Future Cars May Predict Driver Actions

Vehicles may one day be able to anticipate their drivers' actions, according to researchers at Cornell University.

The scientists says technology they are developing is so far about 77% accurate in predicting when drivers plan to turn, change lanes or continue forward about 3.5 seconds before they do so. They will report on their work to date during the 2015 Robotic Science and Systems conference in Rome on July 16.

The goal is to link such information to the vehicle's traffic sensors and GPS mapping capabilities. Doing so could enable the car to warn the driver about a dangerous maneuver turning the wrong way into a one-way street, for example before it happens.

The Cornell team compiled video footage of 10 drivers during 1,200 miles of travel. Face detection and tracking software correlated head movements with vehicle actions. External radar and cameras monitored conditions outside the vehicles. The team then developed an algorithm to calculate the probability that a driver will take a particular action.

There is still work to be done. The researchers note their technology can be confused by shadows and the driver's interaction with passengers.

The Cornell project is funded in part by the U.S. Army Research Office, the Office of Naval Research and the National Science Foundation.