A new algorithm developed at the Massachusetts Institute of Technology can accurately predict how likely a person is to run a red light.
Traffic data from a busy intersection in Virginia showed that the system was accurate 85 percent of the time, roughly 15 to 20 percent more accurate than current methods. The algorithm takes into account a vehicle’s speed, deceleration and proximity to a stop light. The system is able to detect a potential light runner about two seconds in advance and could even send out a warning message to nearby vehicles about to enter the intersection.
The latter, of course, would depend on deployment of vehicle-to-vehicle (V2V) wireless communications technology that would allow automobiles to “talk” to each other. Ford, GM and BMW are all actively researching methods to integrate such technology into future cars and even make aftermarket additions available for older cars. An aftermarket solution could also be used for cyclists and pedestrians.
As MIT processor Jonathan How suggests, the system as a whole would have to be reliable enough not to produce too many false positives; that is, alerting drivers when a threat isn’t present. Doing so too often would likely prompt motorists to dismiss the feature as annoying and ultimately disable it.
MIT is also looking into the possibility of bringing the technology to aviation where it could aid in collision detection or prioritizing runways.
A 2008 study from the National Highway Traffic Safety Administration reported that 2.3 million automobile crashes occurred at intersections across the country, resulting in nearly 7,000 deaths. 700 of those were a result of someone running a red light.
