Skip to content

Sandbox warcraft 3 without any graphics. Trying to code a way to use reinforcement learning with Q Networking and MCTS in Real time strategy games. Bellman Equation to predict future prediction of state and action pairs.

Notifications You must be signed in to change notification settings

Skyfrei/Warcraft3

Repository files navigation

Simplified Warcraft 3 version with DeepRL

Sandbox warcraft 3 without any graphics. Trying to code a way to use reinforcement learning and MTCS in Real time strategy games.

Game will not have any graphics for a long time if ever because the point of the project is to create an RTS game which can be deeplearned with different algorithms, following previously published licensed papers. At the end of the project the AI will be able to go toe to toe with a human enemy and make the best possible decisions to win the game.

How can you collaborate

  • Bug finding in the game in case i have missed something such as times when Id get undefined behaviours.
  • Probably safe here as i have implemented null pointers but just in case that would be nice to check.
  • Ideas on how to continue the Reinforcement learning and stuff i could add to it.
  • Better writing of structs and enums (this would actually be really helpful)

What has currently been implemented

  • Game is fully finished and playable although not very interesting without graphics as its a real time strategy game.
  • Algorithm for reinforcement learning has been implemented
  • PyTorch has been sucessfully started running.
  • Currently changing the input of the game so it fits in my tensor model.

About

Sandbox warcraft 3 without any graphics. Trying to code a way to use reinforcement learning with Q Networking and MCTS in Real time strategy games. Bellman Equation to predict future prediction of state and action pairs.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published