AI Learns to Exploit a Glitch in a Racing Game

Video Description

I trained an AI in Trackmania with reinforcement learning, and tried to make it learn the hardest technique in this game: the noseboost. To support my work on Patreon: • https://www.patreon.com/Yoshtm Note: This video was re-uploaded a few hours after the original upluoad due to a copyright issue with the music Contact • Discord: yosh_tm • Twitter:  yoshtm1   • Mail: [email protected] Link to download the last map: https://tmnf.exchange/trackshow/12306053 TAS runs shown in the video: A01-Race, 17.61 by igntul & Lukalyc: https://www.youtube.com/watch?v=-GXxiQ7QtQA A09-Race, 21.08 by gingerroar & plastorex: https://www.youtube.com/watch?v=1QsSRru6mwM D12-Speed, 19.94 by threadd: https://www.youtube.com/watch?v=DxPGft3ygNw Other reinforcement learning projects mentioned in the video: GT Sophy: https://www.youtube.com/watch?v=l948hMaTPuo OpenAI Five: https://www.youtube.com/watch?v=UZHTNBMAfAA RLGym: https://www.twitch.tv/rlgym?lang=fr Linesight: https://www.youtube.com/@linesight-rl Thanks to Donadigo, for TMInterface! https://donadigo.com/tminterface/ The setup used in all maps to put the car on its nose was inspired by this map: https://tmnf.exchange/trackshow/9042183 On a more general note, this video would not have been possible without the passion and creativity of the TAS community, who discovered the noseboost in 2021. Special thanks to them :) More information on Trackmania TAS here: https://discord.com/invite/tD4rarRYpj You can find a list of the musics I used at the end of the video.