“Anticipatory” Vehicles

In the U.S., ninety percent of accidents are caused by driver error. With more cars on the road and more distractions than ever for drivers it’s no wonder that auto manufacturers are looking for ways to enhance driving safety through automatic steering and braking systems. Now a project known as Brain4Cars is exploring ways to make these systems even more intelligent by predicting what a driver is about to do next.

An experimental dashboard computer has been designed which uses deep machine learning to recognize the actions, body language and behavior that precede a particular driving maneuver. For example, a lane change may be preceded by a glance over the shoulder and a check of the rearview mirror as well as changes in speed, steering and acceleration. The system would combine this information with the car’s built-in sensors and cameras to alert a driver when someone has pulled into that lane, or even prevent them from performing the maneuver. On the other hand, if the system senses that a driver is distracted, but there is no imminent threat ahead of them, it may suspend unnecessary alerts. The algorithms were developed from data collected on ten different individuals driving nearly 1,200 miles and were found to be over 90 percent accurate in predicting a driver’s intention to change lanes. The researchers intend to make the data collected freely available to auto researchers and academics.

Some may think that monitoring drivers will become less important with the dawn of autonomous vehicles. However, the opposite is actually true. As drivers are allowed to increasingly divert their attention while their vehicles do the driving, it will be more important than ever to be able to assess their behavior during those times when they do need to retake the wheel, since it can take several seconds for a person to fully engage, depending on their level of distraction.