Open cong1024 opened 3 months ago
@cong1024 I will be quick, but happy to help further.
First, checkout origin/main to get the latest version (we merged the PR). Then, redo the calibration precedure by deleting this file: https://github.com/huggingface/lerobot/blob/main/lerobot/common/robot_devices/robots/koch.py#L305 Click on the image urls during the procedure and carefully match the positions with your arms.
Secondly check that the index of the motor corresponds to what you have with DynamixelWizzard
Without Dynamixel Wizzard, you can try to order each one to move seperately by reading their position and writing a delta position:
goal_position_array, self.leader_arms[name].read("Present_Position")
# update each dim seperately
self.leader_arms[name].write("Goal_Position", goal_position_array)
You can also write one value each time and the name of the motor similarly to:
self.leader_arms[name].write("Goal_Position", GRIPPER_OPEN, "gripper")
You can also access their ID and make sure it goes from 1 to 6: https://github.com/huggingface/lerobot/blob/main/lerobot/common/robot_devices/motors/dynamixel.py#L38-L39
We are working on a precedure to ensure the motor indices are well set.
Thanks for help ! @Cadene , yes ,I have seen the main branch has merged the control of koch robot. For controlling aloha robot, I see the pr is still in draft, maybe working on branch user/rcadene/2024_07_11_control_aloha is better to debug? Since the configuration for aloha and koch is different. For the rest suggestion, Yes, I will try to working on ! thx~
this branch is not stable, I dont recommend using it
Once this PR merges, Aloha will be natively supported: https://github.com/huggingface/lerobot/pull/316 Sorry for the wait!
System Info
Information
Reproduction
Hi Guys, I am using the pr #316 written by Cadene to control the real aloha robot, when running cmd : python control_robot.py teleoperate --robot aloha, I found the follower move fast to HORIZONTAL_POSITION, all arms in HORIZONTAL_POSITION like a line. When I control the follower arm with the master arm, I find that the movement direction of the follower arm is exactly opposite to that of the master arm. I thinke there may be some bug in code, or my own problem. I tried to figure out the following:
Expected behavior
^^