Abstract: There has been a recent explosion in the capabilities of game-playingartificial intelligence. Many classes of RL tasks, from Atari games to motorcontrol to board games, are now solvable by fairly generic algorithms, based ondeep learning, that learn to play from experience with minimal knowledge of thespecific domain of interest. In this work, we will investigate the performanceof these methods on Super Smash Bros. Melee (SSBM), a popular console fightinggame. The SSBM environment has complex dynamics and partial observability,making it challenging for human and machine alike. The multi-player aspectposes an additional challenge, as the vast majority of recent advances in RLhave focused on single-agent environments. Nonetheless, we will show that it ispossible to train agents that are competitive against and even surpass humanprofessionals, a new result for the multi-player video game setting.
Hailing from 'Fatal Fury', Terry loves to fight with his friends. In 'Super Smash Bros. Ultimate', Terry is the 74th fighter in the roster and the fourth fighter in the 1st Fighters Pass. Ken is Ryu's friend and rival in the 'Street Fighter' franchise. He loves to have a good time in a fight. In this game, Ken is an Echo Fighter for Ryu.
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs and how to get involved.