Why Not Use GPS?

When we start talking about robots and the need for sensors, the question that many people ask is: ‘Why don’t we just use GPS, surely that’s enough?’. GPS is everywhere—it’s in our phones, it’s in our cars … it tells us where we are. So let’s talk a little bit about GPS.

Imagine we have a robot on the ground and it’s got a GPS receiver in its head. In the sky are a number of satellites in low earth orbit, and they are all sending radio waves which the GPS receiver in the robot’s head can pick up.

Using the time that it takes the radio wave to travel from each satellite to the robot we can work out the distance from each satellite to the robot.

If we know enough of these distances—we know where the satellites are in space—then we can work out where the robot is. That’s the fundamental principle of GPS.

Now these radio waves are at a very high frequency; they are around about 1.5 gigahertz, and that’s significant and we’ll talk about why that’s significant shortly.

Imagine now we’re looking upwards from the robot. We’re looking up to the sky, and there are a number of satellites that we can see. We only need four satellites in fact, to work out where we are on the planet. That’s the minimum requirement in order to get what’s called a ‘fix’.

Now imagine that our robot is in an urban environment. Imagine we’re on a street in Manhattan or something like that. So what happens now is that some part of the sky is obscured. We can’t see all of the satellites. Now we can only see two satellites and that’s not enough to obtain a fix. This is a very common phenomenon, that’s why we have problems with GPS in urban environments. But it occurs in other environments as well. It can occur perhaps in a very, very deep mining pit where the walls of the pit obscure a significant fraction of the sky.
Now imagine that we have our robot and it’s at some big industrial complex. There are large walls and chimneys and structures made out of metal. Now what happens in this case is that the signals from the satellite may not travel in a direct line to the robot. They may bounce off some of these metallic structures before they hit the robot. The problem with this is that the signals have travelled a longer distance than the actual distance between the robot and the satellite. The path length has increased through what’s called ‘multi-pathed reflection’. So the GPS receiver within the robot doesn’t know this, so it will come up with an erroneous estimate of where the robot is located.

This is, again, quite a common problem with GPS next to a big structure that reflects radio waves. The GPS estimated robot position can be significantly in error.

Now let’s consider another scenario. Consider that our robot is underground, perhaps doing some mining work, a really important domain for robots. Now the problem that we have here is that the radio waves from the satellite cannot penetrate the earth.

Another application where we might want to use robots is underwater. And again we have the same problem as we have in the underground case. The radio waves at 1.5 gigahertz cannot penetrate very far into water. They perhaps penetrate a few millimetres. So a robot that’s at any depth can make absolutely no use of GPS information.

Now let’s consider a situation we’re outdoors but there are trees above the robot. The problem we have here is actually very similar to the last one. There is a lot of water in the leaves of the trees, and particularly after it has been raining then there’s a lot of the water on the surface of the leaves. And this water absorbs the 1.5 gigahertz radio waves from the satellite, and they won’t penetrate. So with heavy tree canopy GPS radio waves are absorbed, and we’re not going to get enough information for the robot to obtain a fix.

But there’s another and perhaps deeper reason why GPS is not the solution to all robotic problems. Consider this kind of typical case where I’ve got a robot and it wants to pick up an object.

So this robot wants to pick up that banana. Now I can add a GPS receiver to the robot, so the robot now knows where it is in the world. But it doesn’t really help, because it doesn’t really know where the banana is. So I can add a GPS receiver to the banana, and now the banana knows where it is; the robot knows where it is. But it still hasn’t helped with this problem of this robot knowing where the banana is so that it can move and grasp it. So in order to get information about the banana’s position to the robot I could add a radio transmitter to this GPS receiver on the banana, and I could add a receiver to the robot. So now the robot knows where the banana is, it has the coordinates of the banana; it knows where it is, and having those two pieces of information, the robot can then plan a path in order to get from here to there.

Now while this might work, the fundamental problem is that every object that the robot would want to pick up, work with and manipulate needs to be fitted with this kind of instrumentation, and clearly this is not the way we solve this problem. If I want to pick up the banana I simply reach out, I use my eyes to work out where it is and guide my hand in order to pick it up.