Aarian Marshall is a transportation reporter at WIRED and former CityLab contributor. She lives in San Francisco.
And it might be coming to an autonomous vehicle near you.
The GPS in your smartphone is okay. Yeah, I said it: just okay. It’s crummy while inside a building, and because of errors that pile up between the satellite that sends out its signal and the equipment that receives it back on Earth, it isn’t perfect outside, either. The errors mean that your standard GPS equipment is generally only accurate within 10 meters. That’s why your phone may take a while to figure out whether you’re walking, biking, or driving in the right direction, and why it sometimes says you’ve actually been sitting in the house next door this whole time.
But GPS can be much more accurate than that. In fact, it has been for at least two decades. Surveyors have used GPS to measure down to the centimeter for some time now. More precise GPS identifies signals at a higher frequency, called the carrier phase, but must use complex bits of mathematics to figure out locations. The problem is that this process is slow, and speeding it up to make it widely available to the public is expensive.
So while our cell phones probably have it in them to make super-accurate measurements, they don’t, says Jay Farrell, a professor and electrical and computer engineering chair at the Bourns College of Engineering at the University of California, Riverside. Taking very precise GPS measurements would jam up a phone and make it difficult for owners to also, say, Snapchat, or make a call—“whatever they do on their phones,” Farrell says.
But earlier this month, Farrell and his colleagues at Riverside published a breakthrough in the journal IEEE. The researchers’ innovation lies in how they combine GPS measurements with data from an inertial measurement unit, or IMU, which is an electronic device that can measure force and angular velocity. To make these two data points useful and precise, scientists have to figure out the integer number of signal wavelengths traveled between the satellite (in space) and the receiver (on Earth). The UCR team has found a way to clarify that integer number with many, many fewer computations. The upshot is that very precise GPS just got a lot faster, and a lot cheaper.
Why should you care? Farrell’s work focuses on autonomous vehicles, and he says this kind of technology will be vital in getting safe self-driving cars on the road. Autonomous vehicles need not only “road accuracy”—that’s precision down to the 10 meters, which lets the car know which road it’s on—and “lane accuracy”—precision down to the meter, which lets the car know which lane it’s in—but also centimeter accuracy, so that the car knows when it has enough leeway to securely switch lanes or make turns. This breakthrough means this accuracy is more firmly within manufacturers’ reach.
GPS is far from the be-all-end-all for autonomous vehicle navigation. “You’re not going to trust safety of life to one sensor,” Farrell points out. Researchers are also scrambling to improve upon and speed up visual odometry, which uses images and motion sensors to estimate position; lidar, which measures distance using laser light; and radar.
Indeed, interest in piloting technology will only grow. “People think of it as their right on the cellphone to know where they’re at,” says Farrell. The researcher has already gotten an email from a cellphone company—so it’s possible his breakthrough could eventually make it to your lowly smartphone, as well.