Fourth of a five-part series.
If you were expecting the future to be like the the flying cars of “The Jetsons,” it’s turning out more like the robot-driven Johnny Cab in “Total Recall.”
Cars that can drive themselves are here, though maybe without a talking android. Automakers have been working on the technology for decades.
In 1956, GM introduced the Firebird 2, a concept car that would follow a white line down the center of an express lane from Phoenix to Chicago.
Today, with adaptive cruise control, lane assist and collision avoidance, thousands of cars can pretty much drive themselves. But they never built those express lanes. Without them, how do you get the car to stay on the road without hitting anything?
“I’ve been obsessed with self-driving cars for a long time and I think deeply about what it would take to get a fully autonomous car on the streets of Boston," said John Leonard, a professor at MIT.
Leonard identifies several challenges to getting a car to work in its surroundings. He calls the first one “robustness of perception. A self-driving car has to classify things in the world. Is that a cyclist? Is that a pedestrian? Is that a car? And a self-driving car has to do with with a very high probability of detection and a very low probability of false alarms.”
The second challenge is creating highly detailed maps.
“Google uses its maps to say, ‘Oh, I expected that mailbox over there. I expected that tree. I expected that sign,'" Leonard said. "But this map information must be accurate.”
If it sounds like it would take an army of people to create maps of that detail, that’s exactly what’s happening. A company called Mobile Eye has hired 600 people to map objects gathered from millions of miles of test driving.
But even with the most detailed maps, the car still has to be able to see, including in bad weather. In Massachusetts, that means snow. Leonard cites one possible solution: “MIT Lincoln Labs has something using ground-penetrating radar to look through the snow."
Another way to guide vehicles through a storm is by traffic lights sending signals to cars. And there are others—at the University of Arizona, Professor Jonathan Sprinkle is working on the problem with research supported by the National Science Foundation.
“It might be traffic lights, it might be an indicator in your car that tells you what lane to use.," Sprinkle said. "It might be organizing the traffic into platoons.”
The buzzword behind traffic flow is connectivity; enabling vehicles to talk to each other—to create platoons—and also talk to traffic signals.
But Leonard says the biggest challenge of all “is interacting with people. For example, the police officer waving me through a red light. Interacting with humans. Robots are really not good at that. Interpreting nods, waves, winks is a very hard thing.”
And, of course, reacting to pedestrians, who are perhaps the most unpredictable element of all. As far as interacting with people inside the car, the problem is the hand-off: determining when humans should take control.
I learned something about the hand-off when I got behind the wheel of a friend’s Tesla with Autopilot on Route 128 last spring. My first question was, “What do you do with your hands?” He replied: “I do not text!”
My ride-along—I’m not sure you can call it driving—was four days before Tesla announced the first fatal accident involving one of its cars on Autopilot where the driver was similarly riding hands-free. I wouldn’t think of taking my hands off the wheel today.
The Tesla with Autopilot is not a fully autonomous car. But for the experimental vehicles that are, Sprinkle says the final challenge is getting people to trust the car, just like riding with someone who’s just gotten a learner’s permit.
“Many of the feelings that you have when you’re sitting in the passenger seat with a learner driver are the same," he said. "I wonder if they’re going to slow down when this turn comes up. I wonder if they’ll stop before we reach that vehicle that’s right in our path."
In fully autonomous test vehicles, Sprinkle says those feelings go away once the ride is completed safely.
As for trusting other devices from the future—like that crazy treadmill dog walker that drove George Jetson crazy every week—we still have a ways to go.
Robin Washington is a longtime Boston transportation journalist. He may be reached at robin@robinwashington.com or via Twitter @robinbirk.