Robomimic: Robot Manipulation Demonstration Datasets

A collection of demonstration datasets for robot manipulation domains, released with the Robomimic framework. The datasets are intended for benchmarking tasks and developing offline robot learning algorithms. The repository was last updated on April 13, 2025.

Use Cases

• Benchmarking robot learning algorithms based on standardized demonstration datasets.
• Developing offline reinforcement learning methods for manipulation tasks based on recorded demonstrations.
• Training imitation learning models based on robot manipulation demonstrations.

Strengths

• Datasets are released with a framework designed for reproducibility and broad accessibility.
• The description indicates a broad set of demonstration datasets across multiple robot manipulation domains.

Limitations

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

Provenance

Source

robomimic

Collection Method

Demonstration data collected on robot manipulation domains.

Freshness

Last updated 2025-04-13 21:45:01; freshness should be verified.