RoboManipBaselines

Software that integrates various imitation learning methods and benchmark task environments to provide baselines for robot manipulation

RoboManipBaselines_VideoVer100.mp4

Quick start

This quick start allows you to collect data in the MuJoCo simulation and train and rollout the ACT policy.

Install

See the installation documentation.

Policies

SARNN

Spatial attention recurrent neural network

ACT

Action Chunking with Transformers

DiffusionPolicy

Diffusion Policy

MT-ACT

Multi-Task Action Chunking Transformer

Data

Publicly available datasets

See the dataset list.

See the learned parameters for policies learned from these datasets.

Data collection by teleoperation

See teleop.

Environments for robot manipulation

See the environment catalog for a full list of environments.

See envs for installation procedures for each environment.

Utilities

See utils.

Evaluation results

See the evaluation results.

Contribute

If you would like to contribute to this repository, please check out the contribution guide.

License

Files that originate from this repository are subject to the BSD 2-Clause License. If a file explicitly states a different license, or if there are different license files in a directory, those licenses will take precedence. For files in third-party directories, please follow the respective licenses.

Citation

You can cite this work with:

@software{RoboManipBaselines_GitHub2024,
author = {Murooka, Masaki and Motoda, Tomohiro and Nakajo, Ryoichi},
title = {{RoboManipBaselines}},
url = {https://github.com/isri-aist/RoboManipBaselines},
version = {1.0.0},
year = {2024}
month = dec,
}