Flutterby™! : status update 2011-01-20 23:36:10.869477+00

status update

2011-01-20 23:36:10.869477+00 by Dan Lyke 11 comments

Reading up on GPS accuracy. Anyone got favorite whitepapers on RTK implementations or other dGPS technologies?

And I had to double-check to make sure I wasn't looking through my spam inbox...

Shadow, check out Trimble's agricultural guidance systems. And I don't think you need edge markers, I have a picture somewhere around here extracted from a PowerPoint which shows circles of sub inch and inch or so coverage around beacons from a commercial provider, it may actually be Trimble, covering a good portion of northwestern Ohio.

NOAA CORS will give you sub-centimeter precision in many areas of the country, if you don't mind 30 day latencies...

The thing about a field with a tractor or combine in it is that you don't have to adjust to avoid potholes or tire fragments or the ladder that just fell off that pickup truck, or worry about the repaving that's in progress...

If you have some control over the terrain, you can increase the alignment between the map and the terrain. This is necessary for farming applications. It's also possible in highways, especially where variation is controlled in things like lane markers. For a farm, if you plant some kind of beacon in the corners of the fields, then you can correct GPS quite finely.

And I would have thought that NOAA or USDA would have setup public RTK stations. That would be a farm subsidy worth being taxed for!

At the far end of the location puzzle are systems like Ubisense which work wonders in very closely controlled environments (i.e., inside warehouses and production facilities).

I would say a system of VOR/DME beacons planted along some grid points in his field would be the best way for a farmer to set up for an autonomous tractor.

So what's a good commercial solution for the "farmer" type set-up mentioned by Dan above?

I think you might be surprised on the economic room that is left. After my previous comment about "8 miles from their barns" I found that there are private companies broadcasting differential GPS corrections on a good portion of the U.S. midwest, allowing correction with low enough latency to offer inch or so accuracy to tractors with guidance systems (depending on the distance from the broadcaster), but large regions aren't covered.

Personally, I'm of the opinion that the map is not the terrain, and that local navigation systems which key off of local features (ie: lane markers) are necessary to automated vehicle behavior. On the other hand, there are all of these robotic tractors...

I'm not sure there's much economic room left in milking greater accuracy out of GPS signals (unless you're working for someone like Trimble but their engineers wouldn't be starting out as raw as you are - they would have a PhD or so in geodesy, aerospace or computer engineering).

What you can get out of a $75 Personal NAV device is pretty close to state of the art when it comes to vehicle tracking. If you really need something more accurate, the next step is the kind of devices TeleAtlas and NavTeq use in their survey vehicles. Beyond that, car-jack one of Google's self-driving cars and reverse engineer it.

But you'll probably find across all of these platforms is some kind of "imperfect information from GPS correlated to a more accurately surveyed dataset". Which means "back to the Garmin Nuvi".

#Comment Re: made: 2011-01-21 04:51:02.905333+00 by: Dan Lyke [edit history]

Yeah, I'm looking at Gold code and ionosphere propagation issues as measured by the difference between L1, L2 and L5 signals and...

So, yeah, I now know way the hell too much about GPS, and it's only going to get worse.

Wow... You've looked a lot closer at GPS than I ever have. Being a "cartographer", I really don't care how my data is actually collected. ;)

What you've figured out is pretty much my general understanding. GPS isn't terribly accurate but can be made accurate within the bounds of certain applications. This is why the DARPA Grand Challenge wasn't a matter of just hooking a GPS up to a remote control car. It's also why the military uses laser guidance and video guidance rather than just GPS. Almost used to count in horse shoes, hand grenades and bombs but I think it's only acceptable in horse shoes now.

In general, GPS gets more accurate the longer it sits in one place. For moving objects, you have to work with the error. Your Garmin nav device automatically sticks you on the road network and uses a digital compass to put you in the correct lane. And the "more accurate the longer it sits in one place" also kind of works if you stay in the same lane. Your nav device does a best fit line through the GPS points.

To explain the CORS X,Y,Z, DGPS works is by storing every raw satellite signal at the base station and calculating a diff for each coordinate (x,y,z) for each satellite packet received. I'm betting the coordinates are in some reference system common to GPS satellites. That is, the reference system the satellites use to report their position.

If you really want to learn the details of that coordinate system, I'd grab the source code from GPSbabel and start looking there.

#Comment Re: made: 2011-01-21 01:03:10.865831+00 by: Dan Lyke [edit history]

I've been asked to write an intro to error and accuracy in GPS, especially as it applies to real-time vehicle position information.

Since I asked this question I've discovered a whole lot of "right hand doesn't know what the left hand is doing" or, perhaps more accurately: the FAA cares about real-time position in the fifty or hundred foot range, but only over 200 feet AGL, the Coast Guard cares about real-time position somewhat more accurately, but mostly only near the coasts and at sea level for slow moving vehicles, NOAA cares about centimeter level accuracy but doesn't care if they only figure that out a month later (and they care less about that accuracy the further you get from a fault line), and large scale farmers care about inch accuracy in real-time for vehicle control, but only within a mile or 8 of their barns.

And that some of how NOAA describes the CORS stations working seems inconsistent with how Trimble describes their agricultural products base stations working.

And some of these applications use RTCM_SC-104, either via RS232 at 4800 baud or even as static files, but then there's also RINEX or "o" or "n" or "sp3" or "Met" files, and, and when this stuff is done in real time it's either sent in the 300KHz range or 450MHz or 900MHz, and probably also at 2.4GHz and in all sorts of other formats.

And now that I have this base level of understanding I have to go look at how various receivers deal with this, most likely at the chipset level.

While I have your attention: In the CORS data files, I'm running across specs like:

This is from a site in the KETCHIKAN A-5 USGS quad. Is that XYZ coordinate likely in the mythical GPS space that I've heard about but never actually seen in the wild? It's gotta be that, because 5 million meters would make no sense as altitude in whatever the projection the USGS quad is.

Differential GPS and, by extension, RTK, only correct for systemic errors in the GPS signal. I don't have any particular references. What's the application?