ASU engineer is developing artificial intelligence that will make driving safer — a system that will augment autonomous cars
In 2015, Ashraf Gaffar’s daughter was walking across the street at a stop sign when she was hit by a driver who was texting.
“He hit her pretty badly,” said the Arizona State University engineering professor (pictured above). “She stayed in the hospital for a long time, and she had several surgeries to her face. It made me wonder.”
At that point, Gaffar had been working on developing an artificial intelligence car brain for three years.
“We better do it now,” he said. “We don’t want to wait until autonomous vehicles are all over the street in 10, 20 years. We want to start helping our kids on the street tomorrow.”
Gaffar is developing an artificial intelligence system for cars, which will augment autonomous cars (when they arrive) or act as a virtual mom sitting beside you as you drive.
“A vehicle that will help you drive better, drive safer and avoid accidents to a certain degree” is how the assistant professor in the SchoolThe School of Computing, Informatics and Decision Systems Engineering is part of ASU's Ira A. Fulton Schools of Engineering. of Computing, Informatics and Decision Systems Engineering described his system. “It will monitor your way of driving, and it will intervene as needed in an intelligent way.”
Video by Ken Fagan/ASU Now
Gaffar’s system will learn how you drive, then learn on its own.
“It will come pre-loaded with what is good driving and what is not,” he said. “Then the first few weeks it will learn your style of driving and adjust to it. Then it will be able to detect any anomalies and analyze them.”
Over time it will be able to help. You’re still going to be driving the car, but you’ll get some assistance when you need it. If you start showing fatigue or suddenly are unable to drive because of, for example, a heart attack, the car will intervene.
Intervention can range from a warning to pay attention to the car taking over and pulling itself over to the side of the road.
“You can see it as having your mom next to you,” Gaffar said.
Don’t mistake him as a hater of autonomous vehicles (“They really are the future,” he said), but they come with challenges. It’s no coincidence they’re being tested in sunny southern Arizona where it rarely rains and not in Canadian snow or New England fog.
Or amongst a combination of hazards — what Gaffar calls “cumulative adverse conditions” — say, a night where it’s raining on a busy road with no median where headlights are coming right into your eyes.
“When we look at conditions like these, we realize we are not getting rid of the driver anytime soon,” he said.
Dealing with everything found on roads from sea to shining sea means excruciating algorithmic programming that accounts for everything. If this happens, do this. If that happens, do that. If there’s a squirrel, a log, an alligator, a soccer ball in the road, do that. That’s why Gaffar favors artificial intelligence over software.
“There are areas where we have to recognize all of them in advance and put them in a very long and complex software, or we use artificial intelligence,” he said. “It’s easy to program. It simulates human intelligence and, more importantly, it has learning ability. When we drive, we get experience over time. Our intelligent co-driver works the same way. ... This is one of the areas where we really believe an intelligent car is not a replacement of an autonomous vehicle, but they can co-exist.”
And Gaffar’s daughter now?
"While the entire family suffered, and everyone was strongly affected by the accident, she did not just survive; she excelled beyond herself, despite extensive hospitalization, therapy and lost time,” he said. “She is now a successful second-year medical student in one of the best universities in North America. It wouldn't be too hard to guess what she wants to specialize in."
Top photo: Assistant Professor Ashraf Gaffar takes a spin in an automotive simulator on the Polytechnic campus on Sept. 27. He and his doctoral student are writing artificial-intelligence programming for cars that will learn the driver's driving style. A few weeks later, the programming will enable the car to be safer by detecting if the driver is fatigued or distracted. Photo by Charlie Leight/ASU Now