Why Every GPS Overestimates Distance Traveled

[Rex Sanders]

Fairly technical article in IEEE Spectrum:

Why Every GPS Overestimates Distance Traveled

Scientists ran one test walking around a 10-meter square 25 times: “… that pedestrian-course error of 10 to 20 percent is exaggerated because of the low-cost GPS receiver used and the short reference distances.” However, they found the same problems over longer distances in a car.

The less technical explanation is that small zigs and zags in position due to random errors always add up positive, so that GPS distances are longer than real life. I’ve known this for years, they prove it with formulas and big words like variance and autocorrelation that I used to know more about.

The good news is that the scientists think GPS receivers could compute much more accurate distances with a simple software change: “… moment-by-moment GPS velocity measurement is not subject to the same sources of error, so that calculating distance traveled by integrating velocity should yield reasonably accurate results.”

I’m skeptical, given the wildly inaccurate velocities I’ve seen on GPS receivers.

The full paper is free and linked in the IEEE article if you want to geek out. And I’m sure Bob Gross would have more to say.

— Rex

 

 

[Cayenne Redmonk]

I’m seeing differences of a few hundredths of a mile on walks I do all the time.

Their 10m box test is a neat way to publish, but seems largely irrelevant in real world pedestrian usage.

Also, It would seem to me that If a path is close to, but not exactly the straight line shortest distance path, then the GPS distance could be shorter than the distance traveled, if the error falls closer to the straight line than the actual path.

[Ken Thompson]

+ or – 15% has always been my experience with paper maps or GPS.

[Richard May]

Hmmm.. Okay, let me see if i get it. The over-estimate is caused by the accumulation of small errors in location that add up – like ounces equals pounds. It’s like drawing a jaggedy line from point A to point B and measuring every small line.

So, wouldn’t “smooting”, like when I drew a curve through dots on graph paper, make a reasonable approximation? Or is this what the paper proposes but by using speed to make the “guess” more accurate?

[Jerry Adams]

10 m square isn’t good, because it could “cut corners” producing an error.  Better to do a straight line that’s, for example, 1 mile long.  Then you can find the true distance just by looking at the end points which you can calculate by averaging a bunch of readings.

Or, it doesn’t need to be perfectly straight, broad curve is okay, get the true distance by just taking a few points along the curve.

I’ve done that many times, yeah, about 10% over estimate.

And you can reduce that by smoothing, or just deleting most of your points.

A problem is you don’t want to “smooth away” switchbacks.  I’ll smooth a track and then go back and walk it again and make sure I haven’t smoothed away any switchbacks.

 

[Jerry Adams]

and then there’s the odometer function on the GPS that tells you how many miles you’ve walked.  They use some sort of smoothing software.  An especially big error is if you stop for a rest, the GPS will keep taking points that bounce around a little, so if it counts that as distance traveled, it can accumulate a significant distance just sitting there.

the odometer reading has a much bigger error – can under read by 10% or over read by 20%.  Or even more.

I think they could have much better software to do this but they probably don’t prioritize this

[Art …]

I’ve seen the GPS watches of other runners “” underestimate “” distance travelled on trails numerous times.

this is typically over distances of 10 to 30 miles.

the error is usually in the 5% range.

rarely does a GPS watch of my running partners over estimate distance.

[Author]

Posts