While there have not been any effective systems for autonomous driving put into production, driver assistance programs which help motorists control their vehicles have evolved a lot lately.
Researchers from Cornell and Stanford University have developed a system that uses cameras and a new computer algorithm which literallly anticipates what a driver is going to do seconds before that actually happens.
At the moment, there are cars equipped with safety systems that monitor the vehicles’ movements and state a warning if there is an unsafe turn or a lane change, but the warning comes after the driver has acted, which, unfortunately, is too late.
According to the researchers from the two universities, by observing the body language of a driver and taking into account the context of what is happening outside the vehicle, a computer algorithm can calculate the probability of a driver turning, changing lanes or going straight ahead.
Ashutosh Saxena, assistant professor of computer science at Cornell says that “There are many systems now that monitor what’s going on outside the car. Internal monitoring of the driver will be the next leap forward.”
Together with Ashesh Jain, a Cornell graduate student, Saxena has recorded video of ten drivers along with tape of the road ahead over a period of two months. A computer uses face detection and a tracking software identifies head movements associating them with turns and lane changes.
In a test against another data set of videos, the system predicted correctly 77.4% of the time, with an average of 3.53 seconds in advance, seconds which might turn to save lives. The system needs yet to be refined as 6% of the time the tracking gets confused by passing trees or lighting changes but also by the drivers’ interactions with other car passengers.
By Gabriela Florea