Algorithm Accurately Predicts Which Cars Will Run Red Lights
Researchers at MIT have devised an algorithm that predicts when an oncoming car is likely to run a red light.
The team developed an algorithm based on parameters such as the vehicle’s deceleration and its distance from a light to determine which cars were potentially going to run a red light, and which were going to be “compliant.”
They tested the algorithm on data collected from an intersection in Virginia and found that it accurately identified potential violators within a couple of seconds of reaching a light.
The researchers said that it gave drivers at the intersection enough time to react to the threat if alerted.
The algorithm generated fewer false alarms than other tools to predict driving behavior, which could be an important asset in incorporating the algorithm into systems that help guide human drivers.
“If you had some type of heads-up display for the driver, it might be something where the algorithms are analyzing and saying, ℠We´re concerned,´” Jonathan How, the Richard Cockburn Maclaurin Professor of Aeronautics and Astronautics at MIT, said in a press release. “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.”
He said that in order to implement these warning systems, vehicles would need to be able to “talk” with each other, wirelessly sending and receiving information like a vehicle’s speed and position.
Researchers are developing algorithms to analyze vehicle data that would be broadcast through systems like vehicle-to-vehicle (V2V) communication. The U.S. Department of Transportation (DOT) is already exploring V2V technology, as well as some car manufacturers.
Georges Aoude, a former student of How’s, designed an algorithm based on a technique that has been successfully applied in many artificial intelligence domains.
Aoude’s algorithm is able to capture a vehicle’s motion in multiple dimensions using a highly accurate and efficient classifier that can be executed in less than five milliseconds.
How and Aoude tested the algorithm using an extensive set of traffic data collected at an intersection in Virginia.
The team applied their algorithm to data from over 15,000 approaching cars at the intersection and found that it was able to correctly predict red-light violators 85 percent of the time.
The team was able to predict within a couple seconds of whether a car would run a red light. They even found “a sweet spot” when the algorithm has the highest accuracy and when a driver may still have enough time to react.
The MIT researchers found that its algorithm generated fewer false positives when compared to similar safety-prediction technologies.
“The challenge is, you don´t want to be overly pessimistic,” How said in a press release. “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 team is now investigating ways to design a closed-loop system, and are also planning to adapt the existing algorithm to air traffic control.
On the Net: