I was thinking about a robot localization akin to a DIY DGPS. If I had two identical GPS units [like this running in reasonable proximity (say 100 meters), would the error/drift be the same on both devices? I was thinking one receiver would live on a robot. The other would be mounted on a window near my home computer, which has a wireless link to the robot. By monitoring the quantifiable error on the fixed unit, the robot can hopefully zero out the error/drift coming from the mobile unit.
Does anyone have a pair of identical receivers that they can log for 20 minutes, simultaneously (motionless of course)? I’ll even do the data crunching!
After giving it some thought I’ve come to the “conclusion” that they’ll probably differ because the american military “obfuscates” the signal (more precise, adds noise to the timing) a little bit in order to reduce civil-use precision.
lehmanna:
…the american military “obfuscates” the signal (more precise, adds noise to the timing) a little bit in order to reduce civil-use precision. Admittedly, this info may be far outdated…
It is a bit outdated. The deliberate signal error was called Selective Availability (SA) and was shut off about 10 years ago. The argument was that it was hampering civilian navigation with no real defensive benefit, and they could always turn it back on in a crisis.
However, I don’t believe SA is of any concern. DGPS was invented (by civilians) specifically to deal with SA in small regions where precise navigation was needed (crowded harbors, narrow canals, etc.). WAAS does a similar function on a wider scale (though it too is controlled by the US military). The point being that all receivers within a small region are subject to the same signal error, whether artificial or natural.
“Noise” in an individual receiver’s sensing is, I suspect, the most likely cause for side-by-side units to disagree on location, if in fact they do.
Yes, two receivers within a certain area will measure almost the same position, IF they are locked on the exact same set of satellites. If using consumer GPS-receivers there is however no way to know which satellites a a certain position calculation has used.
So, the final answer is no, unless you have “professional” receivers which can give you the raw data from the satellites.
Both receivers must be exposed to the same conditions. Having one in the house and one outside will not work. They will receive potentially different GPS signals, and be will exposed to different multipath induced error.
In general, this will not work well if your stationary and mobile units do not have an unobstructed view of the sky, and special [read expensive] high mounted [eg at least 8 feet off the ground] antennas designed to mitigate multipath errors.
A better solution for your robot is combining GPS with a dead-reckoning navigation device E.g. 6 DOF system using accelerometers and gyros.
A better solution for your robot is combining GPS with a dead-reckoning navigation device E.g. 6 DOF system using accelerometers and gyros.
That's essentially what I had planned, but it's still subject to the weaknesses of the GPS for starting conditions. Given the expense and complexity of marrying all those sensors together with a kalman filter, the mini-DGPS idea seemed a lot simpler...at first. :( I may explore the idea of using active beacons. That's an approach I can get my head around at least.
I’m still curious to see the difference in error for a pair of identical GPS receivers though.
Yep pretty sure most of the error in commercial GPS receivers is from multipath.
Maybe you could have two mounted to your mobile device, cleverly arranged and with blocking “shields” so that each receiver was unlikely to pick up the same reflected signals as the other.
I posted elsewhere, but I ended up doing the experiment with three identical Wintec GPS units via bluetooth, right next to each other, powered on at the same time (more than once, after getting a lock).
They indicated similar satellite visibility, etc. but they ended up having different errors as they said they were apart from each other by 2 or so meters. (I didn’t spend a lot of time analyzing, just checked if they were going to be off by a microdegree or two - the last digits for lat/lon should have been nearly the same but weren’t).
It uses a ground based reference if I understand right but I think it could be adapted.
The raw data usually includes things like the carrier phase. Comparing the post-processed data won’t work, however the raw measurements might be easier and allow differential measurements (instead of distance from satellite, unit 1 is X closer or father to it than unit 2).
“Pseudoranges” are the calculated distances to each individual satellite. Most consumer GPS units spit out a lat/long position that has been calculated from these pseudoranges but not the pseudoranges themselves. At that point, any errors caused by the atmosphere or the signal being reflected off of buildings (multipath effects) are hidden in the position you get. These effects are very localized, so comparing coordinates isn’t effective.
If you had access to the pseudoranges themselves from both your stationary and mobile units, then you could theoretically calculate an offset a la DGPS. But the algorithms for processing these signals are complicated and usually built into units that already support DGPS.
I wish there were a proven open source method for getting better GPS positioning. Like I said, I hope RTKLIB proves to be just that…
