The ego car's own LIDAR/camera rays are raycast every frame toward the pedestrian. A parked delivery truck at the corner geometrically blocks every ray until the ego reaches an angle with a clear line of sight — real occlusion, not a scripted delay. A connected vehicle parked past the truck already has a clear view of the crossing spot. With V2X on, the instant that remote vehicle detects the pedestrian it broadcasts the hazard's position; if the ego is inside the broadcast range it "knows" about the hazard immediately, long before its own sensors could ever see around the truck.
d_stop = v·t_reaction + v²/(2·a_brake)
knownAt = min(ownSensorTime, v2xEnabled ? v2xTime : ∞)
margin = d_ego(knownAt→stop) vs d_ego(knownAt→crossing)
- Ego speed — higher speed both shortens the time to the crossing and lengthens the braking distance needed once the hazard is known, so it takes a bigger head start to stay safe.
- Occluder size — a bigger truck hides the pedestrian longer, delaying the ego's own-sensor detection and making V2X's early warning matter more.
- V2X range — the broadcast only reaches the ego car if it is inside this radius when the remote vehicle reports the hazard; too short a range and V2X can't help in time.
- V2X on/off — toggling it isolates exactly how many meters of stopping margin come from communication versus onboard perception alone.
Real-world relevance: this is the core argument for Vehicle-to-Everything (V2X) communication in autonomous driving — sight lines are physically limited by geometry, but radio isn't, so a connected vehicle or roadside unit can extend perception around corners that no onboard sensor can see through.