Wouldn’t it be handy to know when another driver planned to run a red light? Don’t count on having all the answers any time soon, but students and scientists at the Massachusetts Institute of Technology are working to develop a system to prevent crashes caused by drivers running red lights and stop signs.
The system hinges on two things. The first is a complex algorithm that can predict when a driver is about to run a red light or stop sign. The other is a vehicle-to-vehicle (V2V) communication system that warns drivers with the right of way to stop to avoid being struck.
MIT says that the algorithm can quickly and accurately predict which cars are “compliant,” which means it will stop at the appropriate signal or sign, and which are “violators,” or those that won’t properly stop and could cause an accident. It was developed using real traffic data from an intersection in Christianburg, Virginia, by collecting a massive set of data about the cars that crossed the streets both with and against the traffic light, as well as the timing of the commanding traffic lights. All told, the dataset included more than 15,000 vehicles.
The resulting algorithm has a success rate of 85 percent–that is, it can correctly identify a violator 85 percent of the time. By finding the sweet spot of data collection, scientists also discovered the optimal place in the intersection to make the decision of violator vs. compliant driver, which means the algorithm is exceptionally fast–it takes just five milliseconds to make a decision.
For this algorithm to actually do its job, you’d need an intersection kitted out with the proper sensors and computer hardware, but also a way of communicating this to drivers. Enter V2V, the concept of vehicle-to-vehicle communication. As we’ve reported here before, there are plenty of manufacturers that are toying with the idea of letting cars communicate information about hazards, tailbacks, severe weather and other issues with each other, so this could be an addition to that platform of information. If V2V could access information about non-compliant drivers quickly, it’s possible that it could reduce side-impact crashes. We say “quickly” and mean it, too: the time between an algorithm making the decision and in-vehicle alert would have to be below two seconds, according to MIT.
Now, the technology is the typical few years away from any real product integration, but with the National Highway Traffic Safety Administration claiming that 2.3 million crashes happen at intersections every year, it might not be a bad idea to work on making that number a lot smaller.