Little AI boxers trained through reinforcement learning in Burn, competing live in your browser with Rust and WebAssembly.
Rust Boxing is a browser-based boxing simulation where AI agents trained with Deep Q-Networks (DQN) compete against each other in real-time. This project was built to explore Rust's capabilities in both machine learning and web development, combining reinforcement learning with high-performance WebAssembly.
The project features 78 trained models, with 8 carefully selected fighters available in the browser demo. Users can spectate matches between different AI opponents, each with unique fighting styles learned through thousands of training episodes.
- Rust - Core game engine, physics simulation, and training infrastructure
- Burn - Deep learning framework for implementing and training DQN models
- WebAssembly - Compiles Rust to run at near-native speeds in the browser
- React + TypeScript + Tailwind - Frontend UI for fighter selection and match visualization
- Vite - Fast development and optimized production builds