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

TL;DR

This paper presents a novel architecture integrating reinforcement learning into a classical robotics stack for robot soccer, demonstrating successful real-world application and empirical analysis of design choices leading to victory in the RoboCup SPL Challenge.

Contribution

It introduces a new RL integration framework with multi-fidelity sim2real transfer and behavior decomposition, advancing practical robot decision-making in complex multi-agent environments.

Findings

Achieved victory in the 2024 RoboCup SPL Challenge Shield Division.

Demonstrated effective integration of RL into complete robot behavior architectures.

Provided empirical analysis of key design decisions impacting success.

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.