SO-ARM101 Cube2Plate Policy

This model is a LeRobot policy trained for the SO-ARM101 robotic arm on the Cube2Plate task.

The policy is used to control the SO-ARM101 robotic arm to pick up a cube and place it onto a plate.

Model Details

  • Robot: SO-ARM101
  • Task: Cube2Plate
  • Framework: LeRobot
  • Model type: Imitation learning policy
  • License: Apache-2.0
  • Training dataset: Hiwonder-robot/cube_to_plate

Intended Use

This model is intended for running and evaluating the Cube2Plate task with the SO-ARM101 robotic arm.

It can be used for:

  • SO-ARM101 robotic arm control
  • Cube-to-plate manipulation task
  • LeRobot policy evaluation
  • Imitation learning experiments

How to Use

You can run evaluation or inference with LeRobot:

lerobot-record \
  --robot.type=so101_follower \
  --dataset.repo_id=${HF_USER}/eval_so101 \
  --policy.path=${HF_USER}/Cube2Plate \
  --episodes=10

If the following folder already exists locally:

C:/Users/Admin/Desktop/lerobot/${HF_USER}/eval_so101

please delete it before running the command again.

Training Data

This model was trained using demonstration data collected for the Cube2Plate task with the SO-ARM101 robotic arm.

Dataset:

Hiwonder-robot/cube_to_plate

Evaluation

The model can be evaluated by running the trained policy on the SO-ARM101 robotic arm and checking whether the cube is successfully placed onto the plate.

Limitations

  • This model is designed only for the Cube2Plate task.
  • Performance may be affected by camera position, lighting conditions, object position, and robot calibration.
  • The model may not generalize well to different objects, environments, or robot configurations.
  • Please run the robot carefully to avoid collision or hardware damage.

More Information

This model was trained and uploaded using the LeRobot framework.

Downloads last month
32
Safetensors
Model size
51.7M params
Tensor type
F32
·
Video Preview
loading

Dataset used to train Hiwonder-robot/Cube2Plate