Centimeter GPS accuracy

I just noticed the Indiegogo campaign for the Reach GPS receiver at www.indiegogo.com/projects/reach-first-affordable-rtk-gps-receiver which promises a small GPS receiver with cm level accuracy. I think it gives a view of what will be possible with GPS in the future.

The device is small, cheap, and uses open-source software and an ublox chip that can record at 18 Hz. Nice start. But the really interesting part is that it uses RTK GPS. I read a little about this, and did not really understand most of it. But it seems that the potential precision is in the millimeter range because the a more than 1000-fold higher frequency is used (compared to the normal GPS units) to get rid of atmospheric distortions. This also requires the use of two receivers - a stationary one that acts as reference, and the normal moving one. However, the stationary one can be used to many moving receivers. What's more, the cost for two receivers (with enough CPU power to do the calculations) is less than $500.

The "RT" part in RTK GPS stands for "real-time". This requires that the two units are connected by radio, Wi-Fi, cellular, or similar (the stationary unit can basically be connected to a web server). That might be impractical for windsurfing. However, it seems that post-processing is also possible, so the corrections could be done later at home.

The technology promises absolute position accuracy of 2 cm. I think that requires that the position of the stationary unit is known to this level of precision. However, I think the relative position accuracy should always be in this range (assuming a good GPS signal, etc). I think at this level of accuracy, the positional speed data could actually be more accurate than doppler speed data.

The units will probably be first available for drones. The developers (emlid.com/) are already selling an autopilot chip, the Navio+, for drones. that includes a ublox GPS for less than $200. The chip includes not just a ublox-based GPS, but also a bunch of other things the autopilot needs, including wi-fi and servo outputs.

Any geeks out there looking for a project? The Reach receiver should be available within a month or two, maybe even before the new Locosys becomes available. You'll have to hook up your own screen etc, though. But you do get a "9DOF IMU" with 3-axis accelerometer, so you can spend many windless days in front of a computer, trying to figure out how to use the acceleration data to further improve speed accuracy .

Essentially, as I understand it, Carrier wave RTK with differential correction is what they Survey Grade GPS's use. The Trimble units that are used by Macquarie Innovation and Hydroptre are based on this system and technology.

I know of other projects that are developing RTK GPS devices that could be quite inexpensive, practical and affordable for GPS Windsurfing, and possibly work easily with other current hardware and devices for levels of accuracy that are order of magnitudes better than the current best we use.

It is my understanding that such devices can produce Doppler speed data with incredible accuracy as well. The advantage of Doppler may still remain as it is less susceptible to atmospheric interference.

It is amazing that despite our small market to drive specialist GPS development, we may well have what is now almost unimaginable accuracy available to us in the not too distant future at a very modest cost.

It'll be interesting to see how it develops.

As roo said, without the second GPS providing corrections off a known point, the gps that surveyors use really aren't mucho better than a gt31, accuracy wise. They've just got a lot more fuctions....

As I understand it, survey gps already use real time too

Select to expand quote

Subsonic said..
It'll be interesting to see how it develops.

As roo said, without the second GPS providing corrections off a known point, the gps that surveyors use really aren't mucho better than a gt31, accuracy wise. They've just got a lot more fuctions....

As I understand it, survey gps already use real time too

Difference is price. The Commodore PET was a few thousand dollars, the Sinclair ZX with similar power about $100 or $200. Neither was even 1/1000th as "smart" as any $60 smartphone you may use for GPSLogIt.

The stationary GPS is no big deal. For speed, we only need relative accuracy. Leave one of the two units you can buy for less than $500 on the beach, take the other one sailing. Don't even worry about the "real time" part - just post-process the data. You can still get feedback that's at least as good as the GPSLogIt or GT-31 data on the water just from the ublox GPS, even without the carrier-phase magic.

Survey grade GPS are capable of 1cm and have been for quite some time*. Initially, some may have needed 15+ min's of geostationary time to resolve to that level, but now I think they can get it in real time. When Macquarie Innovation was using the Trimble system back in 2007-8, they were claiming better than 10cm accuracy in Dynamic mode with differential correction at 10hz.

The figure of 1.9mm quoted for RTK is the theoretical possible accuracy. I don't think anyone is claiming they have actually attained that level yet, at least not in a practical application.

Current (in the last couple of years) RTK developments using just L1 and Differential correction in real time from a dedicated second unit in a known location nearby, can do 1cm now, or very close to it, in real time. This capability has already been demonstrated using consumer grade GPS chips and relatively small antenna. With the developments underway, it is just a matter of time before solutions become available at very affordable cost that we can use.

After looking into it some more, I agree with Boardsurfr that centimetre level positional accuracy (with accuracy error data), looks like it will far surpass what is currently possible with Doppler speed data. At this very moment though, and probably for the next couple of years, the Ublox Doppler data at 10hz is about the best we can get access to.

However, it is worth noting that Survey grade GPS's have gone from around $60,000 in the early 2000's, to under $7,000* now, and even to $320!** And the models available have increased a lot.

For about $7,500 you can even have this hand held beauty and send data to your smartphone or tablet! www.gpsworld.com/gnss-systemaugmentation-assistancenewsgeneq-introduces-palm-sized-gpsglonass-rtk-receiver-real-ti/

* gpsworld.com/centimeter-level-rtk-accuracy-more-and-more-available-for-less-and-less/

** gpsworld.com/hemisphere-gps-announces-rtk-network-compatible-s320-survey-system/

This website, gpsworld, is a great source of interesting info.

Here is a bit about an NMEA dongle that can connect to a GPS source, like a Survey grade GPS, and transmit the data via bluetooth to a IOS iPad or iPhone, over riding the internal IOS GPS and providing the IOS apps with extremely accurate locational data.

The interesting comment is this:

"Garg explained the need for the dongle on his Kickstarter page: "The accuracy and precision of the internal GPS on iPads and iPhones is highly unreliable - it works fine for navigational purposes but fails miserably for mapping and asset management applications. The accuracy varies in range from a few meters to a few hundred meters depending on operating conditions, and there is no easy way to reliably ascertain that. Tests have proven that the accuracy rating on the location data returned by Apple is more of a general estimate than a reliable metric."

You can see two important points here:

1. Accuracy not that great.
2. Can't quantify accuracy.

That is the same problem with any phone I'd say.

www.gpsworld.com/kickstarter-launched-for-nmea-dongle-to-connect-gps-ios/

Like this little tidbit:

www.gpsworld.com/gnss-the-new-gps/

China has launched its own GNSS system and it now has the most satellites in view in the sky to use (with GPS and Glosnass), more than Europe and North America. More sats used = better accuracy and reliability.

But guess what?!

Australia is covered by those Chinese sats as well! That means when, or if we have GNSS devices that can use China's BeiDou system as well as GPS and Glosnass, we will have potentially greater accuracy as well! Many of you will have noticed that most phone now use both the GPS and Glosnass satellite constellations. I don't think it will be long before they use the Chinese BeiDou system as well!

The S320 retails for about $7000 (more in Australia, of course). The Piksi RTK system is available at store.swiftnav.com and costs $1000 for two receivers.

But there has been at least one report that "RTK GPS for Under $200 Works!" at www.isoblue.org/posts/2014/04/RTK-GPS-Under-00-Works/

Opps! I blame may ageing eyes! "Should have gone to Spec Savers!" lol!

You know I actually went back to check what I thought was the price because it sounded to low to be true, but I still saw it as $320 instead of S320.

But you are quite right. These things will get a lot cheaper in the near future and the high accuracy technology will merge into consumer priced devices costing a few hundred dollars.

I think this guy is correct, and this was way back in 2010, long before the idea of low cost RTK took off:

In response to this claim: "Consumer-grade GPS receivers and survey-grade GPS receivers are not the same quality and never will be."

He says:

"This isn't the case according to several GNSS receiver designers I've spoken to. There's no reason a "consumer-grade" L1/L5 GPS receiver can't achieve cm-level precision (horizontally and vertically) with a good quality antenna. Of course, it will need a source of correction, but in the 5-10 year window, RTK corrections will be more available and less expensive than they are today. The RTK corrections will likely be free so the only expense will be the wireless data plan.

There will be dozens, maybe hundreds of GPS chipsets designed for L1/L5. Many will be open systems where companies will be able to load their own firmware into the receiver to add specific features (e.g., more robust ambiguity resolution for surveying). The baseline L1/L5 GPS receiver may be only a few hundred dollars, but a customized version for specific applications will have a premium of a few hundred, or maybe $1,000+. For example, a GPS L1/L5 receiver that also includes Galileo signal and is customized for machine control with specific features might be $1,500. Commercial users will pay that premium. They already justify paying tens of thousands of dollars for the same performance today. The difference is that a low price point will attract a much larger audience.

There's no doubt that there will be boutique, niche GPS L1/L5 receivers that will be able to garner a premium price, but surely the days of many thousands of dollars for high-precision GPS receiver are heading to an end."

Extract from:

gpsworld.com/surveynewsfollow-whats-going-happen-when-gps-accuracy-cheap-10082/

You're probably right about the 1 cm precision; typical real-world precision claims are 2 cm to 8 cm. Although your "not in my life time" reminds me a lot of what people said early in the Human Genome Project. Several articles come to mind where the authors painstakingly explained by it would be impossible to exceed the current performance by much, claiming physical limits had been reached or similar. They were proven wrong within a few years, as others discovered how to get around the limits.

In the United States, many states already have public base station networks, where the necessary reference data can be downloaded for free*. The state I live in, Massachusetts, has base stations spaced about 50 km apart**. Two of these are within less than 10 km of a speedsurfing location, and thus close enough. They state "the MaCORS Network Solution typically achieves an RTK accuracy of 1-2 cm horizontally and 2-3 cm vertically."

Your comments regarding simplicity are certainly valid. There are several members in our GPSTC team who find posting data quite challenging. Some of them like the ka72.com site, which simplifies things for them. Setting up a similar web service that analyzes raw GPS data is not much harder. The service could either load the reference data from the web automatically (if available), or require the upload of a second file from a reference station. So maybe that's twice as hard.

Regarding the accuracy of doppler data, I usually have a hard time finding anything that contains actual data, other than Tom Chalko's original paper. Sure, spikes are much less common, and having accuracy estimates is great. But what is the actual accuracy?

For GT-31 data, typical errors (given by GPSResults) for single data point in speed around 30 knots are +-0.5 to 1 knots. 10 second run errors are typically +- 0.3-0.5 knots. That's 1-2% accuracy.

For 5 Hz ublox data, I get errors of +- 0.3-0.6 knots at speeds around 30 knots. That's 1-2%. That goes down to +- 0.15 knots when looking at 10 second runs. 0.5% accuracy.

If we had positional data with an accuracy of +-3 cm (relative to the base station, not absolute), we'd also have an accuracy of +-1-2% for 1 second data points, since 30 knots is about 3 m/sec. But the error model for average speed changes! To get Doppler speed for 10 seconds, we average the individual speeds. By assuming that errors are random, we get lower errors when looking at more data points - either longer runs, or higher frequency data. Of course, if there is a systematic error in the doppler speeds, our accuracy estimates will be incorrect.

For positional data, however, the errors do not accumulate. If any point has a precision of +- 3m cm, the expected error for two points that are 10 seconds apart is +-6 cm or less. 6 cm of 30 meters is 0.2%. That's better than 5 Hz doppler data.

Note that for 500 m record runs, using positional data is actually conceptually more similar to gate timing than using Doppler speeds. Again assuming 3 cm errors, but allowing for 3 sigma on both ends, the observed time different to cross a distance of 500.18 m would give us a reliable estimate of the minimum achieved speed. Adding the 18 cm "safety" would add an expected error of 0.036%. For comparison, the +- error margins I get in GPSResults for 500 m runs are around 0.8% for GT-31 data, and 0.3% for 5 Hz ubox data. So positional data could actually be substantially more accurate, even at +- 3 cm.


* www.gpsworld.com/finally-a-list-of-public-rtk-base-stations-in-the-u-s/
** macors.massdot.state.ma.us/spiderweb/frmIndex.aspx

How long do you plan to live Roo?

Actually, it is here right now. 1cm positional accuracy in real time in a single, stand alone device.

A device receiving signals from many satellites plus a base station without the need for the user to have any other device is no different from what we do now. What is the difference if one of the many signals a GPS receives comes from a free, land based station or even over a mobile phone connection from the internet?

The WSSRC accepts the Trimble GPS units for their records, and, at least when they were first approved, they were only 8-10cm claimed accuracy. It is possible to get that now from other RTK devices that don't use a signal from a land based station, just from satellites.

Unfortunately for us, all that still still costs 7,000 dollars plus, but it will not for too long.

So the only real question is, will the stand alone GPS with 1cm real time precision become affordable for windsurfing in your lifetime?

I certainly hope it will because I hope to see you still around here for at least a few more years.

I agree entirely about simplicity. But of course, simplicity is just related to the user experience. For user simplicity it is just a matter if having clever designers who can mask the underlying ultra complexity from the user so as to provide him with a solution at the press of a few buttons. As you mention, we already have that available if you so desire. It's also easy to conceptualise systems which have even more complexity under the skin, but still with the same simple interface that anyone who can program a VCR can use. (Or use apps on a mobile phone? )

But of course, all this does not change the current reality that 5hz and 10hz Doppler based devices are the best we can do for speed precision at the moment for a few hundred dollars.

I had to search around a bit to find confirmation of my memory of conversations with Tim Daddo from the MI team.

This is from: www.trimble.com/technologyandmore/i3-2010/Download/TM%2016HQWEB_.pdf

the relevant quote is:

According to the International Sailing Federation World Sailing Speed Record Council (the governing body of the WSSRC) Rules 2010, the GPS survey equipment receiver logging rates must be set at 10 hertz, which results in a GPS position every 1/10th of a second. The GPS must also provide a time stamp with each position accurate within .001 second. The data needs to be so precise that a horizontal position error (hpe) of greater than 0.10 meter (one sigma) will not be accepted.

Note that no other GPS devices had been approved for use by the WSSRC at the time this article was written. I suggest this is mainly because someone would have to put a lot of effort into the validation of the accuracy in a form that the WSSRC would accept. It's probably a lot easier to just use the device that already has approval. I am pretty sure there would be many other GPS device around now that could perform at least as well, if not quite a bit better than the Trimble device approved over 10 years ago.