Minesweeper AI

An AI agent for playing Minesweeper by making safe logical moves to avoid mines—a fun demo.

This project is an AI-powered Minesweeper solver that uses propositional logic to play the classic game. The AI reasons about the game board using logical sentences and makes safe moves whenever possible, falling back to random moves only when necessary.

Here’s a demo video

Project Overview

The project consists of three main components:

1. Minesweeper Game Engine

  • Implements the game board with a customizable height, width, and number of mines.
  • Randomly places mines and tracks revealed cells.
  • Provides utility methods:
    • nearby_mines(cell) → counts how many mines are adjacent to a given cell.
    • is_mine(cell) → checks if a cell contains a mine.
    • won() → determines if all mines have been flagged.
  • Includes a text-based board visualization for debugging.

2. Sentence (Logical Representation)

  • Encodes logical knowledge about the game.
  • Each Sentence represents:
    • A set of cells on the board.
    • A count of how many of those cells contain mines.
  • Provides inference capabilities:
    • known_mines() → identifies cells that must be mines.
    • known_safes() → identifies cells that must be safe.
    • mark_mine(cell) → updates sentences when a cell is confirmed as a mine.
    • mark_safe(cell) → updates sentences when a cell is confirmed safe.

3. Minesweeper AI

  • The reasoning engine that plays the game.
  • Maintains sets of:
    • Moves made (to avoid repeats).
    • Known safes and known mines.
    • A knowledge base of sentences about the board.
  • Key Methods:
    • add_knowledge(cell, count)
      • Marks a cell as safe.
      • Adds a new logical sentence based on nearby unrevealed cells and mine counts.
      • Updates the knowledge base by inferring additional safe/mine cells.
      • Uses subset reasoning to derive new knowledge (e.g., if {A, B, C} = 2 and {A, B} = 1, then {C} = 1).
    • make_safe_move()
      • Selects a cell that is logically guaranteed to be safe.
    • make_random_move()
      • Selects a random move when no safe moves are known, avoiding flagged mines.

How the AI Thinks

  1. Player clicks a cell → game reveals number of nearby mines.
  2. AI adds knowledge → creates logical constraints from that number.
  3. Logical inference → deduces safe cells or mines.
  4. Decision-making:
    • Prefer safe moves (guaranteed by logic).
    • If no safe move is available, make a random move.

This allows the AI to play Minesweeper in a human-like reasoning process.

Skills Demonstrated

  • Artificial Intelligence & Logic → Representing knowledge with logical sentences.
  • Game Development → Implemented a custom Minesweeper engine.
  • Algorithm Design → Inference rules for subset relationships.
  • Python Programming → OOP design with classes, sets, and randomization.

Takeaways

This project showcases how knowledge-based AI can be applied to reasoning under uncertainty. Instead of brute-force simulation, the AI uses logical deduction to safely progress through the game — just like a human player.