Top-down view of the same perception-planning loop as the 3D version. Each frame the car casts LIDAR/radar/camera rays against every obstacle's bounding circle (real distance tests). Detected obstacles feed a repulsive potential field; an attractive term pulls the car back toward the lane centerline.
F_total = k_att * (x_goal - x_car)
- Σ k_rep * (1/d_i - 1/d0) * (1/d_i²) * n̂_i for d_i < d0
steer = clamp(atan2(F_total.x, F_total.y), -δmax, δmax)
d_i is distance to obstacle i, d0 is the sensor range (influence radius), n̂_i the unit vector away from the obstacle, δmax the steering cap. Turning off a sensor removes its rays from the fusion set — fewer detections, worse avoidance.
- Vehicle speed: forward velocity; higher speed shortens the reaction window.
- Sensor range: the LIDAR/radar influence radius d0 in the repulsive field.
- Traffic density: number of obstacle vehicles on the road.
- LIDAR / Radar / Camera: toggle each modality to see fused detections and avoidance quality drop.
- V2X: when on, obstacles are seen ahead of line-of-sight, extending effective range 1.6×.