As the name suggests, this project trains and implements an AI model that can drive the car on a track without collission. Firstly, the model is trained using NEAT-python. After the training, the model can drive the car by itself.
- Python
- Pygame
- NEAT (NeuroEvolution of Augumenting Topologies)
The project is inspired from Code Bullet's Youtube video on car simulation. He used Q-learning to train the AI. I made use of NEAT to train the AI model.
The first challenge was making the car move on a still background. The rotational dynamics of the car were difficult to implement. The second challenege was implementing a good fitness function. It took the most time and required a lot of research and trials.
I will work on a manual mode where the user can race against the AI.
Watch this video to see the model being trained and then being implemented to drive the car:- https://youtu.be/LbtvLXYB938