RegoAI

• This project is the development of an AI that can play the Blokus board game through algorithms and machine learning.

• Through appropriate sorting algorithms and pruning of weaker moves, RegoAI aims to have multiple variables that determine aggression, maximization of perceived area and the blocking of enemy piece vertices.

• Similar to Stockfish, an evaluation bar will be able to determine the approximate game state for each player.

  • License
  • Frequently Asked Questions
  • Planned Features
  • Release History
  • Contributions
  • Bugs

License

GNU GPLv3

I do not own or profit off of Blokus©, the intellectual property of Mattel© and this project does not reflect the ideas or beliefs of Mattel© or any other intellectual property owned by Mattel©. All other product names and associated designs are trademarks and copyrighted properties of their respective owners. ©2013 Mattel. All Rights Reserved.

Frequently Asked Questions

• WIP

Planned Features

• Bound by 20 x 20 tile grid (400 tiles ² )

• Only allow player starts be from the corner

• Only allow piece vertex collision

• Have each piece be its own variable

• Place all pieces on the board to minimise score

  (best score is 0; each player starts with 89 points [w])

• Decide on which algorithm to use (depth/best first)

  • Monte Carlo Search Tree
  • minimax (4 player ?)
  • Heuristic Search (Machine Learning)
  • Multiplayer alpha beta pruning
  • Combinatorial Game Theory
  • A* Depth First Search

• Develop an AutoSolver in the endgame (10 > moves available)

  • Skip function if no moves are available

• Evaluation algorithm :

Increase "m" score if approaching middle

+1 if in (20 x 20) - (10 x 10) of square

+2 if in the (10 x 10) - (4 x 4) outline inside

+3 if in the 4 x 4 middle

Decrease "p" score when using -5 pieces

Increase "e" score if % area of 20x20 where x > 25%

Area will always be within 400 t ²

Use of integration for wave-like functions or pythagoras for right-angled triangles or length x width

How can I maximize inward facing vertexes?

• Find a way to condense games into simplifed form similar to Chess
• Aggression metre through AI behaviour to limit opponent block verticies
• Pattern recognition through learning to block incoming threats and pressure
• UI/UX

Release History

• 0.0.3 (2022-10-01)
  • Created red, green, blue, yellow polyomino assets for RegoAI
• 0.0.2 (2022-09-29)
  • Discussed development plan and gathered advice from university professors on how to execute the project.
• 0.0.1 (2022-09-20)
  • Officially began development

Contributions

Refer to the Contributing.md file located in the root file of RegoAI.

Bugs

Currently no known bugs. (wow)