Artificial Intelligence engine playing Quoridor. The project was developed as coursework of Reasoning and Agents course at the University of Edinburgh. It received first prize in the class competition for the best coursework.

You need GHC, the Cabal build system and the Stack tool. See https://www.haskell.org/platform/.

The easiest way to play the game is to go to the src directory, run ghci Main and execute the main function.

There are two test suites:

• Basic tests (stack test :basic-tests).
• Minimax tests (stack test :minimax-tests).

I added a couple of extensions to improve the Quoridor and its AI.

I have implemented A* available in the AStar.hs module. The algorithm uses a priority queue to schedule nodes according to:

• cost needed to go to that node, plus
• heuristic - straight vertical distance to a winning row.

To optimize further, the already visited nodes are stored in a set and checked before visiting again.

The queue is implemented using the heaps package from Hackage. This package is included in the cabal file. You may however need to install it on your machine:

cabal install heaps

My implementation adds additional types in Types.hs and is used in the utility function.

To check the correctness of my A* implementation (in particualr its optimality), I added a new test suite with unit tests. Check tests/Extension/AStarTest.hs and run:

stack test :extension-tests

I added a module Evaluator.hs in which you can type any (partial) game and check the evaluation of its final position. The module gives:

• static evaluation - evaluation with regard to just the final position (i.e. utility from Minimax.hs)
• dynamic evaluation - lookahead evaluation using Minimax

The game need to be written in a comma-separated format, e.g. "e2,d9,e3,c9". Check Reed.txt for more examples.

This evaluator is a great tool to analyze own games as well as explore possible openings.

My extended utility is calculated not only with regard to the current player, but also the opponent. If you want to play my Minimax, or use my Evaluator.hs please use it instead of the basic one.

I made sure that no player can be cut off in the game. To do this I added playable function in Game.hs and extended the validWallAction function.

Since I use lowFirst throughout my implementation, I changed the given highFirst unit tests to be relevant to lowFirst ordering. See tests/Minimax/OrderingTreeTest.hs