Can someone explain to me how the kind of errors described in this story about the police tracking Scott Peterson, the murder suspect
A coupl of relevant passages:
That doesn’t sound right to me at all. Commercial airliners and all sorts of military equipment use GPS receivers all day every day in mission critical applications and would not be able to tolerate errors like those. Likewise, I have never even heard of a consumer grade device that has those sorts of errors.
If the errors are true like they stated, then they have to lie in the mechanism that transmits the location to the police and not within the GPS itself.
It depends upon the type of unit involved. I worked for a company which sold a variety of GPS tracking devices for vehicles. The cheap units were crap, and produced all kinds of erronious readings. If the PD was using one of those, then it’s not surprising they got a lot of bad readings.
Except that we have no idea how big those errors are. GPS does not, to my knowledge, directly measure velocity. What it does measure is position, from which velocity is inferred. And there can be errors in those position measurements. For instance, suppose that the GPS makes a position measurement once every millisecond. And now suppose that one of those measurements is exact, while the next one has an error of 50 feet. If you naively computed a velocity from just those two points, then you would find a speed of 50 feet / 0.001 second, which is about 30,000 MPH. In other words, a huge error in velocity, despite a reasonably small error in position. Now, of course, this isn’t the most prudent way to use your data points, but I wouldn’t put it past a lawyer to use such an example to try to show the device unreliable.
I wonder if they are talking about a combination GPS/inertial device? Such a device uses GPS to measure its position, but also uses an inertial sensor to detect turns and such, so that the exact position on a road can be calculated (the device can also continue to track if the GPS loses signal - a possibility in a downtown area).
I can imagine that a door slam or other vibration could possibly confuse an inertial sensor.
That is a really bad way to design a GPS though. I would think that anyone that designs one would have the sense to smooth the data over time and throw out absurd values. It is not as if you need calculate a new velocity measurement every second and broadcast it. I am not saying that it didn’t happen. I am just saying that it is ridiculous for soemone to design one that way and for the police believe that it is just an inherent limitation of the technology.
Wind and slamming doors sound bogus. You may get spurious readings on a GPS if you don’t get a strong enough signal which can happen if you’re under tree cover or if the satellites are in a bad configuration (too low in the sky for good reception or, I think, too clustered together). But the position is sampled constantly, so you just throw out the bogus data points. For the purpose of tracking someone’s movements by car, it’s absurd to say that GPS isn’t sufficiently accurate. You might be able to get a speeding ticket conviction thrown out if it were based on GPS, but you’d have a hard time arguing that you weren’t within 10-15 yards of where it said you were.
The vbox (http://www.racelogic.co.uk/2003/vbox/aboutvbox.htm) can measure acceleration and velocity with prety good accuracy.
From this page: http://www.racelogic.co.uk/2003/vbox/gps.htm
Also:
So, yes, wind and slamming doors can affect the GPS, but that would only matter when you are trying to do pressicion measurements. In any other case, the GPS might be malfunctioning.
When I hear that “wind and slamming doors” are affecting a positional measurement, I think of barometric altimeter readings. I do quite a bit of skiing and walking, and I have a watch with an altimeter built in. It is normally very accurate, as long as the atmospheric pressure remains stable (i.e. no weather fronts moving in), but if I’m, say, on a very windey ridge, the height reading can go all over the place. Same if you’re in a car and slam the door.
Now, most consumer GPS units also have a built-in barometric altimeter. I wonder if someone somewhere has got their wires crossed and conflated the two?
You can get really screwy results on a GPS. I use one on my motorcycle all the time and sometimes it will tell me I’m traveling at 150mph when I know that even if I wanted to my bike couldn’t go that fast. I think I’ve still got my fastest speed recorded at 1800mph when I was in Europe!
I have however, never heard of wind or a slamming door causing any problems with a GPS. The upper end models, like the ones pilots use, cost a lot more and are supposed to be more accurate. I can’t imagine the police using an expensive GPS when for the most part the cheaper ones will work just as well. I can’t imagine anyone falling for it saying they don’t work right because of a couple bad readings. Then again if they couldn’t get off the jury…
I wouldn’t say most, though the Intersema “calibrated pressure transducer on a chip” and variants are being seen in an increasing number.
GPS professional chiming in here…
It sounds like a software bug or two compounded with the natural problems with the setup.
For this kind of system, you have the constraint that the driver isn’t supposed to see it. That immediately rules out all of your good antenna locations. In fact, being that most cars are not L-band transparent, most of your signals will probably be bouncing off the ground. You have to make all kinds of compromises with low signal power, letting in some really bad measurements you would normally be filtering out.
The software should be smart enough to figure out the really bad outliers. But you can see how they might work under the philosophy to collect everything, bad or good, and filter the data later. You might be able to more intelligently sort out the good parts that way, but it probably looks bad to the lawyers. If you’d filtered before you recorded, they would never have seen the bad stuff.
They’ve also admitted some software glitches. There are probably some worst-case errors coming in that weren’t accounted for.
You also have the problem of limited battery life. That’s where I think the “wind and slamming doors” comes in. It sounds like there could be an accelerometer based switch to turn on the tracking unit only when the vehicle is moving so you can save power. A cheap, low-power accelerometer could be confused into thinking it’s moving from “wind and slamming door” jerks, and could be thinking it’s stopped when it’s at constant velocity (just looks like a drift within the spec of the part).
Now, that’s just speculation as to how they designed it. But that’s the only problem source I can think of because atmospheric delays from weather and jerk from slamming doors isn’t going to affect signal quality significantly at this level of accuracy.
A commercial GPS receiver will not operate at these speeds. Most of the units I am familiar with have a cut out at something like 600 knots. Above that speed, they stop giving positions. This is to prevent someone from using a commercial GPS receiver to guide a missile.
If they got that kind of speed report, then there is a problem in the processing that was done of the GPS data, rather than an error from the receiver itself.
I can think of plenty of ways to screw up calculations involving the time and position data to come up with a huge error. Thing is, though, if you are using the velocity output of a GPS receiver you get that kind of error.
I can think of plenty of ways to screw up calculations involving the time and position data to come up with a huge error. Thing is, though, if you are using the velocity output of a GPS receiver you won’t get that kind of error.
Not exactly on point, but some of those errors might have been on purpose
Warning: contains fodder for the tin foil hat crowd. 