You may have noticed a problem when you try to use your smartphone to navigate in a big city: your GPS location is usually super inaccurate. Sometimes it is only a few feet, but if you are in a particularly dense part of the city where satellite signals are blocked by high-rise buildings, the discrepancy can be orders of magnitude larger. For most people, it is just one of the many annoyances of urban life today. But for companies that rely on two people with smartphones to find themselves in a steel-concrete labyrinth, like Uber, the inaccuracy of GPS is a source of endless pain and frustration.
Like the boats that pass at night, a driver may be in a corner looking for a cyclist on the other side of the block. This can often lead to canceled trips. Uber calls it "wasted supply," which is money from the driver's pocket and also from Uber. This becomes exponentially more problematic when you have several pilots in a car, to the UberPool. And when you start to think of cars without driver wandering aimlessly through our urban canyons, desperately looking for riders with unreliable GPS coordinates … talk about your dystopias.
Meet the Uber ShadowMaps team
Recently, I sat down with two Uber engineers who may have a solution for all this chaos. Andrew Irish and Danny Iland were PhD students at UC Santa Barbara when their startup Shadow Maps was acquired by Uber in 2016. Since then, they have been working on integrating their technology into the Uber application. Recently beta tests began in 15 cities around the world and, according to the first results, they now get GPS signals that are twice as accurate as before.
"I usually say that when someone asks me what I do, I move the blue dot," jokes Iland. That's important because when I open the Uber application while I'm sitting at the desk on the 14th floor in the Verge office in lower Manhattan, I see my blue dot floating out of a hotel that is more than 200 feet away. What Irish and Iland are doing is trying to move that blue dot down the street to make it as accurate as possible when an Uber driver comes looking for me.
The Global Positioning System project was launched in the early 1970s as a way to overcome the limitations of previous navigation systems. It was originally designed for things that fly, like airplanes. So, one of the basic assumptions was that all the satellites would have a direct line of sight, which means that the signal would always travel in a straight line. But now, those assumptions have changed, thanks to the ubiquity of smartphones and the rise of location-based services like Uber.
Suddenly, the satellites went from tracking airplanes to tracking the smartphones of people walking through dense cities. These satellites not only lost the advantage of their line of sight, but had to deal with a forest of tall buildings that acted as mirrors to refract and distort the signals. This phenomenon, commonly called "shading", can create a location error of up to 100 meters or more, especially in high-value markets such as New York and San Francisco. That can wreak havoc on the most sensitive aspect of Uber's business: retirement.
"Then, obviously, this presents problems for us because we might think that a driver is on a different path than what it really is," says Iland. "And that can cause the ETAs to go out completely."
"That can cause ETAs to be totally out of control."
To solve the problem, Iland and Irish used a process called occlusion modeling, whereby the Uber algorithm examines a complete 3D representation of the city and makes a probabilistic calculation of where it is located, which satellites it can see and which it can ". There are about 30 satellites in the US GPS constellation, as well as a constellation of Russian GLONASS satellites (China and the European Union are in the process of launching their own GPS satellites). Satellites available for software developers, Uber can use a removal process to get a more accurate reading of where he is when I'm trying to take a walk.
"They claim that there are only three satellites," says Iland. "So, if I can see satellites C and D with high signal strength, but I can not see satellite B, then it's probably on the left side of the street where satellite B is blocked by a building, if I can see B and C, but I can not see D, so it's probably on the right side of the street, and then we're using satellite visibility information as part of the algorithm, instead of just assuming that all the satellites are in view. "
The raw GPS data (in red) versus the improved data (in blue) using the ShadowMaps probability algorithms. Image: Uber
He adds, "It's basically like negative information, you can make assumptions based on what you do not see, as well as what you see."
Irish and Iland use ray tracing, in which signals from satellites are color-coded according to their strength, to better illustrate possible locations and cut noise. A confusing effect is multipath fading, in which satellites emit signals that bounce off buildings that cross each other, strengthening certain signals that would not otherwise be in direct line of sight. This can complicate the ShadowMaps algorithm and draw erroneous conclusions about the most likely location of both the driver and the driver.
"So if you draw a trace on a grid," says Irish, "and these grid locations are based on how well they match the satellite signal strength and the 3D construction models, many of these noisy effects We were describing before average. "Each probability calculation takes between 20 and 100 milliseconds, and can run every four seconds for passengers and more frequently for drivers," says Iland. "You just want to have a more accurate estimate to account for how much faster the cars move, "adds Irish.
Ray tracing, multipath fading and heat maps
Later, we'll take a walk on the west side of Manhattan so they can show me how much better the modified version of the Uber app is to find our GPS location. Sitting in a conference room in the Uber office in New York, my head is spinning with technical jargon. But in the open air, it's easier to see how the simple corrections of the ShadowMaps team make a big difference. We walk next to a park, where there is a wide view of the sky and, as a result, a more accurate GPS reading. We go down a street, our view of the sky narrows, and suddenly our blue dot jumps to the other side of the block. Then Iland shows his version of the application: our real position and the blue dot are more or less synchronized.
ShadowMaps is not necessarily a unique approach to the problem of GPS accuracy. Paul Groves, associate professor at University College London, leads the research on robust positioning and navigation within the Geodesy and Space Navigation Laboratory. He praised the ShadowMaps team for bringing GPS improvements to a wider audience, but failed to offer them full credit for their work.
"Their approach is not unique," says Groves. "It's partially based on the work we've published in the open literature, there are also several other researchers working on similar approaches, we now have a demonstration system running on an Android smartphone in real time, though restricted to central London."
Much of this boils down to a preference in the methodology, Groves said. "The more people use this approach, the better in my mind, it gives you a better performance." If people are really using our research, instead of leaving and forgetting about it, then we are making a contribution to the world, rather than just generate piles of paper. "
"The more people who use this approach, the better in my mind."
Iland, Ireland, and their team began working on the integration of occlusion modeling in the Uber application more than a year ago. In February 2018, the beta test of the corrective algorithm began in 15 cities in which Uber operates. (Some of those pilot cities closed, however, after Uber sold his business in Southeast Asia to rival Grab.) Early results show a two-fold improvement in GPS accuracy and thousands of cancellations less each month . Now they are waiting for the green light to take on their global project.
Improving GPS accuracy has nothing to do with giving Uber more or less permission to track your location. Last year, the company brought out a highly criticized feature of its application that allowed it to track passengers up to five minutes after a trip in an effort to fix its reputation for poor customer privacy. In contrast, Iland and Irish say they are simply using publicly available data in satellite locations to make the whole process of calling a car on their phone more fluid.
"How often do you have to manually type in your address, versus how often we predict accurately without having to do anything?" Asks Iland. "So that makes it a more magical experience." When you open the application, we already know which is the correct address and all you have to do is press ".