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

GPS Accuracy

John BoardJohn Board Posts: 371
edited 2012-06-11 23:53 in General Discussion
G'day,

I've been looking into using a GPS lately for getting the position of an outdoor robot. I'm wondering what kind of accuracy I could expect from these parallax products:

http://www.parallax.com/Store/Sensors/CompassGPS/tabid/173/CategoryID/48/List/0/SortField/0/Level/a/ProductID/644/Default.aspx
http://www.parallax.com/Store/Sensors/CompassGPS/tabid/173/CategoryID/48/List/0/SortField/0/Level/a/ProductID/645/Default.aspx
http://www.parallax.com/Store/Sensors/CompassGPS/tabid/173/CategoryID/48/List/0/SortField/0/Level/a/ProductID/703/Default.aspx
http://www.parallax.com/Store/Sensors/CompassGPS/tabid/173/CategoryID/48/List/0/SortField/0/Level/a/ProductID/757/Default.aspx

Is there any way to increase this accuracy by having 2 GPS's a fixed distance apart on the robot? I'm looking for accuracy between 1mm - 50mm (0.03" - 1.96"). If this is not achievable with paralodian products, where would I go next to find something with that kind of accuracy (at a decent price - No more than $200).

Thanks,

-John

Comments

  • LeonLeon Posts: 7,620
    edited 2012-06-09 03:58
    1 mm - 50 mm accuracy from GPS just isn't feasible! You might manage a few metres.
  • GadgetmanGadgetman Posts: 2,436
    edited 2012-06-09 04:05
    A single 'nomal' GPS won't give you better accuracy than 7meters no matter what you do.
    A WAAS-enabled GPS should be able to get you accuracy into the 1 - 3 meter range depending on weather and location. (A lot of handhelds support this)

    Differential GPS (using the US CoastGuard service) should be able to get you close to 10cm accuracy if you're not too far from one of their transmitters. This stuff is beyond your budget...

    If you want to bleed money, you could get a Real-time Kinematic GPS system...
    http://en.wikipedia.org/wiki/Real_Time_Kinematic
    Then you get relative accuracy in the 1cm range...
  • wjsteelewjsteele Posts: 697
    edited 2012-06-09 04:25
    Actually, that is not correct, Leon. GPS signals are transmitted on two channels (L1 & L2.) Depending on the use of the GPS, you can get accuracy with about 2" if done correctly. Our aviation GPSs use L1 for movement and the combination (or technically, difference) of L1 & L2 for precision. The calculation is slower, but we use it as a factor in our position estimate calculation from L1. The end result is a system that can have <1' accuracy while moving or <2" while stationary.

    John, the GPS most manufacturers have (over the counter style) are only capable of receiving on L1, which are not precise enough for your needs. There are devices available for receiving L2, however, they will be quite expensive and are mostly special purpose devices, like for surveyors, agriculture and aviation customers.

    Bill
  • John BoardJohn Board Posts: 371
    edited 2012-06-09 17:58
    Would I be able to make some sort of equivalent system - A Tri (or quad) laterating system of beacons - So, using RF determin the distance between the robot and the 4 beacons then use trilateration to figure out where it is. (Over a small scale area).

    -John
  • msrobotsmsrobots Posts: 3,709
    edited 2012-06-10 02:29
    Hi John,

    We have somehow the same questions/goals ...

    I try to figure something like this out also. I noticed that the values of my GPS (VPNxxxx) are changeing even when it sits still.
    So I ordered a second unit to test some differential approach. (one GPS stationary, one moving). The second one arived today so I will test that tomorrow. Hopefully I can get some stable numbers. But I still think the resolution is not as good as you/we wish for.

    I use Xbee modules for communication. They provide some sort of signal-level info if used in API-mode. By now I am not using that but in some recent forum-post I've read about somebody using the signal-level for some sort of orientation.(distance? direction?).

    Maybe mounting the Xbee on a servo like the Ping-Thing and shielding one side to figure out a direction/signal strengh to other Xbees...

    Since I want t o track a couple of things in a hillside area of about 14 acre I play with those long-range-Xbees at 900Mhz? from Sparkfun.

    But Xbees are expensive and I figured out I need a lot of them to halfways cover the area, so I am not shure if I will proceed with them.

    Duane Design - here on the foras - is using some Nordic-Modules. They are not as advanced as the Xbees (them providing alot more then just serial communication), but compared to the Xbees are rotten-cheap (about $2 compared to $20-$80). You even can power(!) them from Prop-IO-pins. I do have no info about range or so yet.

    Since I live in CA, I think about hanging units with Xbees/Nordic-Modules together with a simple prop (or prop+gps) and a solar-panel into those Oak-Trees on the site, Setting up some simple network. They could be used as rf-beacons too. And with a stationary GPS could provide differential values for better accuracy of the GPS-data..

    The things I need to track have some common features - they are independent, groundbased and roaming and they are aware of their energy-level. If this level goes down they try to come 'home'. One type tries to leave the property sometimes, the other type is more ore less undecided about that. And both types tend to stay together, apperantly.

    Short - I want to track my autonomus robots, and this pack of dogs I am living with.

    So - we have somehow the same questions/goals ...

    Enjoy!

    Mike
  • Duane DegnDuane Degn Posts: 10,588
    edited 2012-06-10 07:05
    msrobots wrote: »
    Duane Design - here on the foras - is using some Nordic-Modules. They are not as advanced as the Xbees (them providing alot more then just serial communication), but compared to the Xbees are rotten-cheap (about $2 compared to $20-$80). You even can power(!) them from Prop-IO-pins. I do have no info about range or so yet.

    "Duane Design", that has a nice ring to it.

    I've been working on a better Nordic driver lately. I've found several problems with my previous driver and this new one should be much faster and hopefully easier to use.

    The Nordic modules also include a "Received Power Detector" (RPD). I used to think this type of information wouldn't be useful to determine location, but now I'm not so sure. I've seen several examples on the internet where it looks like this information can be very useful in determining location.

    One of my goals is to create message hopping software for these modules. I'm hoping to use the RPD values to have the various modules figure out the best path from one device to another. I'd really like to make these $2.50 modules (see post #1 of my index) as easy to use as XBees. I think they are faster than XBees. They do require more pins to control them but that's often a trade off you have to make when you want to increase communication speed with a device.

    I think it's possible to power these modules directly from a Prop's IO pin, but most of the modules I've tested wont work that way without a capacitor on its power pin. Since the modules require a cap, I'm thinking they might draw too much current in short bursts and could possibly harm the Prop. I think if one were to add a 100ohm resistor and the necessary capacitor, the Prop could safely power the module from an I/O pin.

    Normally these modules use about 17mA or current in receive mode.

    As I mentioned, I'm working on an improved driver for these modules. I'll make sure and add a way to easily access the RPD information with my object.

    It would be great if these modules could aid in locating your robot as well as allow you to communicate with it.

    BTW, I think it was Xanadoo (Jon) who used the XBee's received power information as a locating aid.
  • msrobotsmsrobots Posts: 3,709
    edited 2012-06-10 20:35
    @Duane - sorry for misspelling your name. Would be a nice company-name...

    hadn't had any time for playing ... so no news about GPS. Two more dogs are coming! 8 weeks old irish wulfhounds. what a blast they are ...

    Enjoy!

    Mike
  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2012-06-10 21:01
    It might be possible to set up your own differential GPS system using two GPS receivers: one on the robot, and another at a known, fixed location. Any non-differential drift between the two would cancel. A problem, however, comes during satellite "handoff" when one loses (or gains) contact with a particular satellite before the other one does. In such a case the position-calculation data between the two will be different, which will lead to differential errors.

    -Phil
  • W9GFOW9GFO Posts: 4,010
    edited 2012-06-10 21:31
    I remember posting about that idea a couple years ago. As I remember it the conclusion was that there are far too many variables for it to work reliably. Each GPS would have to be exactly the same model, chipset, firmware version etc.. In additiona each receiver is going to be experiencing separate signal quality issues.

    Still, might be worth trying - there's gotta be some improvement...
  • John BoardJohn Board Posts: 371
    edited 2012-06-10 23:32
    Thanks for the responses - Let me sum up the problem a bit more,

    I can't work with one fixed GPS and another moving - I need to be able to move the GPS's around.

    I just need to find the position of an object within 3 (or 4) beacons to a degree of accuracy of about 1-5 cm.

    Thanks,

    John
  • max72max72 Posts: 1,155
    edited 2012-06-11 05:45
    I can get 1cm with my "stellar" DGPS.
    The two dual frequency GPS are constantly in contact and the base sends a compressed stream of satellite raw data to the rover. The rover compares them with its own data stream, and tries to calculate an accurate relative position.
    Errors come from multipath (you have both direct and indirect signal) and other sources, so there is an insane amount of calculations to be performed, and even then this precision is not so stable (sometimes it is lost, you have to stop and wait for fixing and so on) even in open field.

    Errors for a single GPS are mainly related to atmosphere crossing, so two similar units not too far away will reasonably have similar error sources, so there might be a relative positioning improvement, but not so accurate.
    I agree, worth a try.

    Massimo
  • Heater.Heater. Posts: 21,230
    edited 2012-06-11 06:06
    John Board,

    I love the way you put this so casually:

    "I just need to find the position of an object within 3 (or 4) beacons to a degree of accuracy of about 1-5 cm."

    Given that GPS is global and the circumference of the earth is about 40000Km so you are asking for an accuracy of 1 part in 4 billion or 2.5e-08 percent !!!

    Now, given the size of a robot or dog, especially irish wulfhounds, I'm sure you can relax that requirement a bit:)
  • msrobotsmsrobots Posts: 3,709
    edited 2012-06-11 23:53
    John,

    I appologise for hijacking your thread.

    Heater - the OP needed a resolution of 1-2 cm. For my irish wulfhoundsI am fine with less. - except I would like to measure how much they grow each hour or so...

    Enjoy!

    Mike
Sign In or Register to comment.