Dota2 Reinforcement-Learning Bot
KEYWORDS: DOTA2, bot, duel, reinforcement learning
Compared to hardcode bot, Dota2 bot created in machine learning way is proved to be much stronger[OpenAi blog]. The challenge of Dota2 ML-bot is vast amount of information available at every frame, and the continuous set of actions possible. In this case, Dota2 ML-bot is usually simplified. For example:
IN THIS PROJECT
we focus on training a high performance bots(custom items and abilitys, fixed area) on duel, which is useful to break rumours like Better Duel Hero
or The Best Duel Hero
.inaccessible in Mac OS
)eg. rumourbreaker_duel
)content/../panorama/
content/../particles/
game/../scripts/
into the corresponding path (overwrite the file, and addons path is easy to find in steam cilent dota2 properties->local file->brower local file
)dota_launch_custom_game <name of your addon> dota
https://github.com/ModDota/Dota2AIFramework
https://github.com/BeyondGodlikeBot/CreepBlockAI
https://developer.valvesoftware.com/wiki/Dota_2_Workshop_Tools