SMARTS: An Open-Source Scalable Multi-Agent RL Training School for Autonomous Driving

Ming Zhou, Jun Luo, Julian Villella, Yaodong Yang, David Rusu, Jiayu Miao, Weinan Zhang, Montgomery Alban, IMAN FADAKAR, Zheng Chen, Chongxi Huang, Ying Wen, Kimia Hassanzadeh, Daniel Graves, Zhengbang Zhu, Yihan Ni, Nhat Nguyen, Mohamed Elsayed, Haitham Ammar, Alexander Cowen-Rivers, Sanjeevan Ahilan, Zheng Tian, Daniel Palenicek, Kasra Rezaee, Peyman Yadmellat, Kun Shao, dong chen, Baokuan Zhang, Hongbo Zhang, Jianye Hao, Wulong Liu, Jun Wang
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:264-285, 2021.

Abstract

Interaction is fundamental in autonomous driving (AD). Despite more than a decade of intensive R&D in AD, how to dynamically interact with diverse road users in various contexts still remains unsolved. Multi-agent learning has recently seen big breakthroughs and has much to offer towards solving realistic interaction in AD. However, to realize this potential we need multi-agent AD simulation of realistic interaction. To break this apparent chicken-and-egg circularity, we built an AD simulation platform called SMARTS (Scalable Multi-Agent Rl Training School), which is designed to accumulate behavior models of road users towards increasingly realistic and diverse interaction that in turn enables deeper and broader multi-agent research on interaction. In this paper, we describe the design goals of SMARTS, explain its key architectural ideas, illustrate its use for multi-agent research through experiments on concrete interaction scenarios, and introduce a set of benchmarks and metrics. As an open-source, industrial-strength platform, the future of SMARTS lies in its growth along with the multi-agent research it enables in the years to come.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-zhou21a, title = {SMARTS: An Open-Source Scalable Multi-Agent RL Training School for Autonomous Driving}, author = {Zhou, Ming and Luo, Jun and Villella, Julian and Yang, Yaodong and Rusu, David and Miao, Jiayu and Zhang, Weinan and Alban, Montgomery and FADAKAR, IMAN and Chen, Zheng and Huang, Chongxi and Wen, Ying and Hassanzadeh, Kimia and Graves, Daniel and Zhu, Zhengbang and Ni, Yihan and Nguyen, Nhat and Elsayed, Mohamed and Ammar, Haitham and Cowen-Rivers, Alexander and Ahilan, Sanjeevan and Tian, Zheng and Palenicek, Daniel and Rezaee, Kasra and Yadmellat, Peyman and Shao, Kun and chen, dong and Zhang, Baokuan and Zhang, Hongbo and Hao, Jianye and Liu, Wulong and Wang, Jun}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {264--285}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/zhou21a/zhou21a.pdf}, url = {https://proceedings.mlr.press/v155/zhou21a.html}, abstract = {Interaction is fundamental in autonomous driving (AD). Despite more than a decade of intensive R&D in AD, how to dynamically interact with diverse road users in various contexts still remains unsolved. Multi-agent learning has recently seen big breakthroughs and has much to offer towards solving realistic interaction in AD. However, to realize this potential we need multi-agent AD simulation of realistic interaction. To break this apparent chicken-and-egg circularity, we built an AD simulation platform called SMARTS (Scalable Multi-Agent Rl Training School), which is designed to accumulate behavior models of road users towards increasingly realistic and diverse interaction that in turn enables deeper and broader multi-agent research on interaction. In this paper, we describe the design goals of SMARTS, explain its key architectural ideas, illustrate its use for multi-agent research through experiments on concrete interaction scenarios, and introduce a set of benchmarks and metrics. As an open-source, industrial-strength platform, the future of SMARTS lies in its growth along with the multi-agent research it enables in the years to come.} }
Endnote
%0 Conference Paper %T SMARTS: An Open-Source Scalable Multi-Agent RL Training School for Autonomous Driving %A Ming Zhou %A Jun Luo %A Julian Villella %A Yaodong Yang %A David Rusu %A Jiayu Miao %A Weinan Zhang %A Montgomery Alban %A IMAN FADAKAR %A Zheng Chen %A Chongxi Huang %A Ying Wen %A Kimia Hassanzadeh %A Daniel Graves %A Zhengbang Zhu %A Yihan Ni %A Nhat Nguyen %A Mohamed Elsayed %A Haitham Ammar %A Alexander Cowen-Rivers %A Sanjeevan Ahilan %A Zheng Tian %A Daniel Palenicek %A Kasra Rezaee %A Peyman Yadmellat %A Kun Shao %A dong chen %A Baokuan Zhang %A Hongbo Zhang %A Jianye Hao %A Wulong Liu %A Jun Wang %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-zhou21a %I PMLR %P 264--285 %U https://proceedings.mlr.press/v155/zhou21a.html %V 155 %X Interaction is fundamental in autonomous driving (AD). Despite more than a decade of intensive R&D in AD, how to dynamically interact with diverse road users in various contexts still remains unsolved. Multi-agent learning has recently seen big breakthroughs and has much to offer towards solving realistic interaction in AD. However, to realize this potential we need multi-agent AD simulation of realistic interaction. To break this apparent chicken-and-egg circularity, we built an AD simulation platform called SMARTS (Scalable Multi-Agent Rl Training School), which is designed to accumulate behavior models of road users towards increasingly realistic and diverse interaction that in turn enables deeper and broader multi-agent research on interaction. In this paper, we describe the design goals of SMARTS, explain its key architectural ideas, illustrate its use for multi-agent research through experiments on concrete interaction scenarios, and introduce a set of benchmarks and metrics. As an open-source, industrial-strength platform, the future of SMARTS lies in its growth along with the multi-agent research it enables in the years to come.
APA
Zhou, M., Luo, J., Villella, J., Yang, Y., Rusu, D., Miao, J., Zhang, W., Alban, M., FADAKAR, I., Chen, Z., Huang, C., Wen, Y., Hassanzadeh, K., Graves, D., Zhu, Z., Ni, Y., Nguyen, N., Elsayed, M., Ammar, H., Cowen-Rivers, A., Ahilan, S., Tian, Z., Palenicek, D., Rezaee, K., Yadmellat, P., Shao, K., chen, d., Zhang, B., Zhang, H., Hao, J., Liu, W. & Wang, J.. (2021). SMARTS: An Open-Source Scalable Multi-Agent RL Training School for Autonomous Driving. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:264-285 Available from https://proceedings.mlr.press/v155/zhou21a.html.

Related Material