Where Are You—Exactly?

Have you ever tried to order pizza on your cellphone from the basement of a large building? If so, it was probably a frustrating experience. If you then tried to find the nearest pizzeria from the Google Maps app, your frustration probably increased.

Although wireless connectivity and localization are improving, most of us have had the experience of going into a large building and having our cellphone signal deteriorate—or even disappear. Not only is this annoying when we want to order pizza, but it can also be dangerous if, for example, rescue workers can’t locate you in a disaster.

And large buildings aren’t the only culprits of poor wireless signals. Depending on the device and location, mountains, trees, clouds, and other blockages can cause similar disruptions. And indoors or outdoors, when signals decrease, so does our ability to locate the devices we care about—from cellphones to refrigerators to self-driving cars.

An ECE team led by Harpreet Dhillon and Michael Buehrer is working to bring reliable wireless connectivity and localization to these low-signal areas.

The team is exploring two new technologies: the millimeter wave (mmWave) wireless signals that are becoming more common with 5G, and reconfigurable intelligent surfaces that can help direct signals wherever we need them.

Millimeter wave

As their name suggests, mmWave signals have a wavelength an order of magnitude smaller than traditional wireless signals and travel at a much higher frequency—10+ GHz instead of the more common 2 GHz. This higher frequency means more information can be transmitted faster, which is spurring its adoption.

The higher frequency also makes these waves better for localization since they can more accurately measure the time it takes signals to travel from a base station, explains Buehrer. “Millimeter waves can also create very narrow beams, which can accurately estimate the angle from the base station,” he continues. Both are important for locating a device.

The standard way to locate a wireless device is to measure how long signals take to reach it from base stations with known locations, like cell towers. Using signals from three known locations, you can pinpoint the device’s location. In the messy world of reality, however, these signals can take multiple paths and reflect off surfaces, notes Dhillon—and you might not always have three signals to work with, making the angle the signal travels from the base station critical information.

Reconfigurable intelligent surfaces

A downside of mmWave is that it is easily blocked by physical objects—like trees or walls. To mitigate this, especially inside buildings, the team is investigating the use of reconfigurable intelligent surfaces to redirect signals around corners. According to Dhillon, “mmWave signals have excellent reflection properties, giving mmWave devices a way to receive strong non-line-of-sight signals and allowing us to locate a device using a single anchor node. This becomes even more useful when we can control these reflections with intelligent surfaces.”

These surfaces might be about two meters tall and wide, very thin, and can be placed on almost any surface. Each pixel is about 1cm², and can be individually controlled. “We can control the angle a signal takes when it leaves that pixel and steer around corners,” says Buehrer. “This lets us steer signals around corners and overcome the limits of non-line-of-sight propagation.”

“One of the primary challenges of geolocation,” says Buehrer, “is that the signal doesn’t take a direct path from transmitter to receiver. But if we can use these surfaces to direct signals, we can use that information to get a better estimate of where the receiver is located.”

These surfaces might also allow the team to accurately locate a device vertically—not just horizontally like current systems. This is extremely important for applications in buildings, for example. “One of the hardest open problems in localization is reliable localization in the vertical dimension,” says Dhillon. And these new technologies might hold the answer.

Algorithms and modeling

The team is developing these location algorithms, starting by modeling the perfect scenario involving reconfigurable surfaces. “There is a fundamental limit to how accurately we can measure these signals,” Buehrer explains. “We have to start by knowing what that limit is, what is the best we can possibly do.” After determining this, the team will move on to crafting the practical location algorithms.

Building on this, the team will use stochastic geometry to capture randomness in the placement of surfaces and characterize the reflections from them. “The effect of reflections on the mmWave signals isn’t well understood,” says Dhillon. “We need to develop fundamentally new stochastic geometry tools for the system-level analysis of mmWave localization.”

The team will be developing stochastic models for all key network elements, so their analysis is not limited to a specific network topology, explains Dhillon. “Point elements like base stations can be visualized as collections of points and modeled as point processes. The intelligent surfaces can be modeled using germ grain models.” One advantage to this approach, he notes, is that it can determine the network-wide feasibility of the localization algorithms.

Ultimately, the team hopes to achieve centimeter-level accuracy for locating wireless devices. If you’ve ever looked at your cellphone’s location on a map, you might notice that it’s accurate within a few meters, at best, when indoors. Dhillon and Buehrer also plan to use their findings to create an open source mmWave localization simulator so anyone can explore and test algorithms. "Please deliver my pizza to the basement conference room—you’ll see where my phone is."