This project implements a genetic algorithm (GA) to evolve agents that can find their way through a maze. The agents start with random movements, and through natural selection, crossover, and mutation, they gradually improve their ability to reach the goal.
Key Features:
- Genetic Algorithm Core: Includes selection, crossover, and mutation mechanics.
- Maze Representation: The maze is represented as a 2D grid with walls, paths, and a goal.
- Real-time Visualization: Watch agents as they evolve in the maze with a simple ASCII representation.
- Elite Selection: Top-performing agents are selected as the basis for the next generation.
-
Initialization:
- Agents are created with a random genome (sequence of moves: up, down, left, right).
- Agents are placed at the start position in the maze.
-
Fitness Evaluation:
- Each agent’s fitness is determined by how close it gets to the goal. The closer, the better the fitness.
-
Evolution Process:
- Selection: The top-performing agents (elite agents) are selected to reproduce.
- Crossover: Two parents’ genomes are combined at a random point to create offspring.
- Mutation: Some offspring may experience small random changes to their genome to introduce variation.
-
Next Generation:
- The new generation replaces the old one, and the process repeats until agents successfully reach the goal or the set number of generations is reached.
The maze is represented as an 8x8 grid:
- #: Wall
- A: Agent
- G: Goal
- (Blank Space): Path
- Fix problem where generation stops after some time
- Improve fitness function to prevent local maximum
- Integrate OpenGL to better represent generations
