diff --git a/README.md b/README.md
index 6201136..f4240f6 100644
--- a/README.md
+++ b/README.md
@@ -12,3 +12,21 @@ Key specifications:
 Next:\
 Comparison with Realman official IK method.
 Embedded with current demo.
+
+
+Comparison (04June2026):
+
+- With current dual arm joint limit, 
+```
+ub = np.array([150.0, 110.0, 170.0,  130, 175.0, 125.0,  179.0])\
+lb = np.array([-150.0, -30.0, -170.0, -130, -175.0, -125.0,  -179.0])
+```
+the success rates for qp-based ik and realman Algo ik are 45% and 23%.
+
+- With realman-75 physical joint limit,
+```
+ub = np.array([179.0, 129.0, 179.0, 134, 179.0, 127.0, 359.0])
+lb = -ub
+```
+the success rates for qp-based ik and realman Algo ik are 76% and 51%.
+
diff --git a/kine_ctrl/main.py b/kine_ctrl/main.py
index 38f37f4..cb2cd8e 100644
--- a/kine_ctrl/main.py
+++ b/kine_ctrl/main.py
@@ -87,28 +87,31 @@ def demo_position_control():
 
 #   --------------  for comparison   ----------------
     print(f'in the comparison part')
-    #
-    # lower_limits = np.array([-3.14159, -2.2689, -3.14159, -2.3562, -3.14159, -2.234, -3.14159])
-    # upper_limits = np.array([3.14159, 2.2689, 3.14159, 2.3562, 3.14159, 2.234, 3.14159])
-    # result = np.array([[0, 0], [0, 0]])
-    #
-    # joint_rand = np.random.uniform(lower_limits, upper_limits)
-    #
-    # fk_qp = robot_kine_qp.forward_kinematics(joint_rand.tolist(), tool=tool_name)
-    # fk_qp_p = np.concatenate([fk_qp['position'], fk_qp['rpy']], axis=0)
-    # fk_rm_p = robot_kine_rm.forward_kinematics(q=(joint_rand*180/pi).tolist(), tool=tool_name)
-    # print(f'the fk diversion is { sum(abs( np.array(fk_rm_p) - np.array(fk_qp_p) )) }')
-
-
-
 
     if True:
-        lower_limits = np.array([  -3.14159, -2.2689, -3.14159, -2.3562, -3.14159, -2.234, -6.14159  ])
-        upper_limits = np.array([  3.14159, 2.2689, 3.14159, 2.3562, 3.14159, 2.234, 6.14159  ])
+
+        # ub = np.array([150.0, 110.0, 170.0,  130, 175.0, 125.0,  179.0])
+        # lb = np.array([-150.0, -30.0, -170.0, -130, -175.0, -125.0,  -179.0])
+        ub = np.array([179.0, 129.0, 179.0, 134, 179.0, 127.0, 359.0])
+        lb = -ub
+
+        # lower_limits = np.array([  -3.14159, -2.2689, -3.14159, -2.3562, -3.14159, -2.234, -3.14159  ])
+        # upper_limits = np.array([  3.14159, 2.2689, 3.14159, 2.3562, 3.14159, 2.234, 3.14159  ])
+
+        for i in range(7):
+            robot_kine_qp.model.lowerPositionLimit[i] = lb[i]/180*pi
+            robot_kine_qp.model.upperPositionLimit[i] = ub[i]/180*pi
+
+        robot_kine_rm.robot_kine_rm.rm_algo_set_joint_max_limit(ub)
+        robot_kine_rm.robot_kine_rm.rm_algo_set_joint_min_limit(lb)
+
         result = np.array([[0,0],[0,0]], dtype=np.int32)        #   qp_fk, qp_ik, rm_fk, rm_ik
+
+
+
         for i in range(1000):
             print(f'\n-------------- in i = {i} ----------------')
-            joint_rand = np.random.uniform(lower_limits, upper_limits)
+            joint_rand = np.random.uniform(ub/180*pi, lb/180*pi)
             print(f'the predefined joints are {joint_rand}')
 
             #   -------------- fk ------------------
@@ -129,7 +132,7 @@ def demo_position_control():
 
             #   ----------- ik ----------------
             t_p = fk_rm_p1
-            joint_rand_init = np.random.uniform(lower_limits, upper_limits)
+            joint_rand_init = np.random.uniform(ub/180*pi, lb/180*pi)
             print(f'the guess is {joint_rand_init}')
             joint_solution, success, error, ite = robot_kine_qp.inverse_kinematics(
                 t_p[0:3], target_rpy=t_p[3:6], initial_guess=joint_rand_init,
diff --git a/kine_ctrl/rm75_kine_qp.py b/kine_ctrl/rm75_kine_qp.py
index e3bc8a0..6057b17 100644
--- a/kine_ctrl/rm75_kine_qp.py
+++ b/kine_ctrl/rm75_kine_qp.py
@@ -92,10 +92,10 @@ class KinematicsSolver():
 
         # Joint limits (radians) - expanded for better reachability
         self.lower_limits = np.array([
-            -3.14159, -2.2689, -3.14159, -2.3562, -3.14159, -2.234, -3.14159
+            -3.14159, -2.2689, -3.14159, -2.3562, -3.14159, -2.234, -6.14159
         ])
         self.upper_limits = np.array([
-            3.14159, 2.2689, 3.14159, 2.3562, 3.14159, 2.234, 3.14159
+            3.14159, 2.2689, 3.14159, 2.3562, 3.14159, 2.234, 6.14159
         ])
 
         # Set joint limits in the model