Go Down

Topic: Distance with GPS impossible? <1% accuracy (Read 8264 times) previous topic - next topic


Asking for <10-feet accuracy from any GPS, short of military grade (and that will cost you well into the 1000's of dollars, and will require military clearance) is asking the impossible.  I've dealt with many off the shelf GPS models  (Garmin, Lowrance, Magellan, DeLorme, Pharos , just to name brand names.) and the idea of pinpoint accuracy is not likely with the store-bought type. (even with WAAS (Wide Area Augmentation System))

You have factors like cloud cover, surrounding objects (mountains, buildings, trees) which will obscure the signals in various ways, thus the receiver will only allow that 10-foot minimum. (trust me, when your little Garmin eTrex says 8-ft accuracy, You're lucky to be within 16-feet.)

I tried awhile back to make a robot-tank, based on the old VEX platform, and I could never get a simple navigation code to work, using just plain NMEA (National Marine Electronics Assoc.) format code.  The main problem, was the VEX controller's limited code (programmer 1.0), and speed.
it would always keep auto-correcting as the receiver received new coordinates. Sometimes, the device would just sit there, then suddenly jump, trying to get to corrected coordinates just inches away, then suddenly go back near where it had started. This, all on just one set of coordinates.
(to say nothing of not being able to decipher which way was north, unless it had moved a few feet. Unless you add a digital compass, a GPS will say you're facing any direction til you've moved, then computer in which direction)

Computing distance, is no better. the distance will always change because of the changing coordinates from the receiver. again, looking for less than 10-feet, is asking the impossible.. You'd be better off using a tape measure or a per-measured length of string.


Please pardon any mistakes I've made below, I replied in a rush.

@paul -
I can prove that the absolute error will cancel out (be the same) within a few miles and within a few minutes.  Time matters.  Please help me design an experiment.  I'd be happy to do it if you agree to its validity.  I have 2 in mind.  First, I'll go back and forth between 2 point about 1/4 mi apart.  Each time writing down the L/L.  Of course they will be different by a few meters.  Only a few meters each round trip.  Therefore the distance calculation will only be off by a few meters each time.  Much less than you think, because the errors partially cancel.  In other words my distance measured will not change much.  If I take a lunch break before completing my last single measurement of Point B only, then the distance will be different than the rest.  An outlier.  Agreed?  Do you want me to do it 10x to show you?

Second, at a street corner with known coordinates using multiple methods to calculate it...  The actual reading will be off by a vector say 45 degrees, maybe 10m.  Go to another known corner nearby.  It'll be off in the same direction and approximate magnitude!  Coincidence?  The map is in error?  I don't think so.  I've been doing this every week for 16 years, the results are always the same.

@KE7 -

Path - Driving a car on a curvy road, or hiking on a trail. 100's of data points are required to capture the curves correctly.

Accurately - I'd like less than 2%.  To test this walk the same short trail twice carefully.  Or drive as if the cops were watching.

I've clearly agreed already with the rest that you said in 1).

2) I've solved. There are more than one ways to do it cheaply.  And at least one expensive solution see below.

3) I told you what accuracy I wish for. Cheap and practical.  How does a 5th wheel device determine position within a lane?  Sorry if that's a stupid question.

My expectations are very realistic.  You're just giving me the canned form letter answer that you say to everyone else on this forum.  I'm the one who's telling you GPS cannot do much better than 2%.  Is that realistic?  I have solved the other problems to my satisfaction.  Therefore I am realistic in my expectations.  You don't think I can beat 2%?  It's impossible?  Difficult yes.  Impossible no.  I have a friend who does exactly what we're asking for at Conrail.  His accuracy is 0.1% and his budget is not much bigger than mine for hardware.  The only difference is that it's a curved rail, which of course makes it much easier.  We're talking.

@zoomkat -
1) Yes, it's easy to get within 1% when point A is in TX and B in CO as the crow flies.  But if the recorded path goes from TX to CO, forget it.  You'll be off by 10's of miles.

2) A cheap gyroscope does not work.  My friends have tried.  Over the course of the race, with all the engine vibration, violent changes in angle, and slow changes too, it forgets which way is up.  We have had some success with an expensive military grade unit.

3) I do not understand your solution.  Here's what I have that works.  I also have another theory to test.  With 8 riders having GPS riding 10 laps, just average their tracks.  A human can do this manually.  Picture dragging the whole polygon until it lines up with the others overall.  There are 80 per race so it's labor intensive.  Of course there will be minor differences.  This is what I want to measure!  It can also be automated like this: Average all the points in 1 lap.  For each lap and each rider, the result should be the same.  Shift those which are not.  I am satisfied except for the following limitations: You need access to multiple riders data.  You have to shift the polygon more than once within a lap to get to 1ft accuracy.

@gelfing -
I agree with your assessment of consumer grade WAAS.  More like 30ft most of the time.  I understand the problems in your project.  Mine were similar.  Over time 30ft is all I can expect to measure distance.  But while you're driving you can expect each individual sample to be much more accurate, relatively, than that.  Maybe inches.  Let me exaggerate to illustrate my point.  Let's say we're sampling our position 10x/sec.  Each sample might be off by 1" relatively, compared to the last.  But over the course of a trip to work we might be off by 100m.  These small error are additive.  The errors which matter are perpindicular to our path of travel.  If a point was wrong in the direction we're going, it wouldn't matter to calculate the total distance traveled.  If we assume it's a perfectly straight line (when it is), it's easy to fix the error.  It's not much harder when you're driving around a city with square streets and no lane changes.  What do we do about curves?  My friend at Conrail knows the answer to this when they're constant curves.  I'm getting closer!  If someone could throw me an idea...


Sep 26, 2011, 02:05 am Last Edit: Sep 26, 2011, 03:01 pm by sbright33 Reason: 1
It looks like I'm talking to myself above because...

I think we lost KE7... whatever his Ham call sign was.  It seems he can't hang with me.  Did I say something wrong?  He deleted his account on this Forum entirely!  I'll try to be nicer.  Sorry KE7!  Steve KB3SF

I've noticed some of his posts still remain elsewhere.  His account is NOT gone.  How did he do this?  Manually delete 100 posts 1 at a time?


How did he do this?

Probably the same way you have started deleting some of your more unusal post.
Google forum search: Use Google Search box in upper right side of this page.
Why I like my 2005 Rio Yellow Honda S2000  https://www.youtube.com/watch?v=pWjMvrkUqX0


No, it's very different.  I have never deleted a post.  Mine are being deleted by the Moderator.  Watch this one will be gone in a minute.  So no one will even know that this is true...

The moderator may delete stuff that is seriously off topic or deemed offensive by somebody, but I doubt that other material gets moderated (but i could be wrong).
Google forum search: Use Google Search box in upper right side of this page.
Why I like my 2005 Rio Yellow Honda S2000  https://www.youtube.com/watch?v=pWjMvrkUqX0


I read almost anything that catches my eye with an interesting looking title.
Except unix related PC questions, out of my depth there...
Designing & building electrical circuits for over 25 years.  Screw Shield for Mega/Due/Uno,  Bobuino with ATMega1284P, & other '328P & '1284P creations & offerings at  my website.


I will delete content-deficient, inflammatory content, sbright33 - please read your PMs.
"Pete, it's a fool looks for logic in the chambers of the human heart." Ulysses Everett McGill.
Do not send technical questions via personal messaging - they will be ignored.
I speak for myself, not Arduino.


I know I m going to regret jumping in here, but I have been experimenting with GPS accuracy for another form or racing, although it is MUCH slower than motorcycle racing. Yes, the error is like 6 to 12 feet. But when you are in a small area, the error seems to remain relatively consistent over a short period of time.

So if position A is measured and that measurement is off by x feet on a heading of n, that error remains consistent for a while. In my case, I need to do this for 5 to 10 minutes and 10 feet error, while not desirable, is acceptable.

Just my 2 cents.


Check this thing out:



I agree!  Cool product, but I don't see the price.  I'm sure it is more accurate than 1%, see it is possible!  To all you pessimists.  Especially with the odometer input, that makes it perfect for measuring distance.

Go Up