The goal of this project is to create a simulation of a self-driving car by implementing car physics, sensors, collision detection, and neural networks among other things.
- Implemented a feed-forward neural network from scratch
- Added a chart to track training progress
- Tuned network and sensors to improve model accuracy
- Created a test framework of randomized car traffic
- Added a finish line to denote when training is complete
To see how the model is trained, open index.html in a browser. The training is complete when the car reaches the finish line (100% on the graph)
The "best brain" has already been saved in best-brain.js. Open test.html in a browser to test the trained car. Here, the traffic cars are randomly generated, meaning everytime you refresh, there is a different arrangement of cars. Feel free to refresh multiple times and see how the car reacts!
- This project is based on the self-driving AI course from freeCodeCamp, which can be found here: https://www.youtube.com/watch?v=Rs_rAxEsAvI&t=2s&ab_channel=freeCodeCamp.org. After completing the course, I've made my own improvements, including a test framework, model tuning, more car traffic, sensor tuning, and much more.