Shop OBEX P1 Docs P2 Docs Learn Events
Accuracy of Parallax GPS Module — Parallax Forums

Accuracy of Parallax GPS Module

LittleGrizzyLittleGrizzy Posts: 5
edited 2009-02-27 13:29 in Accessories
I have been working on an autonomous vehicle using the protoboard as a controller. I was planning on using the GPS module as was wondering how accurate it was. Approximately how far does the vehicle have to move before the GPS will register new coordinates?

Comments

  • Erik FriesenErik Friesen Posts: 1,071
    edited 2009-02-12 15:34
    I would say about 5 feet. You might get better results by averaging although it would take more than a few seconds to do so.
  • JonathanJonathan Posts: 1,023
    edited 2009-02-12 16:27
    I think it's going to be more than that, at least the smallest *average* distance that cna be detected. It's going to depend on # sats in view and location, atmospheric conditions, the details of the Parallax gps, with which I am not familiar. 5 feet seems optomistic to me. I'd do some empirical testing here.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    www.madlabs.info - Home of the Hydrogen Fuel Cell Robot
  • John AbshierJohn Abshier Posts: 1,116
    edited 2009-02-12 16:30
    I checked the manufacturer's website (www.polestargps.com). They state an accuracy figure of 4.2 meters at 95% probability. My experience with GPS is that averaging for a few seconds doesn't help because the errors are correlated.

    John Abshier
  • SRLMSRLM Posts: 5,045
    edited 2009-02-12 16:45
    I'd guess about 10 feet.
  • pmrobertpmrobert Posts: 669
    edited 2009-02-12 18:09
    My experience with other GPS devices is that a movement of ~24" can be detected reliably on the next fix. The new fix will reflect a fairly correct delta from the previous fix but both are subject to the 4.2m error range stated. My experience is with WAAS corrected units, I don't know if the unit under discussion is able to acquire WAAS or not - the datasheets are silent on this, AFAIK. If I find some time later I'll disable WAAS and see if my observation holds true.

    -Mike

    For clarity, I am answering the question "Approximately how far does the vehicle have to move before the GPS will register new coordinates?" not the question of absolute accuracy and size of error zone.

    Post Edited (pmrobert) : 2/13/2009 9:23:00 PM GMT
  • Erik FriesenErik Friesen Posts: 1,071
    edited 2009-02-12 23:23
    Yeah that is probably correct, 5 feet would be optomistic. The interesting thing about it is that the signal is messed up on purpose.
  • SRLMSRLM Posts: 5,045
    edited 2009-02-12 23:55
    This is the perfect example of the kind of thread that should be in the new Sensors forum...
  • awesomeduckawesomeduck Posts: 87
    edited 2009-02-13 03:18
    I found the Parallax GPS to be off by 10 feet between my front door and what google maps said the same coordinates should be. I also found that the altitude was off by about 50 feet compared to what the flood plane maps for my county said it should be. No telling how accurate google is. You can just enter in the coordinates, like "31.602233 -83.938983", on google maps and see where you are. I will say the GPS device was precise though...as long as I have at least 5 sats acquired I get the same readings.

    As for the GPS signal being purposely messed up...that is no longer true. That was a military feature called selective availability. The idea was to purposely add error (for the civilian signal) during military operations. However, now that so many critical devices rely on accurate data, like airplane and auto navigation, they no longer purposely add error. Needless to say though, we're dealing with time deltas in nanoseconds from satellites thousands of miles away, on a piece of hardware you can cobble to together on a bread board and write your own code for. So 10 feet is pretty freaking amazing when you consider the complexities involved.
  • Chris SavageChris Savage Parallax Engineering Posts: 14,406
    edited 2009-02-13 16:35
    Jonathan’s answer seems most complete in that the datasheet lists the accuracy under ideal conditions. Number of satellites and other conditions will affect accuracy. We have seen drift as much as 30 feet between readings when inside a building (where it’s not designed to work) which can give you an idea of how things can affect the readings.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Chris Savage
    Parallax Engineering
  • rjo_rjo_ Posts: 1,825
    edited 2009-02-14 07:10
    Jonathan,

    The peculiar thing is that my Parallax GPS and my Telit GPS module seem to drift in similar amounts. And I get drift, when I have 6 or more satellites on line. I get the drift inside and out. I haven't checked to see if two drifts are synchronous or not... but it wouldn't surprise me if they were. The drift can be substantial... giving a bogus velocity measurement of 5mph or more.

    The solution might be to use multiple-sensors... for example you might (but probably wouldn't) use short-term accelerometer information to cross-check the GPS data. When we get cameras... the image data will also be theoretically useful but practically worthless.

    The good news is that if you want an autonomous vehicle, then you have to use GPS.

    The really good news is that by itself, the GPS data is not quite good enough.

    Rich
  • BOTing AroundBOTing Around Posts: 2
    edited 2009-02-20 19:15
    I may be completely out there as I am just starting with GPS units but I want to know MUCH more before I sink a whole lot of money into it's integration.

    Is there a way to increase the sensitivity of the GPS unit, such as a board mod or bigger/more sensitive antenna for a cheap price? How about an alignment feature that could be done as part of a start-up sequence to increase it's accuracy?

    Thanks in advance.

    Sebastian
  • John R.John R. Posts: 1,376
    edited 2009-02-20 20:02
    The error and drifting has less to do with the software and settings, and more to do with the nature of the system and the capabilitiies of the hardware. As awesomeduck wrote, you're basically using delta times of nanoscecond differences to triangulate your position in 3 dimensional space. Differences in the atmosphere could potentially be enought to throw off the timing. Hardware plays into things in that the system can only report position as accurate as the hardware can calculate, including rounding in terms of times and when doing the math. Here is a case where the difference between 2/3 = .66667 and 2/3 = .666666666667 can make a difference.

    The way that systems (such as landing an airplane with GPS) get more accurate is by using "differential GPS". As one of the posters noted, he generally gets the same reading on his door step, even though it's "off". Differential GPS puts another transmitter at a known location, and by combining this signal with the satelite signals, you can a location down to inches, depending on the combination of components.

    I know about this conceptually, but have no idea how to set it up, or what you'd need to do it.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    John R.
    Click here to see my Nomad Build Log
  • BOTing AroundBOTing Around Posts: 2
    edited 2009-02-21 04:57
    That was a great post John.

    Thank you very much, I hope that all replies are just as good. Hmmm now onto Differential GPSing.

    Sebastian
  • FranklinFranklin Posts: 4,747
    edited 2009-02-21 18:37
    I seem to have read somewhere that the main difference between the inexpensive (cheap?) units and the more accurate ones is the accuracy of the clock. The closer you can calculate the difference in the satellite response the more accurate the position. Also the 5m accuracy is not the same as sensitivity to change in location.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    - Stephen
  • LittleGrizzyLittleGrizzy Posts: 5
    edited 2009-02-27 13:29
    Thanks for the many responses so far. Should I be worried about interference from the motors or other components near·by? There are a few inches and 1/8th inch acrylic between·the GPS·and the motors but I just want to be safe.
Sign In or Register to comment.