Sensors

LIDAR noise σ: 0.25 m Camera depth noise σ: 1.40 m

Scene

Obstacles: 7 Process noise (accel σ): 1.2 m/s²

Live tracking stats

Confirmed tracks0

Tentative tracks0

Tracks deleted (total)0

Fused tracking error—

Raw LIDAR error—

Raw camera error—

Sim step—

Raw LIDAR hit Raw camera hit Fused Kalman track

Each obstacle is tracked by its own Kalman filter: a predict step advances position/velocity with process noise, then LIDAR (precise, isotropic) and camera (precise in bearing, noisy in depth) measurements are fused according to their real covariance. Translucent dots are raw noisy hits; the solid orange boxes are the fused, smoothed tracks. Toggle a sensor off to see tracks coast on prediction alone, and drag to orbit the scene.