MuJoCoPickAndPlace V1: Robot Manipulation Episodes from Simulation

A dataset of 10 episodes containing 4,357 frames of robot pick-and-place tasks generated using the LeRobot framework. The data was created by johnsutor and last updated on July 17, 2026. It includes both data files and video files, with splits defined for training.

Use Cases

• Training reinforcement learning agents for robotic pick-and-place tasks based on the described episodes.
• Benchmarking robot control policies using the provided simulation data and video recordings.
• Developing imitation learning models from the demonstration data generated by LeRobot.
• Analyzing robot manipulation trajectories and state transitions within the 4,357 frames.

Strengths

• Contains 10 complete task episodes, providing multiple demonstrations.
• Includes 4,357 frames of data, offering a sequence of states and actions.
• Provides associated video files (200 MB) recorded at 30 FPS for visual analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

johnsutor via Hugging Face.

Collection Method

Created using the LeRobot framework, likely in a MuJoCo simulation environment.

Freshness

Last updated 2026-07-17 02:41:31; freshness should be verified.