Toward Real-World Cooperative and Competitive Soccer with Quadrupedal Robot Teams

Zhi Su, Yuman Gao, Emily Lukas, Yunfei Li, Jiaze Cai, Faris Talubah, Fei Gao, Chao Yu, Zhongyu Li, Yi Wu, Koushil Sreenath
Proceedings of The 9th Conference on Robot Learning, PMLR 305:4723-4742, 2025.

Abstract

Achieving coordinated teamwork among legged robots requires both fine-grained locomotion control and long-horizon strategic decision-making. Robot soccer offers a compelling testbed for this challenge, combining dynamic, competitive, and multi-agent interactions. In this work, we present a hierarchical multi-agent reinforcement learning (MARL) framework that enables fully autonomous and decentralized quadruped robot soccer. First, a set of highly dynamic low-level skills is trained for legged locomotion and ball manipulation, such as walking, dribbling, and kicking. On top of these, a high-level strategic planning policy is trained with Multi-Agent Proximal Policy Optimization (MAPPO) via Fictitious Self-Play (FSP). This learning framework allows agents to adapt to diverse opponent strategies and gives rise to sophisticated team behaviors, including coordinated passing, interception, and dynamic role allocation. With an extensive ablation study, the proposed learning method shows significant advantages in the cooperative and competitive multi-agent soccer game. We deploy the learned policies to real quadruped robots relying solely on onboard proprioception and decentralized localization, with the resulting system supporting autonomous robot-robot and robot-human soccer matches on indoor and outdoor soccer courts.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-su25a, title = {Toward Real-World Cooperative and Competitive Soccer with Quadrupedal Robot Teams}, author = {Su, Zhi and Gao, Yuman and Lukas, Emily and Li, Yunfei and Cai, Jiaze and Talubah, Faris and Gao, Fei and Yu, Chao and Li, Zhongyu and Wu, Yi and Sreenath, Koushil}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {4723--4742}, year = {2025}, editor = {Lim, Joseph and Song, Shuran and Park, Hae-Won}, volume = {305}, series = {Proceedings of Machine Learning Research}, month = {27--30 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v305/main/assets/su25a/su25a.pdf}, url = {https://proceedings.mlr.press/v305/su25a.html}, abstract = {Achieving coordinated teamwork among legged robots requires both fine-grained locomotion control and long-horizon strategic decision-making. Robot soccer offers a compelling testbed for this challenge, combining dynamic, competitive, and multi-agent interactions. In this work, we present a hierarchical multi-agent reinforcement learning (MARL) framework that enables fully autonomous and decentralized quadruped robot soccer. First, a set of highly dynamic low-level skills is trained for legged locomotion and ball manipulation, such as walking, dribbling, and kicking. On top of these, a high-level strategic planning policy is trained with Multi-Agent Proximal Policy Optimization (MAPPO) via Fictitious Self-Play (FSP). This learning framework allows agents to adapt to diverse opponent strategies and gives rise to sophisticated team behaviors, including coordinated passing, interception, and dynamic role allocation. With an extensive ablation study, the proposed learning method shows significant advantages in the cooperative and competitive multi-agent soccer game. We deploy the learned policies to real quadruped robots relying solely on onboard proprioception and decentralized localization, with the resulting system supporting autonomous robot-robot and robot-human soccer matches on indoor and outdoor soccer courts.} }
Endnote
%0 Conference Paper %T Toward Real-World Cooperative and Competitive Soccer with Quadrupedal Robot Teams %A Zhi Su %A Yuman Gao %A Emily Lukas %A Yunfei Li %A Jiaze Cai %A Faris Talubah %A Fei Gao %A Chao Yu %A Zhongyu Li %A Yi Wu %A Koushil Sreenath %B Proceedings of The 9th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Joseph Lim %E Shuran Song %E Hae-Won Park %F pmlr-v305-su25a %I PMLR %P 4723--4742 %U https://proceedings.mlr.press/v305/su25a.html %V 305 %X Achieving coordinated teamwork among legged robots requires both fine-grained locomotion control and long-horizon strategic decision-making. Robot soccer offers a compelling testbed for this challenge, combining dynamic, competitive, and multi-agent interactions. In this work, we present a hierarchical multi-agent reinforcement learning (MARL) framework that enables fully autonomous and decentralized quadruped robot soccer. First, a set of highly dynamic low-level skills is trained for legged locomotion and ball manipulation, such as walking, dribbling, and kicking. On top of these, a high-level strategic planning policy is trained with Multi-Agent Proximal Policy Optimization (MAPPO) via Fictitious Self-Play (FSP). This learning framework allows agents to adapt to diverse opponent strategies and gives rise to sophisticated team behaviors, including coordinated passing, interception, and dynamic role allocation. With an extensive ablation study, the proposed learning method shows significant advantages in the cooperative and competitive multi-agent soccer game. We deploy the learned policies to real quadruped robots relying solely on onboard proprioception and decentralized localization, with the resulting system supporting autonomous robot-robot and robot-human soccer matches on indoor and outdoor soccer courts.
APA
Su, Z., Gao, Y., Lukas, E., Li, Y., Cai, J., Talubah, F., Gao, F., Yu, C., Li, Z., Wu, Y. & Sreenath, K.. (2025). Toward Real-World Cooperative and Competitive Soccer with Quadrupedal Robot Teams. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:4723-4742 Available from https://proceedings.mlr.press/v305/su25a.html.

Related Material