Closed Xu-robot-learning closed 2 months ago
You can add a dimension to start_angles
in default_xyz.yaml
, e.g. start_angles: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
for ur5right
and start_angles: [0.126, -0.942, -1.88, -4.15, -4.78, 0.0, 0.0]
for ur5left
.
Adding a dimension can avoid errors, but UR5 has a pose problem and should put the added dimension in the first place. eg start_angles: [0.0, 0.126, -0.942, -1.88, -4.15, -4.78, 0.0]
Hello, sorry for the late reply here. From what I recall, there may be a yaml mismatch with the device.py file. If you are still interested, you can find a newer/working version of irl_control here: https://github.com/ir-lab/bimanual-imitation, which shouldn't have this problem.
When I execute force_test.py file, the problem appears on line 79: if self.name == "ur5right" or self.name == "ur5left": self.sim.qpos[self.joint_ids] = np.copy(self.start_angles) rerurn the error: shape mismatch: value array of shape (6,) could not be broadcast to indexing result of shape (7,). Can you tell me why and how to modify code?Thanks!