R2D-RL: A RoboCup 2D Soccer Environment for Multi-Agent Reinforcement Learning
Researchers have developed R2D-RL, a new reinforcement learning environment designed to bridge the gap between the RoboCup 2D Soccer Simulation (RCSS2D) platform and modern Python-based multi-agent reinforcement learning (MARL) workflows. This environment facilitates easier integration with RCSS2D and HELIOS player clients through shared-memory communication and cycle-level synchronization. R2D-RL offers features such as configurable opponents, hybrid action spaces, action masks, reward shaping, and parallel execution, supporting both full-field and scenario-based training. AI
IMPACT Simplifies MARL research for complex robotic systems like robot soccer.