GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots

Gilbert Feng, Hongbo Zhang, Zhongyu Li, Xue Bin Peng, Bhuvan Basireddy, Linzhu Yue, ZHITAO SONG, Lizhi Yang, Yunhui Liu, Koushil Sreenath, Sergey Levine
Proceedings of The 6th Conference on Robot Learning, PMLR 205:1893-1903, 2023.

Abstract

Recent years have seen a surge in commercially-available and affordable quadrupedal robots, with many of these platforms being actively used in research and industry. As the availability of legged robots grows, so does the need for controllers that enable these robots to perform useful skills. However, most learning-based frameworks for controller development focus on training robot-specific controllers, a process that needs to be repeated for every new robot. In this work, we introduce a framework for training generalized locomotion (GenLoco) controllers for quadrupedal robots. Our framework synthesizes general-purpose locomotion controllers that can be deployed on a large variety of quadrupedal robots with similar morphologies. We present a simple but effective morphology randomization method that procedurally generates a diverse set of simulated robots for training. We show that by training a controller on this large set of simulated robots, our models acquire more general control strategies that can be directly transferred to novel simulated and real-world robots with diverse morphologies, which were not observed during training.

Cite this Paper

BibTeX
@InProceedings{pmlr-v205-feng23a, title = {GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots}, author = {Feng, Gilbert and Zhang, Hongbo and Li, Zhongyu and Peng, Xue Bin and Basireddy, Bhuvan and Yue, Linzhu and SONG, ZHITAO and Yang, Lizhi and Liu, Yunhui and Sreenath, Koushil and Levine, Sergey}, booktitle = {Proceedings of The 6th Conference on Robot Learning}, pages = {1893--1903}, year = {2023}, editor = {Liu, Karen and Kulic, Dana and Ichnowski, Jeff}, volume = {205}, series = {Proceedings of Machine Learning Research}, month = {14--18 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v205/feng23a/feng23a.pdf}, url = {https://proceedings.mlr.press/v205/feng23a.html}, abstract = {Recent years have seen a surge in commercially-available and affordable quadrupedal robots, with many of these platforms being actively used in research and industry. As the availability of legged robots grows, so does the need for controllers that enable these robots to perform useful skills. However, most learning-based frameworks for controller development focus on training robot-specific controllers, a process that needs to be repeated for every new robot. In this work, we introduce a framework for training generalized locomotion (GenLoco) controllers for quadrupedal robots. Our framework synthesizes general-purpose locomotion controllers that can be deployed on a large variety of quadrupedal robots with similar morphologies. We present a simple but effective morphology randomization method that procedurally generates a diverse set of simulated robots for training. We show that by training a controller on this large set of simulated robots, our models acquire more general control strategies that can be directly transferred to novel simulated and real-world robots with diverse morphologies, which were not observed during training.} }
Endnote
%0 Conference Paper %T GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots %A Gilbert Feng %A Hongbo Zhang %A Zhongyu Li %A Xue Bin Peng %A Bhuvan Basireddy %A Linzhu Yue %A ZHITAO SONG %A Lizhi Yang %A Yunhui Liu %A Koushil Sreenath %A Sergey Levine %B Proceedings of The 6th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Karen Liu %E Dana Kulic %E Jeff Ichnowski %F pmlr-v205-feng23a %I PMLR %P 1893--1903 %U https://proceedings.mlr.press/v205/feng23a.html %V 205 %X Recent years have seen a surge in commercially-available and affordable quadrupedal robots, with many of these platforms being actively used in research and industry. As the availability of legged robots grows, so does the need for controllers that enable these robots to perform useful skills. However, most learning-based frameworks for controller development focus on training robot-specific controllers, a process that needs to be repeated for every new robot. In this work, we introduce a framework for training generalized locomotion (GenLoco) controllers for quadrupedal robots. Our framework synthesizes general-purpose locomotion controllers that can be deployed on a large variety of quadrupedal robots with similar morphologies. We present a simple but effective morphology randomization method that procedurally generates a diverse set of simulated robots for training. We show that by training a controller on this large set of simulated robots, our models acquire more general control strategies that can be directly transferred to novel simulated and real-world robots with diverse morphologies, which were not observed during training.
APA
Feng, G., Zhang, H., Li, Z., Peng, X.B., Basireddy, B., Yue, L., SONG, Z., Yang, L., Liu, Y., Sreenath, K. & Levine, S.. (2023). GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. Proceedings of The 6th Conference on Robot Learning, in Proceedings of Machine Learning Research 205:1893-1903 Available from https://proceedings.mlr.press/v205/feng23a.html.

Related Material