• RootAI
  • Table of Contents
  • Running the Project
    • Prerequisite
    • Setup
      • For IntelliJ
      • For VSCode
    • Running the Game
      • Option 1 - Run inside IntelliJ
      • Option 2 - Run via command line
    • Exporting
  • Download and Running the exe
  • Controls
  • Technical
    • The AI agent
    • Action generation

It is RECOMMENDED that you:

• know Root boardgame's rules
• or have played before
• or have a rule book with you

• python 11.3+
• libraries
  • pygame 2.4.0+ — pip install pygame
    • (not recommended, un-maintained) conda install -c conda-forge pygame
  • yaml — pip install pyyaml
  • numpy — pip install numpy
  • pathos — pip install pathos
  • scipy — pip install scipy

1. Add a new Python configuration (top right)
   • set name: Run main.py
   • set Script path: .\main.py
   • set Environment > Working directory to absolute path to src
     • eg. D:\GitHub\RootAI\src
2. Go to Settings > CodeStyle > Cogs icon > Import scheme > import codestyle.xml

1. You should be able to just run via command line with the instruction below

1. Run Run main.py to run the game

1. cd to .\src
2. run python with config file path as arg python -m main ".\config\config.yml"

1. cd to root of project
2. Run pyinstaller .\src\main.py
3. Copy .\src\assets to .\dist\main and \.src\config\config.yml to .\dist\main\config

1. download the RootAI from the releases tab: https://github.com/iambaangkok/RootAI/releases
2. either:
   1. double click RootAI.exe
   2. run via command line ./RootAI.exe
3. config file is in /config/config.yml, change to match your needs.
   1. Eyrie is set to be the mcts AI agent
   2. Marquise is the player
   3. Eyrie starts first

No mouse controls

• Press UP/DOWN --- change selected action
• Press R --- random action
• Press RETURN/SPACE --- execute selected action
• Press O --- print_game_state
• Press C --- new_game_from_current_game_state
• Hold F --- continuously run player
• Hold A --- continuously run agent
• Press N in game-end state --- new game
• Press Q in game-end state --- quit game

The AI agent uses Monte Carlo Tree Search (MCTS) to simulate into to future and pick the "best" possible action at the current game state.

Each action generation sequence corresponds to a 5 digit-number.

• Digit 0: faction = 1 (Marquise) | 2 (Eyrie) | 9 (Neutral)
• Digit 1-4: distinct id for that method

e.g.

• 10001 = action generation sequence 1 for Marquise