README.md

metadata

configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*.jsonl

TOCA Demo Dataset (LeRobot Compatible)

This dataset was converted from the BOTA driver SQLite recording.

Structure

• observation.state: Joint positions (6 DOF)
• observation.velocity: Joint velocities (6 DOF)
• observation.wrench: Force/Torque sensor data (Fx, Fy, Fz, Tx, Ty, Tz)
• action: Joint positions (Target for Imitation Learning)
• episode_index: ID of the episode
• frame_index: Frame number within the episode
• timestamp: Time in seconds
• next.done: Boolean indicating end of episode

Loading with LeRobot / Hugging Face Datasets

from datasets import load_dataset

dataset = load_dataset("./lerobot_dataset")
print(dataset['train'][0])