Wii Reinforcement Learning (github.com)

A new GitHub repository has been launched, allowing Reinforcement Learning (RL) agents to play Wii games, with a specific emphasis on Mario Kart Wii's Luigi Circuit against Hard CPUs. The project features the "Beyond The Rainbow" (BTR) algorithm, which achieves first-place finishes after approximately one day of training on powerful hardware like the RTX 4090, though it remains usable on less advanced systems with extended training times. The BTR algorithm has been accepted for poster presentation at ICML 2025, highlighting its significance in advancing RL techniques in gaming environments. This repository offers detailed instructions for setup, requiring Python 3.12 and various dependencies to enable interaction with the Dolphin emulator. Users are expected to manage their game ROM independently, as legal restrictions prevent the repository from providing it. The project also introduces the capability to run multiple instances of Dolphin for training, accommodating various machine configurations. Utilizing this framework not only fosters accessibility to reinforcement learning methodologies for gaming but also encourages the broader AI/ML community to explore innovative applications within the gaming sector, paving the way for future developments in autonomous agents and gameplay strategies.