Can AI Technology Help Solve the Distracted Driving Epidemic?

In the United States, distracted driving is an epidemic that needs to be solved. According to the U.S Department of Transportation, approximately 9 people are killed each day, with 1,000 more being injured, from a car crash that is the result of a motorist driving distracted. While many will attribute this to society’s increased use of technology, one researcher believes it is actually technology that can help solve the outbreak.

Thanks to the new groundbreaking software developed by a team of scientists at the University of Waterloo, distracted driving is about to become much harder to ignore. Spearheaded by Dr. Fakhri Karray, the group has created a new artificial intelligence (AI) software that detects when vehicle operators drive distracted.

The stand alone system, according to Dr. Karray, is “able to intelligently tell that the person is texting, drinking water, eating, reading a newspaper or making a phone call. The system will alert [the driver] that [he or she] is doing a dangerous type of behavior, one that is not allowed while driving.”

But how exactly can one small gadget accurately determine what the person behind the wheel is doing? According to Dr. Karray, the new AI software uses a database of pictures to detect when the driver is engaging in dangerous activities while behind the wheel. The team recorded individuals completing different tasks ranging from drinking water to texting, talking on a cell phone to driving normally. These video clips were then fed into a system that segmented them into photographs. Once broken down into photographs, Dr. Karray explained that he was able to load them into the AI brain, which is tasked with assigning each behavior under various classes. Each class is labeled to distinguish the safe activities from the potentially dangerous ones.

While mounted on the dashboard, the system continuously observes the driver’s actions and compares them to what is already stored into the program (it’s memory of past experiences). If an action recorded is flagged as being in an unsafe class, the driver will get an alert informing them that they are participating in a dangerous activity.

However, as Dr. Karray explained, this is only the very beginning of what the technology is able to do. The system could also be integrated fully within a suite of sensors designed by car manufactures to operate next generation of fully autonomous vehicles. It is up to the automotive designers, insurance companies or the federal government safety board to decide how far they want the technology to go. “These people are going to make the decisions, it’s not up to me–I just give them options,” Karray remarked. “So possibly they would like an option where the system alerts the parent, the car close by, or even the police when someone is driving dangerously or is continuously engaged in the distracting behavior. The system [could] be connected or tethered to the cloud so any time it observes that you are texting, it sends a faceless picture of the person and says ‘we are going to ticket you, this is forbidden.’ Or, the system could simply alert the driver to tell them don’t do this.” The system developed here is part of a comprehensive research work dealing with the assessment of the driver status while driving. In the future, this system could allow the car to take full control when the driver is fully distracted or on the verge of causing a car crash.

Dr. Karray and his team are currently preparing to auction this technology, which, if sold to a commercial entity willing to work with the scientists, could then make the systems available to the masses within the next 12 months. “We have tested the system and its working, not perfectly, but it’s working to a really high level of precision,” he said. While this would just be the separate technology to mount in your vehicle, Dr. Karray noted that in two to three years, with cooperation from automobile manufacturers, this technology has the ability to become a standard feature for next generation driverless cars.