Puzzle Solver Project

Welcome to the Puzzle Solver Project repository! This is a central hub that introduces all my puzzle game solver projects. Each solver is designed to tackle a specific type of puzzle using efficient algorithms, data structures, and innovative strategies.

Whether you are passionate about solving puzzles or intrigued by the technical challenges of developing algorithmic solvers, this project aims to inspire and offer comprehensive solutions. Explore the innovative methods and strategies employed to tackle complex puzzles programmatically.

🧩 Featured Puzzle Solutions

1. Minesweeper Solver

An intelligent solver that plays Minesweeper using computer vision and probability-based decision-making.

beginner_gameplay.mp4

Features

• Computer Vision Integration: Uses OpenCV and Tesseract OCR to detect and analyze the game board.
• Dynamic Decision-Making: Employs probability-based heuristic algorithms for optimal moves.
• Visualization Tools: Provides real-time progress updates and debugging visualizations for enhanced interactivity.
• Automated Gameplay: Integrates screen capture, mouse control, and real-time game state analysis for seamless play.

2. Solitaire Solver

An intelligent program for solving Klondike Solitaire games using computer vision, OCR, and heuristic algorithms to analyze and execute optimal moves.

gameplay_example.mp4

Features

• Game Solving Engine: Advanced heuristic-based solver prioritizing empty tableau spaces, uncovering face-down cards, and building foundation sequences.
• Computer Vision System:
  • Robust card detection using advanced contour techniques.
  • Accurate suit and value recognition with ORB feature matching and Tesseract OCR.
  • Handles overlapping and partially visible cards.
• Automation: Seamless interaction with Solitaire through PyAutoGUI, including real-time card manipulation and move execution.

3. Wordbox Solver

An automated solver for Wordbox puzzles that uses computer vision and optimized algorithms to detect the game board, find word combinations, and maximize scores.

Features

• Automated Game Detection: Leverages computer vision and OCR to analyze the board.
• Efficient Word Finding: Uses a Trie data structure to quickly identify all valid words.
• Score Optimization: Accounts for bonus tiles and word lengths to maximize scoring potential.
• Natural Gameplay: Simulates human-like mouse movements with subtle variations for realism.

4. Maze Solver

A high-performance C++ maze generation and solving engine that creates complex mazes and solves them using classical search algorithms as well as ML powered by LibTorch.

Features

• Maze Generation Engine

  • Generates perfect mazes using DFS backtracking.

  • Adds natural variability using Monte Carlo wall-removal, creating more complex and realistic structures.

• Multiple Solving Algorithms

  • DFS: Fast recursive exploration, finds a path but not always optimal.

  • BFS: Guarantees the shortest path, but explores many states.

  • A* Search: Uses Manhattan heuristic to find the shortest path with significantly fewer visited nodes.

  • DQN-based ML Solver: Trained with Deep Q-Learning to solve unseen mazes. Sometimes discovers efficient paths with fewer visits, but performance depends on training quality.