PreCo: Enhancing Generalization in Co-Design of Modular Soft Robots via Brain-Body Pre-Training

Yuxing Wang, Shuang Wu, Tiantian Zhang, Yongzhe Chang, Haobo Fu, QIANG FU, Xueqian Wang
Proceedings of The 7th Conference on Robot Learning, PMLR 229:478-498, 2023.

Abstract

Brain-body co-design, which involves the collaborative design of control strategies and morphologies, has emerged as a promising approach to enhance a robot’s adaptability to its environment. However, the conventional co-design process often starts from scratch, lacking the utilization of prior knowledge. This can result in time-consuming and costly endeavors. In this paper, we present PreCo, a novel methodology that efficiently integrates brain-body pre-training into the co-design process of modular soft robots. PreCo is based on the insight of embedding co-design principles into models, achieved by pre-training a universal co-design policy on a diverse set of tasks. This pre-trained co-designer is utilized to generate initial designs and control policies, which are then fine-tuned for specific co-design tasks. Through experiments on a modular soft robot system, our method demonstrates zero-shot generalization to unseen co-design tasks, facilitating few-shot adaptation while significantly reducing the number of policy iterations required.

Cite this Paper

BibTeX
@InProceedings{pmlr-v229-wang23b, title = {PreCo: Enhancing Generalization in Co-Design of Modular Soft Robots via Brain-Body Pre-Training}, author = {Wang, Yuxing and Wu, Shuang and Zhang, Tiantian and Chang, Yongzhe and Fu, Haobo and FU, QIANG and Wang, Xueqian}, booktitle = {Proceedings of The 7th Conference on Robot Learning}, pages = {478--498}, year = {2023}, editor = {Tan, Jie and Toussaint, Marc and Darvish, Kourosh}, volume = {229}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v229/wang23b/wang23b.pdf}, url = {https://proceedings.mlr.press/v229/wang23b.html}, abstract = {Brain-body co-design, which involves the collaborative design of control strategies and morphologies, has emerged as a promising approach to enhance a robot’s adaptability to its environment. However, the conventional co-design process often starts from scratch, lacking the utilization of prior knowledge. This can result in time-consuming and costly endeavors. In this paper, we present PreCo, a novel methodology that efficiently integrates brain-body pre-training into the co-design process of modular soft robots. PreCo is based on the insight of embedding co-design principles into models, achieved by pre-training a universal co-design policy on a diverse set of tasks. This pre-trained co-designer is utilized to generate initial designs and control policies, which are then fine-tuned for specific co-design tasks. Through experiments on a modular soft robot system, our method demonstrates zero-shot generalization to unseen co-design tasks, facilitating few-shot adaptation while significantly reducing the number of policy iterations required.} }
Endnote
%0 Conference Paper %T PreCo: Enhancing Generalization in Co-Design of Modular Soft Robots via Brain-Body Pre-Training %A Yuxing Wang %A Shuang Wu %A Tiantian Zhang %A Yongzhe Chang %A Haobo Fu %A QIANG FU %A Xueqian Wang %B Proceedings of The 7th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Jie Tan %E Marc Toussaint %E Kourosh Darvish %F pmlr-v229-wang23b %I PMLR %P 478--498 %U https://proceedings.mlr.press/v229/wang23b.html %V 229 %X Brain-body co-design, which involves the collaborative design of control strategies and morphologies, has emerged as a promising approach to enhance a robot’s adaptability to its environment. However, the conventional co-design process often starts from scratch, lacking the utilization of prior knowledge. This can result in time-consuming and costly endeavors. In this paper, we present PreCo, a novel methodology that efficiently integrates brain-body pre-training into the co-design process of modular soft robots. PreCo is based on the insight of embedding co-design principles into models, achieved by pre-training a universal co-design policy on a diverse set of tasks. This pre-trained co-designer is utilized to generate initial designs and control policies, which are then fine-tuned for specific co-design tasks. Through experiments on a modular soft robot system, our method demonstrates zero-shot generalization to unseen co-design tasks, facilitating few-shot adaptation while significantly reducing the number of policy iterations required.
APA
Wang, Y., Wu, S., Zhang, T., Chang, Y., Fu, H., FU, Q. & Wang, X.. (2023). PreCo: Enhancing Generalization in Co-Design of Modular Soft Robots via Brain-Body Pre-Training. Proceedings of The 7th Conference on Robot Learning, in Proceedings of Machine Learning Research 229:478-498 Available from https://proceedings.mlr.press/v229/wang23b.html.

Related Material