metadata
license: bsd-2-clause
pipeline_tag: robotics
Spatial attention recurrent neural network (SARNN)
This repository contains a SARNN (Spatial attention recurrent neural network) model trained with the MujocoUR5eCable dataset.
The model is part of RoboManipBaselines, a unified framework for imitation learning in robotic manipulation across real and simulation environments.
[Paper] [Project Page] [GitHub]
Install
See GitHub for installation instructions.
Policy rollout
To run a trained policy, navigate to the top directory of the robo_manip_baselines repository and run:
# Go to the top directory of this repository
$ cd robo_manip_baselines
$ python ./bin/Rollout.py Sarnn MujocoUR5eCable --checkpoint ./checkpoint/Sarnn/<checkpoint_name>/policy_last.ckpt
Technical Details
For more information on the technical details of the SARNN architecture, please see the following paper:
@INPROCEEDINGS{SARNN_ICRA2022,
author = {Ichiwara, Hideyuki and Ito, Hiroshi and Yamamoto, Kenjiro and Mori, Hiroki and Ogata, Tetsuya},
title = {Contact-Rich Manipulation of a Flexible Object based on Deep Predictive Learning using Vision and Tactility},
booktitle = {International Conference on Robotics and Automation},
year = {2022},
pages = {5375-5381},
doi = {10.1109/ICRA46639.2022.9811940}
}
Citation
If you use this framework or model in your work, please cite the RoboManipBaselines paper:
@article{RoboManipBaselines_Murooka_2025,
title={RoboManipBaselines: A Unified Framework for Imitation Learning in Robotic Manipulation across Real and Simulation Environments},
author={Murooka, Masaki and Motoda, Tomohiro and Nakajo, Ryoichi and Oh, Hanbit and Makihara, Koshi and Shirai, Keisuke and Ogata, Tetsuya and Domae, Yukiyasu},
journal={arXiv preprint arXiv:2509.17057},
year={2025}
}