jhd0313/diffusion_move-box
Robot Arm Control Policy
1 min readupdated
Trained using the LeRobot library to solve box manipulation tasks using a so_follower robot manipulator. It uses Diffusion Policy (an approach that frames motion control as a diffusion process to generate smooth and consistent trajectories). The model takes the robot's state and images from two cameras (overhead and wrist) as input, outputting control actions. Useful for researchers and engineers working with LeRobot-based robotics.
Key features
- Built on Diffusion Policy for precise contact-rich control
- Uses two cameras: overhead and wrist
- Trained on a dataset of 50 episodes (10,668 frames at 30 FPS)
- Developed and exported using the LeRobot ecosystem
What you can do
- Run the ready-to-use control policy on your so_follower robot
- Analyze the motion trajectory dataset in LeRobot format
- Fine-tune the model on your own demonstrations for new manipulation scenarios
Example
Running the pre-trained control policy on the robot via the command-line utility:
lerobot-rollout \
--strategy.type=base \
--robot.type=so_follower \
--robot.port= \
--robot.cameras="{ overhead: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30}, wrist: {type: opencv, index_or_path: 1, width: 640, height: 480, fps: 30}}" \
--policy.path=jhd0313/diffusion_move-box \
--task="Move the box" \
--duration=60Read more
DetailsLicense: apache-2.0
Robot type: so_follower
Training parameters: 5000 steps, batch size 8, adam optimizer, learning rate 0.0001
Framework: LeRobot v0.6.0










Comments