Cornell and Stanford Researchers Develop Technology That Can Predict Dangerous Driving Behavior

Many comprehensive studies in the past have shown that the vast majority of traffic accidents involve some sort of human error, which means that preventing dangerous driving behaviors can help dramatically improve road safety. That’s exactly what a group of researchers from Cornell and Stanford Universities have attempted to do by trying to develop a technology that can predict human errors and help prevent car accidents. Ashutosh Saxena, assistant professor of computer science at Cornell, with the help of his colleagues, has developed a vehicle safety system that monitors a driver’s movements, so that it can determine whether the driver is about to make a potentially dangerous maneuver, which could lead to a collision with another vehicle or a pedestrian.

In a report that was recently published on Cornell University’s website, researchers say that they have created a system that relies on sensors that are mounted on the car to scan the car’s surroundings and cameras inside the vehicle to track the driver’s movements, allowing it to detect an imminent collision. With its cameras constantly monitoring the driver and looking for specific movements that indicate that the driver is about to make a sharp turn, abruptly change lanes or stop suddenly, combined with the information coming from the sensors outside the car that are monitoring lane markings, as well as the movement of other vehicles, the system’s computer algorithm can detect a potential accident before it happens.

“There are many systems now that monitor what’s going on outside the car,” said Ashutosh Saxena. “Internal monitoring of the driver will be the next leap forward.”

If the system senses that the driver is about to make a move that would result in a crash, it warns the driver through various alerts, which could be visual, audible, or tactile, such as a light, a beep or a vibration, so that the driver can take the appropriate action that is necessary to avoid a collision in a timely manner. “If there’s danger on the left, the left side of the steering wheel or the seat could vibrate,” said graduate student Ashesh Jain, who participated in the development of this system.

Over a period of two months, researchers analyzed videos of 10 drivers covering a total of 1,200 miles of city and highway driving, capturing their body language, as well as footage of the area in front of the car that was recorded while they were driving. Then, they tested the system with different drivers, with the technology predicting what maneuver the drivers are about to take in 77% of the time.

As they continue their research, Saxena and his associates plan to add more advanced cameras, as well as eye-tracking capability to the system, which would help read a driver’s body language and interpret their movements more accurately. The researchers say that this system is still at an early development stage, and that in order for it to become fully functional and commercially available in the near future, automakers will have to join the research efforts.