A crash caused by a driver running a red light caught by a camera in Worth, IL.
It’s happened to just about everyone. The light turns green, you step on the throttle and just as you pull into the intersection a car races across your path. If you’re lucky, you hit the brakes in time, mutter a curse under your breath and then drive on. But many motorists aren’t so lucky, with red light runners causing tens of thousands of accidents annually – many resulting in serious injuries or fatalities.
Efforts to crack down have had only moderate impact, even with the increasing use of automatic traffic enforcement cameras.
But researchers at MIT think they may have a better solution: a newly-developed algorithm they suggest is highly effective at predicting when a motorist will run a red light.
While the system would only be able to provide a second or two’s warning, researchers believe that could be enough to prevent an accident, perhaps even save a life.
“If you had some type of heads-up display for the driver,” explained MIT Aeronautics and Astronautics Prof. Jonathan How, “Even though your light might be green, it may recommend you not go, because there are people behaving badly that you may not be aware of.”
The software was developed by How’s former student, Georges Aoude, after using roadway sensors at a busy intersection in Christianburg, Virginia. The software looks for the rate of deceleration as vehicles approach a red light, among other things. The MIT team claims they have achieved an accuracy rate of 85%.
Flashing a warning might be one effective use of the technology but it might also be integrated into technology known as vehicle-to-vehicle communications. The goal is to not only link vehicles to a roadway infrastructure but have one car talk to another. That could provide advanced warning of a hidden obstacle, a sudden slowdown in traffic – or someone running a red light.
The MIT researchers aren’t the only ones who’ve taken a shot at trying to predict when someone would run a red light. But existing software has had a tendency to be a bit too cautious, issuing to many “false positives,” according to a report from MIT.
“The challenge is, you don’t want to be overly pessimistic,” How said. “If you’re too pessimistic, you start reporting there’s a problem when there really isn’t, and then very rapidly, the human’s going to push a button that turns this thing off.”
The next step is to develop what the researchers call a “closed loop system,” which would actually recommend the best course of action for a driver.
The team is also looking at possible uses for the new algorithm in air traffic control to see if it can predict the behavior of an aircraft to help anticipate possibly dangerous situations, especially in the crowded space around busy airports.