If you broadcast often enough, milliseconds stop mattering.
Pedestrian in Crosswalk Warning fuses signal-phase data with sensor-based pedestrian detection and broadcasts a Personal Safety Message (PSM) stream to nearby connected vehicles. CVs that predict a vehicle-pedestrian collision then apply a two-zone deceleration profile to maintain a safe distance.
An edge computing unit at the intersection reads SPaT data from the traffic signal controller and pedestrian presence/position from advanced sensors (video, LiDAR). The RSU broadcasts the combined picture — PSM + SPaT + MAP — to approaching connected vehicles, which evaluate collision risk in real time.
Aimsun pedestrian-simulator outputs (modeling video/LiDAR sensor output) are read every 0.1 seconds. Pedestrians are mapped to crosswalks; presence during a red phase or in a vehicle's path is flagged.
PSMs at 10 Hz carry pedestrian location/speed. SPaT at 10 Hz, MAP at 1 Hz. Default 1,640 ft range covers the entire intersection approach.
Each CV checks: at my current speed, will I reach the crosswalk before the pedestrian finishes crossing? If yes, collision risk is flagged.
Two distance thresholds: S_th1 = 3× braking distance, S_th2 = 6× braking distance. Zone 1 (inside S_th1): decelerate at 2× d_normal. Zone 2 (between S_th1 and S_th2): decelerate at d_normal. Beyond S_th2 the algorithm doesn't activate.
Built on top of the Red Light Violation Warning network, with pedestrian origins, destinations, and a pedestrian simulation area added. High pedestrian demand is forced to surface multiple collision opportunities.
When pedestrians are detected and broadcast often enough, communications latency and packet loss stop being meaningful failure modes.
In a representative scenario, the unwarned CV approached the crosswalk at ~25 mph at the 50 ft threshold — collision-likely. The warned CV started decelerating earlier and arrived at ~15 mph, with a 50 ft buffer maintained.
Across delays from 0 to 0.5 s, the proposed algorithm still maintains a safe buffer. Delay shifts the start of deceleration but doesn't eliminate the safety margin.
From 0% to 20% packet loss, performance barely changes. Because PSMs broadcast at 10 Hz over a long upstream distance, missing a message or two doesn't change the control action.
The robustness comes from detecting pedestrians frequently and broadcasting their presence at high frequency. Drop those rates and the algorithm becomes fragile to all the things it currently tolerates.
Maintaining a sufficiently long V2X communications range lets CVs receive alerts well upstream — buying the time to decelerate gracefully rather than emergency-brake.
Aimsun's collision-free model overrides the proposed algorithm when actual contact would occur, so quantitative crash-rate metrics aren't possible. Per-collision speed/distance traces are the evaluation tool.
This is the rare V2X application where redundancy on the broadcast side buys forgiveness everywhere else — frequent PSMs make the system tolerant to latency, packet loss, and processing delay all at once.
The clearest insight from the simulation is structural: high broadcast frequency is the lever that does the most work. By pumping PSMs at 10 Hz, the system creates so much message redundancy that the typical degradations — latency, packet loss, sensor processing delay — stop having measurable consequences.
Aimsun's collision-free behavior limits how aggressively safety can be quantified. The simulator stops vehicles at the last instant to prevent contact, which means real near-misses are masked. The team's response — examining individual potential-collision traces in detail rather than counting crashes — produces qualitative confidence but no statistical safety claim.
For deployment, that means investing in the sensor and broadcast stack pays off everywhere. Stretching pennies on detection frequency or upstream range hits the algorithm's only real failure modes; the communications layer is comparatively forgiving.
Run pedestrian detection and PSM broadcast at the highest sustainable frequency — 10 Hz is the simulated default and the source of the algorithm's robustness.
Provision a long upstream communications range. Earlier alerts mean gentler deceleration, which means a comfortable safety buffer rather than emergency braking.
Don't over-engineer the communications layer. With high-frequency broadcasts in place, the system tolerates 0.5 s of combined processing + communications delay and 20% packet loss without meaningful degradation.
Use the two-zone deceleration profile: S_th1 = 3× braking distance, S_th2 = 6× braking distance, with 2× / 1× normal deceleration multipliers.
Plan a safety-metrics study that escapes Aimsun's collision-free model. SUMO with built-in surrogate safety, or post-processing of trajectory data with a custom near-miss detector, would let you quantify the crash-rate improvement that this study can only show qualitatively.