A* Algorithm-Based Gomoku AI Game System: Complete Implementation from Search to Visualization

A Gomoku human-machine game system based on the A* search algorithm, featuring a complete 15x15 chessboard interactive interface, AI intelligent move decision-making, game record export, and search tree visualization functions. It is suitable for artificial intelligence course design and algorithm verification.

• Original Author/Maintainer: qzddmyc
• Source Platform: GitHub
• Original Project Name: gomoku-astar
• Project Link: https://github.com/qzddmyc/gomoku-astar
• Release Date: May 30, 2026

As a classic strategy board game, Gomoku has long been an ideal experimental platform for artificial intelligence algorithm research. Its simple rules yet vast strategy space make it perfect for learning and verifying various search algorithms. This project applies the classic A* (A-Star) search algorithm to Gomoku gameplay, not only implementing human-machine battle functionality but also providing complete visualization tools, allowing learners to intuitively understand the AI's thinking process.

For artificial intelligence course design, this project offers a complete closed loop from theory to practice: students can learn the principles of the A* algorithm, observe its performance in actual games, and deeply understand the algorithm's decision path through search tree visualization.

The A* algorithm is a heuristic search algorithm widely used in path planning, game AI, and other fields. In the specific scenario of Gomoku, the algorithm needs targeted adaptation:

The state of the Gomoku board can be represented by a 15x15 matrix, where each position can be empty, black stone, or white stone. The AI needs to find the optimal move position in this huge state space. Unlike path planning, Gomoku search does not look for a path from the start to the end, but evaluates the value of each possible move position.

The heuristic function is the core of the A* algorithm. In this project, the heuristic function needs to evaluate the "threat value" and "opportunity value" of each empty position on the board:

• Offensive Score: Evaluate the possibility of forming connected stones, live three, live four, etc., for one's own side after placing a stone at that position
• Defensive Score: Evaluate the necessity of preventing the opponent from forming threats
• Position Weight: The strategic value of the central area of the board is usually higher than the edges

This design allows the AI to both actively attack and defend in a timely manner, showing strategic thinking close to that of human players.

Due to the extremely large state space of Gomoku, a complete A* search needs to be combined with pruning strategies:

• Local Search: Only consider empty positions around existing stones instead of traversing the entire board
• Depth Limitation: Limit the search depth to make decisions within a reasonable time
• Heuristic Pruning: Skip branches that are obviously inferior to the current optimal solution

The project adopts a modular design with a clear code structure, making it easy to learn and extend: