Self-Driving car

A simple self-driving car using a neuronal network.

Inspiration

This project is inspired by the self-driving car project on https://www.radufromfinland.com/.

I watched the video about it on YouTube (freeCodeCamp.org, https://www.youtube.com/watch?v=Rs_rAxEsAvI&list=WL&index=1) and immediatelly wanted to create what I just saw on my own.

This video majorly improoved my understanding of neural networks and how the are used.

A lot of parts are implemented from sratch, like all the rendering code, but some of the core algorithms are C# ports of Radus JavaScript code.

Goals, plans and a disclaimer

This project is not intended to have any usefull output like a reusable library.

This is a 100% experimental project to learn neuronal networks and have fun with it.

Key features

Initial implementation done

• Entity graph
• Multilayer renderer
• Distance sensors
• Manual driving with keyboard
• Collisions
• WPF client
• Port original neuronal network

Planned features

• Spawning and despawning traffic
• Training mode
• Testing mode
• Experiment with different sensor arangements and scoring methods
• Alternative implementations using a library with ready to use neuronal networks
• Blazor web implementation
• Manual control with gamepad
• Nicer car rendering

Other ideas

• Fully automated training
• Curved road
• Dynamic traffic
• Different vehicle types (trucks, motor bikes, ...)

Documentation

[t.b.d.]