An assistant professor of computer science at Cornell University and his team have come up with a car safety system that warns a driver when he or she is about to make a mistake, according to a Cornell Chronicle article by Bill Steele. The warning could be a light, a sound, or a vibration, Steele writes. The system can identify when a driver is becoming drowsy or distracted, according to the company’s website, which could help prevent car accidents.
The professor, Ashutosh Saxena, told Steele that there are now many systems that monitor things outside a vehicle. This one — called Brain4Cars, as Ben Coxworth reports for Gizmag — features internal monitoring of the driver, which “will be the next leap forward,” Saxena said.
To develop Brain4Cars, Saxena and researchers from Cornell and Stanford recorded video of 10 drivers and the roads they were driving on over a two-month period. Each person drove 1,180 miles on highways and in urban traffic, Coxworth writes.
A computer using face detection and tracking software identified head movements and learned to associate them with turns and lane changes, so that the final system can anticipate possible actions the driver may take. The computer continuously reports its anticipations to the car’s central safety system.
In a test that compared data from those videos with data from videos of different drivers, Brain4Cars was correct 77.4% of the time in predicting a driver’s actions, Steele writes. The system was able to anticipate driver behavior an average 3.53 seconds in advance, Steele adds. Saxena said that those few extra seconds could save lives.
The system needs further refinement, as it still has some glitches. Steele writes that, for example, 6% of the time, the face tracking component was confused by such things as tree shadows and other aspects of lighting. In another example, Steele writes that the system can be misled by drivers interacting with passengers.
Cornell graduate student Ashesh Jain, part of the research team, told Steele that automakers will need to incorporate Brains4Cars into a complete safety system. Among the improvements it could benefit from are infrared camera for nighttime observation; 3-D cameras for more accuracy; tactile sensors that monitor pressure on the steering wheel; and cameras or pressure sensors that can observe the driver’s feet, perhaps to know when they will hit the brakes or when the driver is looking at a phone, Steele writes. Brains4Cars could also eventually link directly to wearable technologies, Jain told Steele.
Saxon and Jain will give a talk about Brains4Cars in a workshop on “Model Learning for Human-Robot Communication” at the 2015 Robotic Science and Systems conference July 16 in Rome, Steele reports. The other members of the research team include graduate student Hema S. Koppula and Stanford University graduate students Bharad Raghavan and Shane Soh.
In a related article, this blog wrote last month about the Impecca Alert Band, which can detect ahead of time when a driver is getting too drowsy to drive.