SALT LAKE CITY — Francesco Biondi was in the driver’s seat of a partially self-driving car not long ago, talking to his passengers and enjoying the ride while the autopilot did the work. Then the car stopped behind a long line of cars inching their way through a four-way stop at an intersection.
“As long as the car in front would stop and restart, the system worked well, up until the point I was the first car in the line,” he said.
That’s when Biondi’s car simply darted into the intersection.
Luckily, other drivers were attentive enough to stay out of the way.
“They honked at me, but nobody actually ran into me,” he said.
Biondi, an assistant professor of psychology at the University of Utah, had unwittingly demonstrated something his own research has been trying to measure. Today’s partially self-driving cars are hazardous because they lull drivers into a state of complacency, when in reality those drivers need to be alert and ready to take control at a moment’s notice.
Biondi and other researchers at the U., with funding from the AAA Foundation for Traffic Safety, recently conducted tests in which volunteers drove back and forth to Wendover. They were told to drive one direction manually and to let the car handle the driving in the other. They wore contraptions that measured brain waves and had video recorders trained on them.
While the data still is being recorded, a summary of the project said “average heart rate decreased and heart rate variability increased when using autopilot,” which is consistent with mind-wandering and “transitioning into a state of relaxation.”
Cameras showed drivers yawning and struggling to keep their eyes open. The drivers had response times longer than those recorded by drivers in other studies involving distracted driving.
This ought to lend a tone of seriousness, and even urgency, to counteract those television commercials that show cars stopping suddenly and dramatically to avoid accidents despite onboard distractions. Often, modern technology can make up for human shortcomings, but the word “partially” in “partially self-driving” often gets short shrift.
Biondi said the results are consistent with decades-old research involving airplane pilots and autopilot systems. Pilots are used to handling the job of taxiing, takeoffs and landings manually, while letting the plane fly itself the rest of the way. Similarly, drivers of today’s semi-autonomous cars ought to handle urban driving themselves and leave the autonomous stuff for long stretches of boring highway, such as drives to Wendover.
But even then, they need to be alert to system failures or road hazards.
That message isn’t getting through.
“The consumer is in the middle, very confused,” Biondi said.
In a piece for wired.com, Jack Stewart wrote recently that people are getting the wrong message about where we are in the race toward fully autonomous vehicles. “In one example after another, it's clear too many people don’t get, or ignore the limitations of these robocars-in-training,” he wrote. “They zone out, look away, even fall asleep. And they cause crashes.”
He referred to two recent accidents — one involving a Tesla Model S in Culver City, California, that smashed into a parked fire truck that was responding to an accident on the freeway, and the other involving an alleged drunken driver who was found passed out behind the wheel of a stopped Tesla on the Bay Bridge in San Francisco.
Both drivers said they assumed the car was on autopilot and could drive itself. In the case of the alleged drunken driver, the car had a feature that gradually stops the vehicle if the driver ignores prompts to touch the steering wheel.
Biondi hopes the research he and his colleagues do will help manufacturers and consumers alike, perhaps even resulting in onboard features that do a better job of keeping drivers mentally engaged. He wants to do more research involving longer trips and people of all ages.7 comments on this story
He said the U.’s research is unique in that it isn’t funded by any interests with ties to specific automakers, it measures brain waves and it uses real drivers on real roads. Its data may prove helpful as the world plods along in this awkward interim stage before the age of fully automatic cars.
Biondi, at age 30, believes he won’t live to see that age, by the way, and not because he expects to die in a malfunctioning robot car. The obstacles that remain are great, he said.
All the more reason for us to find ways to stay alert.