Project Overview
This is a simple AI car game that simulates the evolution of self-driving 2D cars using the NEAT technique. I implemented the simulation in Python. Initially, I trained the cars on basic maps designed with tools like Paint to provide easy learning challenges. As the training progressed, I gradually introduced more challenging maps. The cars in the simulation have five input sensors and four output neurons representing actions like steering left, steering right, increasing speed, and decreasing speed.
To evaluate the performance of the cars, I employed a fitness metric that rewards or penalizes based on their performance. After each generation, the cars evolve, with the fittest ones surviving and reproducing. Cars that fail to perform well are removed from the population. This simulation creates an environment where the best cars survive and reproduce, leading to continuous performance improvement over time.
The main purpose of this project is to demonstrate how neural networks can be trained by showcasing this game. It is designed for enthusiasts or beginners who are interested in understanding the training process of neural networks through this game.
Feel free to try out the game using the provided model link! Happy learning!