Code for Transportation 1st Place
Updated: Sep 9
Vihaan Joshi | USA
Many wheelchair users, people with walkers, and the elderly cross at half the speed of normal walkers. The traffic light has no idea they are there. The lights countdown to hand the road back to the traffic. Yet, according to studies, pedestrians over the age of 65 are twice as likely to be killed per crossing as those who are younger. The pedestrian button also does not help the traffic light as it only indicates that somebody is over there waiting, not for what purpose. CrossWise puts a single camera on the signal pole to gather information about the crosswalk. The system deploys the YOLO object detection model with tracking to monitor the crossing. Based on the detection of individuals crossing the road, the system makes three decisions that a traffic light timer cannot make on its own. First, it makes a touchless crossing call that detects somebody is waiting to cross the road without having to detect whether they are waiting for a bus or not. Second, it sizes each interval of the crosswalk according to the slowest person who is actually on the crosswalk. Finally, it ensures that the clearance interval for pedestrians will not be ended while there is somebody still on the crossing road. This system also detects if the approaching cars cannot physically stop in time for the pedestrians using the deceleration formula (v²/2d). I have tested this in SUMO, a traffic simulator that traffic engineers use to test traffic lights against the standard traffic light in the same area with the same data. Across four seeds and eighty simulated minutes of traffic, the traditional signal left 73 pedestrians still in the road when the traffic was released. CrossWise, however, left zero in every run. Furthermore, the average waiting time for the pedestrians declined from 45.4 seconds to 34.5 seconds. However, the delay for the vehicles increased from 5.1 seconds to 13.8 seconds. This perception system was separately tested on real video footage of a real crosswalk. It requires one camera and one small computer to sit in the traffic signal cabinet. The system does not store any video footage of the crosswalk. It can only classify the motion and the shape of any objects moving across the crosswalk. If the perception system should fail, the signal will fall back to its original timing. This Contributes to SDG 11(sustainable cities, Target 11.2 "accessible transport for persons with disabilities and older persons", SDG 3 "road safety", Target 3.6, and SDG 10(reduced inequalities, Target 10.2). Crossing wise adapts to the person, instead of asking the person to keep up.
Comments