· 1 min read
ACORN: Teaching AI to Drive in Assetto Corsa
Building an improved AI driving model for Assetto Corsa that actually races cleanly. A look at the technical approach and early results.
The Default AI Problem
Assetto Corsa’s default AI is… functional. But it has obvious flaws:
- Unrealistic braking points
- Inconsistent racing lines
- Robotic overtaking
ACORN (AI Car Optimal Racing Network) is my attempt to fix this using machine learning.
Approach
Instead of hand-tuning AI parameters, ACORN:
- Collects telemetry from fast human laps
- Trains a model to predict optimal inputs (throttle, brake, steer)
- Deploys the model as a plugin that overrides default AI behavior
Early Results
The model already shows:
- Smoother corner entry/exit
- More consistent lap times
- Cleaner battles (fewer phantom collisions)
Still a lot of work to do, but the foundation is solid. Next up: multi-car race strategy and weather adaptation.
Why This Matters
Better AI means better practice for real drivers, more engaging single-player races, and ultimately a more complete sim racing experience.