Self-Driving Simulator
Browser FSD lab. PolicyNet learns by trial-and-error
(dense rewards: distance, crashes, lights, keep-right). Classical Fusion only
emergency-brakes. No expert clone.
Fusion (recommended) = cameras + LiDAR features → RL policy.
- Live camera feeds with detections, free-space, walls/curbs
- Emergency stop on red / near wall — net still steers otherwise
- Fog/night — vision suffers, LiDAR free-path holds up
Grid city — intersections, roundabouts, freeway, suburb/downtown. Traffic slider 0→rush. Score += distance + law (keep-right, headway, lights, speed zone).
R / Next gen = new episode (keeps neural weights). Wipe net = erase brain (v13).
Manual: WASD — A=left D=right (S = reverse) · Pause P · Modes 1–6
Learning
Pure RL: reward for distance / survival; punish walls, red-light runs, off-road. Explore (ε) decays as score rises.
Orbit: drag in the main view (left or right) · scroll zooms.
Click camera tiles to enlarge.