Independent game guide

F1 Neuroévolution

Choose a car and circuit, then watch generations of neural-network drivers evolve from sensor readings while you tune speed, rays, obstacles, and training stages.

Loads the game from the creator's official host

Play on itch.io ↗

Playable here on GameWren.The game is served by the creator’s official host inside this page; GameWren does not rehost its files.

F1 Neuroévolution

Put a neural driver on the circuit

Open the car showroom and select at least one lineage before starting a circuit run. Eight Formula 1-inspired tracks are joined by two extreme layouts and a fifty-level training curriculum that can move the model through progressively different shapes.

The tested Tiyoti Supraa GT began training on Monza at 10× speed. The live panel showed generation zero, one of sixty cars still on track, 22.9 percent progress, 4.3 seconds of survival, and a speed of 190 km/h.

Watch the inputs behind each turn

Five distance rays and the car's speed feed a 6→20→2 neural network. The lineage panel exposes the current ray distances, speed, progress, survival, confidence, and circuit record so the behavior is more than a moving dot on a track.

Change the ray angles and length or open the network editor to experiment with the model. Structural changes erase previous generations, while simple simulation-speed and display changes leave the learning process intact.

Train faster or make the course harder

Speed controls run the simulation from real time through 100× or maximum. Performance presets can draw fewer cars, hide the course, or disable decorative shaders when rapid generations matter more than watching every attempt.

Obstacle tools add random objects, remove them, or open a workshop for preset and hand-drawn shapes. Successful models can be saved, loaded, reset, and compared through per-circuit confidence views.

Guide

Select: Open the car showroom, select at least one lineage, and choose a circuit before training
Train: Start a circuit run or the fifty-stage deployment curriculum and compare progress across generations
Accelerate: Raise simulation speed from 1× through 100× or maximum, and reduce rendering detail when calculation matters more than animation
Tune: Change sensor-ray angles and length or open the neural-network editor; structural changes reset saved generations
Challenge: Place preset, random, or hand-drawn obstacles, then save successful models and review confidence by circuit

Creator's description

Selecting the Tiyoti Supraa GT enabled training on Monza. At 10× speed, the first generation reported one of sixty cars still on track, 22.9 percent progress, 4.3 seconds of survival, 190 km/h, and five live sensor distances feeding the 6→20→2 network.