Navigating better satnav
Many of us would be - literally - lost without a satnav, but sometimes this technology can lead you up the garden path and other times it loses the lock on the satellite signal at the just the wrong moment and you end up on a one way system with no escape! You could look upon it that at least there’s a human behind the wheel to keep things under control. But what about autonomous cars, which will depend upon satellite navigation to determine their position? Will they end up driving into walls or at a standstill because they can’t tell where they are? Katie Haylor went to Queens College in Cambridge to meet Ramsey Faragher who’s developing technology to improve the accuracy of satnav positioning. First up, she asked Ramsey, when we dial up a destination on our smart phones, what’s actually going on in order to get us there?
Ramsay - Firstly, your satnav is a radio receiver that picks up satellite signals from space. They come from 20,000 kilometers away, so they're very, very weak by the time they get to your satnav. And your satnav performs some maths and it allows it to calculate where it is. Then the second important technological part, a layer on top of that, is the actual navigation part. That routes you to the destination you want to go. So completely separate to the maths and the science of satellites and radio are algorithms that try to work out the most efficient way for you to get where you want to go from where you are.
Katie - Now, in theory this is great, but I can't be the only one who's been taken literally around the houses by my satnav. How accurate are they?
Ramsay - You can have all sorts of issues, so there might be a bug in the routing algorithms and the navigation part. The GPS might be suffering problems and when you're in cities it's quite likely that the GPS part will be suffering problems. The basic accuracy of the sort of GPS chip that's in your satnav is about one meter on a very good day, if you stood on top of a mountain with perfect clear view of the entire sky. There are really clever really expensive really complicated ways of processing GPS to get down to like one millimeter positioning but that's not the stuff that's in the cheap receivers in our handsets. The problem with being in a city is that the signals can be blocked by the buildings and you want as many signals as you can possibly get. And the second more serious problem is that the signals bounce around between the buildings before they get to your device. But all of the maths that goes on inside your receiver assumes that the signals are travelled in a straight line. If the signals have bounced off buildings before your receiver picks them up, they've actually gone further than they would have done otherwise, and the calculated position is wrong.
Katie - So we've come out to the back of Queen’s College. Which road are we on?
Ramsay - I think this one's called Queen's Lane, conveniently. But yes, it runs down the side of Queens College and round the back of St. Catharine’s and past King’s College as well. So we're in a very narrow street and we're in what in my world we call an urban canyon. And it means those you can see we have a very thin strip of the sky above us so we can't see the entire set of satellites that are up there. And that's one problem: you want a really good geometry to get a good position fix. You want the satellites to be in all angles in the sky. But these walls on either side of us are making the signals bounce repeatedly before they come down to the ground. So the signals from the sides have travelled further than they should have done. And that's why when we look at the little dots on the map as we walk along you'll see that it's not actually doing a good job of keeping us in the passageway. As you can see it thinks we're inside St. Catharines college at the moment and we're not. And it's just moved to the other side of the street.
Katie - It's not even on the street right now.
Ramsay - That's bouncing around 15 metres either side of the road. If we took this receiver to the top of the mountain it would be accurate to a couple of meters, but it's giving us a 15 metre error just because of these buildings around us.
Katie - Famously, no mountains in Cambridge.
Ramsay - Exactly.
Katie - So we're going to test not just the little blue dot but how your phone is actually going to direct us somewhere, so King's College Chapel is a very nice destination to pick. Once it's found the routes, we'll get going.
GPS audio - Turn right onto King’s Lane
Ramsay - It's certainly putting us near Queen's Lane. It still thinks we're in the middle of St Catherine’s College. But it knows that we need to get ourselves onto King’s Lake even though the GPS bit is struggling at the moment. The routing algorithms that plan the quickest route from A to B are still able to do their bit, even if the GPS is a bit wonky.
GPS audio - Turn right onto Kings Lane
Katie - Of course this is all very well. For a human. You can just use a common sense and say I'm not in St. Catharine’s College, I'm outside in the street. If we are to have autonomous vehicles on the road in the mainstream, I'm guessing a robot doesn't really have that kind of common sense.
Ramsay - That's right. Right now, today, people building these autonomous systems, they can pick a path, right? They can either build robots that don't trust the world and think they're being lied to, and maybe go against the rules sometimes or decide to do their own thing - that might be a dangerous path for us to go along. Or, we make the sensors and the technology that the robots rely on much more reliable and make them tell the truth more often. This is actually a key thing my company is doing: the software change that we make the GPS receivers, we prevent the GPS received from lying to the rest of the system. So we give an error estimate that's true and we improve the accuracy as well. And so that increase in integrity is what's really important for autonomous vehicles. So they don't mind being told an inaccurate measurement, as long as they're told that it has a large error on it. The big problem is if you pass some measurement to an autonomous platform which has a very small error but in fact was a very wrong measurement. So high integrity sensors are what needed to ensure that the future autonomous vehicles are as good as we are
Katie - So you managed to successfully navigate our way back to your office, even though your GPS got lost. I mean, we obviously knew where we were going. How can satnav be made more accurate?
Ramsay - So the good news is there's three important changes that are coming that will make GPS much better for both us humans and for the coming robots as well. So the first one is simply that the more satellites there are in the sky, the better the performance you get. And what we all call satnav GPS, sometimes casually GPS, is the American system. There's a Russian system, a Chinese system, and a European system. And in the future there might be a British system, after Brexit. So there'll be hundreds of satellites in the sky. That will help.
The second important change coming is that the satellites do improve gradually over the decades and have new technologies in them, and there is a new signal type that is rolling out at the moment, and the fundamental performance of that new signal tape is about 10 times higher than what we get today. So up on a mountain you'll get 30 centimetre accurate positioning instead of sort of 1 to 2 metres. In cities you'll still have the problems that we've already discussed about the signals bouncing around, and there being extra path length in there, and the maths being a bit wrong for cities.
And so the third important change is the sort of change that my company is providing, where we provide a software upgrade to GPS receivers that changes how they work; accounts for this sort of physics that goes on in the cities. That means that the receiver can cope and understand and deal with signals that come from different directions and aren't coming from the satellite itself. Over the course of the next few years, all of those things will come together to provide much better performance in cities than we have at the moment.