mcguile / thesis Goto Github PK

This repo houses accompanying files for my postgraduate dissertation. The dissertation itself is available upon request.

To play the game from a random state with all pieces on the board:

• From terminal, CD into thesis/src
• python3 game_with_ui.py will run the game
• Optional flag -f [filename] allows you to log the game to a text file
• Either point and click to select pieces or press keys to execute AI for a single move
  • R key executes a random move
  • M key executes MCTS
  • S key executes Swarm AI

To play a game from a saved log file:

• From terminal, CD into thesis/src
• python3 log_replay.py -f [log_file]
• Optional flag -ai to start control after the random moves
• Use LEFT and RIGHT arrow keys to go back and forward in time
• The terminal displays if a random move or an AI move is being executed

• game - all algorithms and logic to play the game of Hive, mainly insect move algorithms
• game_with_ui - directly executable to play a game using the UI, with each move given the option to be controlled manually or by AI (see instructions above)
• game_without_ui - can be instantiated and started with .play_full_game() to play a game with only AIs and save to logs
• insects - base classes for each playable insect, including meta-insects Stack and Blank
• log_replay - directly executable via the above instructions to replay a log using the UI
• mcts - implementation of root-parallelised Monte-Carlo Search Tree
• state - data class holding the current state of the game
• board - data class holding the matrix of insects/meta-insects
• action - data class holding a single action for MCTS to execute
• swarm - implementation of PSO, allowing for global/local PSO and intentions communications
• tests - test class to gather logs
• utils - helper functions which do not require game state
• count_hives - Depth-first search to count the number of Hives present on the board (One-Hive Rule)