Why RF pattern-matching beats GPS in big cities

Recently, Urgent Communications Editor Glenn Bischoff chatted with Per Enge, Ph.D. — a self-described “radio geek” — who became the chief technical advisor to Polaris Wireless at the first of the year. Enge — who is a professor of aeronautics and astronautics at Stanford University, where he directs the school’s GPS research laboratory — spoke about the challenges of location accuracy and how RF pattern-matching can be utilized to mitigate those challenges.

So, how just how far back does your interest in radio go?
My parents gave me a Heath kit when I was a kid. I don’t know if you remember those, but all of the parts were in there and you’d have to place all of the components onto the printed circuit board — and then hope that it worked. I was very pleased with myself — I got it all put together in about six hours or so, and it did work.

You’ve worked for the last three decades on GPS, which you’ve described as an “extraordinarily gorgeous thing.” Why do you say that?
There are two reasons. First of all, radio navigation technology is deep. Second, everything you learn in your electrical engineering classes … all comes into play and it’s important. So, on the one hand it’s very satisfying for a geek. But on the other hand, it’s something of this world — it’s a real thing. We save lives every day, we save unimaginable amounts of fuel every day. So it’s a very human technology — and it doesn’t get much better than that.

How has GPS evolved over the past three decades?
I was part of the second wave of engineers that worked on differential GPS, which allowed us to improve accuracy from about 100 meters down to 1 meter or so. We started with the maritime sector and then in the 1990s began to move into aviation, specifically for satellite navigation for landing airplanes.

What’s the next phase of the evolution?
When you look at GPS, it traditionally hasn’t made much use of databases, because the computers and the memory weren’t there to accommodate large databases. That’s why I’m excited about my involvement with Polaris Wireless, because they’ve really double-downed on that. It’s where databases meet radio navigation.

How so?
RF pattern-matching, which involves collecting signal strength measurements … and then correlating them with a database that describes what the signal strength should be in a given city.

Why is that important? What new capabilities does it create?
It’s very accurate in urban areas and even indoors. That’s because the pattern gets complicated around buildings, and it is that grain, that detail that RF pattern-matching picks up on. For instance, it knows that the signal from base station No. 8 is very strong, which means it’s probably coming down this street, but the signal from base station No. 5 is weaker, which means that it’s probably coming from the other side of the building. Here’s one way to think about it: crumple up a piece of paper, and then unfold it — all of those wrinkles you see represent the grain, or detail, that the technology picks up on — and that’s a good thing.

How is that information used?
Every cell phone today — every single one — looks for the base stations around it then generates a network measurement report that gets sent back to the server that identifies all of the base stations in the area and their signal strengths — it’s all built in. And based on those measurements, the server can identify where you are.

How accurate is that location identification?
GPS, in an open-sky situation, is accurate down to 1 meter, or maybe a half-meter. But when you go into cities, two things happen: it becomes less accurate — maybe only to 50 or 100 meters — but the real challenge for GPS in cities is coverage — it just doesn’t work [because it depends on line of site]. So, while RF pattern matching is accurate to about 50 meters, the important thing to remember is that the coverage is much, much stronger in these very difficult areas.

One of the limitations of location technology is that it does not identify locations vertically, the so-called “Z” axis. Does RF pattern-matching help in that regard?
We’re working on that, and refining it. Basically, it works in the sense that as the [signal] from the base station goes up in the building, it gets clear of the clutter from the neighboring buildings, which is an effect that we take advantage of.

So, as the signal rises in the building and clears the clutter it gets stronger, and by measuring that change via RF pattern matching, you can determine, with reasonable accuracy, what floor a person might be on?
Correct. That’s a very good statement of it.

What do you think? Tell us in the comment box below.