← Back to projects

AI / Reinforcement Learning

Reinforcement Learning Cars Learning to Drive

A-Level OCR Computer Science programming project built in Unity

Built as an A-Level OCR Computer Science programming project, this Unity simulation trained cars to navigate around a track using custom-built feed-forward neural networks and reinforcement learning rather than third-party ML libraries. The project combined car physics, sensor-style inputs, reward feedback, model loading and saving, training controls and visual feedback. It was accompanied by an approximately 50,000-word project report explaining the system design, development process and evaluation.

OCR A-Level CS

Course

Unity

Engine

C#

Language

50k words

Writeup

Key details

  • Built the car physics, driving simulation and track-navigation environment in Unity.
  • Implemented feed-forward neural networks and reinforcement learning logic manually in C# using default Unity libraries.
  • Added model loading and saving so trained networks could be reused between sessions.
  • Built training controls and visual feedback for inspecting whether cars were learning valid track behaviour.
  • Produced a substantial written A-Level project report explaining the system and development process.