• Classified by Topic • Classified by Publication Type • Sorted by Date • Sorted by First Author Last Name • Classified by Funding Source •
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, and Josiah
Hanna.
In International Conference on Robotics and Automation (ICRA), May 2025.
Robot decision-making in partially observable, real-time, dynamic, andmulti-agent environments remains a difficult and unsolved challenge. Model-freereinforcement learning (RL) is a promising approach to learning decision-makingin such domains, however, end-to-end RL in complex environments is oftenintractable. To address this challenge in the RoboCup Standard Platform League(SPL) domain, we developed a novel architecture integrating RL within a classicalrobotics stack, while employing a multi-fidelity sim2real approach anddecomposing behavior into learned sub-behaviors with heuristic selection. Ourarchitecture led to victory in the 2024 RoboCup SPL Challenge Shield Division. Inthis work, we fully describe our system's architecture and empirically analyzekey design decisions that contributed to its success. Our approach demonstrateshow RL-based behaviors can be integrated into complete robot behaviorarchitectures.
@InProceedings{zhihan_wang_ICRA25,
author = {Adam Labiosa and Zhihan Wang and Siddhant Agarwal and William Cong and Geethika Hemkumar and Abhinav Narayan Harish and Benjamin Hong and Josh Kelle and Chen Li and Yuhao Li and Zisen Shao and Peter Stone and Josiah Hanna},
title = {Reinforcement Learning within the Classical Robotics Stack: A Case Study in Robot Soccer},
booktitle = {International Conference on Robotics and Automation (ICRA)},
year = {2025},
month = {May},
location = {Atlanta, United States},
abstract = {Robot decision-making in partially observable, real-time, dynamic, and
multi-agent environments remains a difficult and unsolved challenge. Model-free
reinforcement learning (RL) is a promising approach to learning decision-making
in such domains, however, end-to-end RL in complex environments is often
intractable. To address this challenge in the RoboCup Standard Platform League
(SPL) domain, we developed a novel architecture integrating RL within a classical
robotics stack, while employing a multi-fidelity sim2real approach and
decomposing behavior into learned sub-behaviors with heuristic selection. Our
architecture led to victory in the 2024 RoboCup SPL Challenge Shield Division. In
this work, we fully describe our system's architecture and empirically analyze
key design decisions that contributed to its success. Our approach demonstrates
how RL-based behaviors can be integrated into complete robot behavior
architectures.
},
}
Generated by bib2html.pl (written by Patrick Riley ) on Sat Mar 08, 2025 23:11:31