Reinforcement Learning Within the Classical Robotics Stack: A Case Study in Robot Soccer

Adam Labiosa, Zhihan Wang, Siddhant Agarwal, William Cong, Geethika Hemkumar, Abhinav Narayan Harish, Benjamin Hong, Josh Kelle, Chen Li, Yuhao Li, Zisen Shao, Peter Stone, Josiah P. Hanna·December 12, 2024

Summary

A novel reinforcement learning architecture was developed for the RoboCup Standard Platform League, integrating RL within a classical robotics stack. This approach, employing a multi-fidelity sim2real method and decomposing behavior into learned sub-behaviors with heuristic selection, led to victory in the 2024 SPL Challenge Shield Division. The system demonstrates how RL can be integrated into complete robot behavior architectures, offering insights for roboticists aiming to apply RL in complex, multi-agent environments.

Key findings

8

Advanced features