Human beings can be pretty bad drivers. We get behind the wheel when we're too tired. We're distracted by cell phones and other gadgets. And sometimes, we try things we're just not skilled enough to attempt successfully.
Fortunately, the augmented cognition research team at Sandia National Laboratories is developing an automotive system that can analyze human behavior and help prevent such dangers. Better yet, they don’t need to reinvent the wheel to do it.
"We utilized data that already existed on the car's computer to collect a wide range of physical data such as brake pedal force, acceleration, steering wheel angle, and turn signaling," says Kevin Dixon, principal investigator. "And specialized sensors including a pressure-sensitive chair and an ultrasonic six-degree-of-freedom head-tracking system measured driver posture."
The researchers combined this information with data gleaned from five drivers wearing caps connected to electroencephalogram electrodes that measured the brain's electrical activity. After these subjects drove for several hours in unstructured conditions, the researchers input all of the data into Sandia software known as classifiers that categorize driving behavior.
Based on the data alone, the classifiers could detect particular situations, such as when drivers approach a slow-moving vehicle or change lanes to pass another vehicle. They also could detect the difficulty and stress of the task the driver is attempting. Then, the system tries to modify the tasks or the environment to lower the stress and improve the performance parameters.
"The beauty of this is that we aren't doing anything new or different to the car," says Dixon. "All the software that can make the determination of 'dangerous' or 'safe' driving situations would all be placed in the computer that already exists in the car. It's almost like there is another human in the car."
Based on this system, future cars may be able to deduce when drivers are becoming tired or hold cell-phone calls to prevent distractions during difficult driving tasks. The project began five years ago with funding from the Defense Advanced Research Projects Agency.
"Every year, tens of thousands of people die in automobile crashes, many caused by driver distraction," Dixon says. "If our algorithms can identify dangerous situations before they happen and alert drivers to them, we will help save lives."
Sandia National Laboratries