees: list[float]) -> None:
+        if len(initial_joint_degrees) != 7 or not all(
+            math.isfinite(value) for value in initial_joint_degrees
+        ):
+            raise ValueError("initial joint pose must contain 7 finite values")
+        self._joint_positions = [
+            math.radians(value) for value in initial_joint_degrees
+        ]
+        self.last_joint_target: list[float] | None = None
         self.last_tool_open: bool | None = None
 
     def connect(self) -> None:
         return
 
-    def get_current_pose(self) -> ArmPose:
-        return self._pose
+    def get_latest_joint_state(self) -> JointStateSnapshot:
+        return JointStateSnapshot(
+            list(self._joint_positions),
+            time.monotonic(),
+        )
 
-    def send_cartesian_target(self, pose: ArmPose, follow: bool) -> None:
+    def read_joint_state(self) -> JointStateSnapshot:
+        return self.get_latest_joint_state()
+
+    def send_joint_target(self, joints: list[float], follow: bool) -> None:
         del follow
-        self.last_velocity = [
-            (pose.x - self._pose.x) / self._dt,
-            (pose.y - self._pose.y) / self._dt,
-            (pose.z - self._pose.z) / self._dt,
-            _angle_delta(pose.rx, self._pose.rx) / self._dt,
-            _angle_delta(pose.ry, self._pose.ry) / self._dt,
-            _angle_delta(pose.rz, self._pose.rz) / self._dt,
-        ]
-        self._pose = pose
+        if len(joints) != 7 or not all(math.isfinite(value) for value in joints):
+            raise ValueError("joint target must contain 7 finite values")
+        self._joint_positions = list(joints)
+        self.last_joint_target = list(joints)
 
     def stop(self) -> None:
-        self.last_velocity = [0.0] * 6
+        return
 
     def close(self) -> None:
         self.stop()
 
-    def configure_peripheral(self, config_file: str, peripheral_arm: str) -> None:
-        del config_file, peripheral_arm
+    def configure_peripheral(self, config: Any, peripheral_arm: str) -> None:
+        del config, peripheral_arm
 
     def set_tool_enabled(self, open_tool: bool) -> None:
         self.last_tool_open = open_tool
 
 
 class RealManAdapter:
-    """睿尔曼 Python API2 的笛卡尔位姿透传适配层。"""
+    """复用一个睿尔曼 Python API2 连接的关节适配层。"""
 
     def __init__(
         self,
         robot_ip: str,
         robot_port: int,
         avoid_singularity: int,
-        frame_type: int,
+        realtime_push_host_ip: str,
+        realtime_push_port: int,
+        realtime_push_cycle_ms: int = 5,
         logger: Any | None = None,
         configure_safety_limits: bool = True,
         max_line_speed: float = 1.0,
@@ -98,7 +110,22 @@ class RealManAdapter:
         self._robot_ip = robot_ip
         self._robot_port = robot_port
         self._avoid_singularity = avoid_singularity
-        self._frame_type = frame_type
+        try:
+            self._realtime_push_host_ip = str(
+                ipaddress.IPv4Address(realtime_push_host_ip)
+            )
+        except ipaddress.AddressValueError as exc:
+            raise ValueError(
+                "realtime_push_host_ip must be a valid IPv4 address"
+            ) from exc
+        if not 1 <= realtime_push_port <= 65535:
+            raise ValueError("realtime_push_port must be between 1 and 65535")
+        if realtime_push_cycle_ms <= 0 or realtime_push_cycle_ms % 5 != 0:
+            raise ValueError(
+                "realtime_push_cycle_ms must be a positive multiple of 5"
+            )
+        self._realtime_push_port = realtime_push_port
+        self._realtime_push_cycle_ms = realtime_push_cycle_ms
         self._logger = logger
         self._configure_safety_limits = configure_safety_limits
         self._max_line_speed = max_line_speed
@@ -114,54 +141,134 @@ class RealManAdapter:
         self._canfd_radio = canfd_radio
         self._scissorgripper: int | None = None
         self._arm: Any | None = None
+        self._joint_state_lock = threading.Lock()
+        self._latest_joint_state: JointStateSnapshot | None = None
+        self._feedback_ready = threading.Event()
+        self._realtime_callback: Any | None = None
+        self._accept_realtime_feedback = False
+        self._feedback_fault_logged = False
+        self._last_motion_status: tuple[Any, ...] | None = None
 
     def connect(self) -> None:
         try:
-            from Robotic_Arm.rm_robot_interface import RoboticArm, rm_thread_mode_e
+            from Robotic_Arm.rm_robot_interface import (
+                RoboticArm,
+                rm_realtime_arm_state_callback_ptr,
+                rm_realtime_push_config_t,
+                rm_thread_mode_e,
+            )
         except ImportError as exc:
             raise RuntimeError(
                 "未安装睿尔曼 Python API2。请安装厂商 SDK，或用 use_mock:=true 先跑模拟模式。"
             ) from exc
 
         self._arm = RoboticArm(rm_thread_mode_e.RM_TRIPLE_MODE_E)
-        handle = self._arm.rm_create_robot_arm(self._robot_ip, self._robot_port)
-        self._check_robot_handle(handle)
-        self._log_info(
-            "RealMan connected: "
-            f"ip={self._robot_ip}, port={self._robot_port}, "
-            f"avoid_singularity={self._avoid_singularity}, "
-            f"frame_type={self._frame_type}, command=rm_movep_canfd"
+        try:
+            handle = self._arm.rm_create_robot_arm(
+                self._robot_ip,
+                self._robot_port,
+            )
+            self._check_robot_handle(handle)
+            self._log_info(
+                "RealMan connected: "
+                f"ip={self._robot_ip}, port={self._robot_port}, "
+                f"avoid_singularity={self._avoid_singularity}, "
+                "command=rm_movej_canfd"
+            )
+            if self._configure_safety_limits:
+                self._apply_safety_limits()
+            if self._move_to_initial_pose_on_connect:
+                self._move_to_initial_pose()
+            self._feedback_ready.clear()
+            self._accept_realtime_feedback = True
+            self._realtime_callback = rm_realtime_arm_state_callback_ptr(
+                self._on_realtime_arm_state
+            )
+            self._arm.rm_realtime_arm_state_call_back(
+                self._realtime_callback
+            )
+            config = rm_realtime_push_config_t(
+                self._realtime_push_cycle_ms // 5,
+                True,
+                self._realtime_push_port,
+                0,
+                self._realtime_push_host_ip,
+            )
+            self._check_return(
+                self._arm.rm_set_realtime_push(config),
+                "rm_set_realtime_push",
+            )
+            if not self._feedback_ready.wait(timeout=2.0):
+                raise RuntimeError(
+                    "RealMan UDP realtime feedback did not receive a valid "
+                    "frame within 2 seconds"
+                )
+            self._log_info(
+                "RealMan UDP realtime feedback ready: "
+                f"host={self._realtime_push_host_ip}:"
+                f"{self._realtime_push_port}, "
+                f"cycle={self._realtime_push_cycle_ms} ms"
+            )
+        except Exception:
+            self._accept_realtime_feedback = False
+            try:
+                self._arm.rm_delete_robot_arm()
+            except Exception:
+                pass
+            self._arm = None
+            self._realtime_callback = None
+            raise
+
+    def get_latest_joint_state(self) -> JointStateSnapshot | None:
+        with self._joint_state_lock:
+            if self._latest_joint_state is None:
+                return None
+            return JointStateSnapshot(
+                list(self._latest_joint_state.positions),
+                self._latest_joint_state.received_at,
+                self._latest_joint_state.read_duration_ms,
+                self._latest_joint_state.update_interval_ms,
+                self._latest_joint_state.motion_ready,
+            )
+
+    def read_joint_state(self) -> JointStateSnapshot:
+        self._require_arm()
+        started_at = time.monotonic()
+        result = self._arm.rm_get_joint_degree()
+        finished_at = time.monotonic()
+        if not isinstance(result, tuple) or len(result) != 2:
+            raise RuntimeError(
+                f"rm_get_joint_degree returned invalid result: {result!r}"
+            )
+        self._check_return(result, "rm_get_joint_degree")
+        return JointStateSnapshot(
+            self._joint_positions_from_degrees(
+                result[1],
+                "rm_get_joint_degree",
+            ),
+            finished_at,
+            (finished_at - started_at) * 1000.0,
         )
-        if self._configure_safety_limits:
-            self._apply_safety_limits()
-        if self._move_to_initial_pose_on_connect:
-            self._move_to_initial_pose()
 
-    def get_current_pose(self) -> ArmPose:
+    def send_joint_target(self, joints: list[float], follow: bool) -> None:
         self._require_arm()
-        state = self._arm.rm_get_current_arm_state()
-        pose = self._find_pose(state)
-        if pose is None:
-            raise RuntimeError(f"无法从睿尔曼状态中解析当前 TCP 位姿：{state!r}")
-        return ArmPose(*pose[:6])
-
-    def send_cartesian_target(self, pose: ArmPose, follow: bool) -> None:
-        self._require_arm()
-        ret = self._arm.rm_movep_canfd(
-            [pose.x, pose.y, pose.z, pose.rx, pose.ry, pose.rz],
+        if len(joints) != 7 or not all(math.isfinite(value) for value in joints):
+            raise ValueError("joint target must contain 7 finite values")
+        ret = self._arm.rm_movej_canfd(
+            [math.degrees(value) for value in joints],
             follow,
+            0,
             self._canfd_trajectory_mode,
             self._canfd_radio,
         )
-        self._check_return(ret, "rm_movep_canfd")
+        self._check_return(ret, "rm_movej_canfd")
 
-    def configure_peripheral(self, config_file: str, peripheral_arm: str) -> None:
+    def configure_peripheral(self, config: Any, peripheral_arm: str) -> None:
         self._require_arm()
-        from .fun_peripheral import load_peripheral_config, peripheral_cfg
+        from .fun_peripheral import peripheral_cfg
 
-        config = load_peripheral_config(config_file, peripheral_arm)
         self._scissorgripper = config.scissorgripper
-        tool_name = list(config.tools_in_ee.keys())[config.scissorgripper]
+        tool_name = config.tool_name
         self._log_info(
             "开始配置 RealMan 末端外设："
             f"arm={peripheral_arm}, scissorgripper={config.scissorgripper}, "
@@ -195,16 +302,144 @@ class RealManAdapter:
     def close(self) -> None:
         if self._arm is None:
             return
+        self._accept_realtime_feedback = False
         self.stop()
         try:
             self._arm.rm_delete_robot_arm()
         finally:
             self._arm = None
+            self._realtime_callback = None
 
     def _require_arm(self) -> None:
         if self._arm is None:
             raise RuntimeError("睿尔曼机械臂尚未连接")
 
+    def _on_realtime_arm_state(self, data: Any) -> None:
+        if not self._accept_realtime_feedback:
+            return
+        try:
+            if data is None or int(data.errCode) != 0:
+                raise ValueError("invalid realtime feedback error code")
+            arm_ip = data.arm_ip
+            if isinstance(arm_ip, bytes):
+                arm_ip = arm_ip.decode("utf-8").split("\x00", 1)[0]
+            if str(arm_ip) != self._robot_ip:
+                raise ValueError(
+                    f"unexpected realtime feedback source: {arm_ip}"
+                )
+            positions = self._joint_positions_from_degrees(
+                data.joint_status.joint_position,
+                "RM75 UDP feedback",
+            )
+            joint_enabled = [
+                bool(value) for value in data.joint_status.joint_en_flag
+            ]
+            joint_errors = [
+                int(value) for value in data.joint_status.joint_err_code
+            ]
+            if len(joint_enabled) != 7 or len(joint_errors) != 7:
+                raise ValueError(
+                    "RM75 UDP feedback must contain 7 joint states"
+                )
+            arm_error_count = int(data.err.err_len)
+            arm_errors = [
+                int(value) for value in list(data.err.err)[:arm_error_count]
+            ]
+            arm_errors = [code for code in arm_errors if code != 0]
+            arm_current_status = int(data.arm_current_status)
+            motion_ready = (
+                0 <= arm_current_status <= 8
+                and all(joint_enabled)
+                and not any(joint_errors)
+                and not arm_errors
+            )
+            motion_status = (
+                arm_current_status,
+                tuple(joint_enabled),
+                tuple(joint_errors),
+                tuple(arm_errors),
+                motion_ready,
+            )
+            received_at = time.monotonic()
+            with self._joint_state_lock:
+                update_interval_ms = (
+                    None
+                    if self._latest_joint_state is None
+                    else (
+                        received_at
+                        - self._latest_joint_state.received_at
+                    )
+                    * 1000.0
+                )
+                self._latest_joint_state = JointStateSnapshot(
+                    positions,
+                    received_at,
+                    None,
+                    update_interval_ms,
+                    motion_ready,
+                )
+            self._log_motion_status_transition(motion_status)
+            self._feedback_fault_logged = False
+            self._feedback_ready.set()
+        except Exception as exc:
+            with self._joint_state_lock:
+                if self._latest_joint_state is not None:
+                    current = self._latest_joint_state
+                    self._latest_joint_state = JointStateSnapshot(
+                        list(current.positions),
+                        current.received_at,
+                        current.read_duration_ms,
+                        current.update_interval_ms,
+                        False,
+                    )
+            if not self._feedback_fault_logged:
+                self._log_warn(
+                    f"RealMan UDP realtime feedback invalid: {exc}"
+                )
+                self._feedback_fault_logged = True
+
+    @staticmethod
+    def _joint_positions_from_degrees(
+        values: Any,
+        source: str,
+    ) -> list[float]:
+        try:
+            degrees = list(values)
+        except TypeError as exc:
+            raise ValueError(
+                f"{source} must contain 7 numeric joints"
+            ) from exc
+        if len(degrees) != 7 or not all(
+            isinstance(value, Number) for value in degrees
+        ):
+            raise ValueError(
+                f"{source} must contain 7 numeric joints"
+            )
+        positions = [math.radians(float(value)) for value in degrees]
+        if not all(math.isfinite(value) for value in positions):
+            raise ValueError(f"{source} contains NaN/Inf")
+        return positions
+
+    def _log_motion_status_transition(
+        self,
+        status: tuple[Any, ...],
+    ) -> None:
+        previous = self._last_motion_status
+        if status == previous:
+            return
+        self._last_motion_status = status
+        arm_status, joint_enabled, joint_errors, arm_errors, ready = status
+        details = (
+            f"arm_status={arm_status}, "
+            f"joint_enabled={list(joint_enabled)}, "
+            f"joint_errors={list(joint_errors)}, "
+            f"arm_errors={list(arm_errors)}"
+        )
+        if not ready:
+            self._log_warn(f"RealMan UDP 报警或掉使能：{details}")
+        elif previous is not None and not previous[-1]:
+            self._log_info(f"RealMan UDP 运动状态恢复正常：{details}")
+
     def _apply_safety_limits(self) -> None:
         # 真机安全限幅尽量下发到控制器；不支持的 SDK 接口会在 _try_call 中降级为警告。
         self._try_call("rm_set_avoid_singularity_mode", int(self._avoid_singularity))
@@ -260,74 +495,3 @@ class RealManAdapter:
     @staticmethod
     def _return_code(ret: Any) -> Any:
         return ret[0] if isinstance(ret, tuple) and ret else ret
-
-    @classmethod
-    def _find_pose(cls, obj: Any) -> list[float] | None:
-        # 不同 SDK 版本返回字段可能略有差异，因此递归查找常见 TCP 位姿字段。
-        if isinstance(obj, dict):
-            for key in ("pose", "tool_pose", "tcp_pose", "current_pose"):
-                pose = cls._as_pose(obj.get(key))
-                if pose is not None:
-                    return pose
-            for value in obj.values():
-                pose = cls._find_pose(value)
-                if pose is not None:
-                    return pose
-        elif isinstance(obj, (list, tuple)):
-            pose = cls._as_pose(obj)
-            if pose is not None:
-                return pose
-            for value in obj:
-                pose = cls._find_pose(value)
-                if pose is not None:
-                    return pose
-        elif hasattr(obj, "to_dictionary"):
-            try:
-                return cls._find_pose(obj.to_dictionary(7))
-            except TypeError:
-                return cls._find_pose(obj.to_dictionary())
-        elif hasattr(obj, "to_dict"):
-            return cls._find_pose(obj.to_dict())
-        else:
-            for key in ("pose", "tool_pose", "tcp_pose", "current_pose"):
-                if hasattr(obj, key):
-                    pose = cls._as_pose(getattr(obj, key))
-                    if pose is not None:
-                        return pose
-        return None
-
-    @staticmethod
-    def _as_pose(value: Any) -> list[float] | None:
-        if isinstance(value, (list, tuple)) and len(value) >= 6:
-            if all(isinstance(item, Number) for item in value[:6]):
-                return [float(item) for item in value[:6]]
-        if isinstance(value, dict):
-            position = value.get("position")
-            euler = value.get("euler")
-            if isinstance(position, dict) and isinstance(euler, dict):
-                keys = ("x", "y", "z")
-                rpy_keys = ("rx", "ry", "rz")
-                if all(key in position for key in keys) and all(key in euler for key in rpy_keys):
-                    return [
-                        float(position["x"]),
-                        float(position["y"]),
-                        float(position["z"]),
-                        float(euler["rx"]),
-                        float(euler["ry"]),
-                        float(euler["rz"]),
-                    ]
-        if all(hasattr(value, attr) for attr in ("position", "euler")):
-            position = getattr(value, "position")
-            euler = getattr(value, "euler")
-            if all(hasattr(position, key) for key in ("x", "y", "z")) and all(
-                hasattr(euler, key) for key in ("rx", "ry", "rz")
-            ):
-                return [
-                    float(position.x),
-                    float(position.y),
-                    float(position.z),
-                    float(euler.rx),
-                    float(euler.ry),
-                    float(euler.rz),
-                ]
-        return None