Comment by s_gourichon | Hacker News Reader

Comment by s_gourichon | Hacker News Reader

Parent Full thread

> In between, mapmakers figure, the error is swamped by the imprecision of mapping and GPS equipment. Future maps may be updated at a rate closer to real-time. “We have the technology now with GPS to be able to make those slight adjustments on a more frequent basis,” Craun said.

A French startup operates an impressive technical solution in this direction:

> new GNSS (Global Navigation Satellite Systems) augmentation services based on the PPP-CNES technology (Precise Point Positioning).

> Absolute positioning accuracy of 4 cm (2D-95%), everywhere in the world, on the ground, on the sea and in the air, all the time, with only one receiver, without any nearby GNSS reference station: Precision, Availability and Integrity

> Depending on the level of service subscribed: Absolute positioning precision of 2, 4, 10, 50 or 80 cm (2D-95%)

> Compatible with GPS, GLONASS, BEIDOU and GALILEO constellations and mono and multi-frequencies receivers

I attended a meetup were they explained part of it. They have models for each of the aspects that can deteriorate precision and update all the parameters of all models in real time based on continuous measurements.

At a point during the development of the solution they had an unexplained discrepancy and tracked it down to some satellites of a common network being classifiable into two clusters of two different latencies (like, in the nanosecond range or even smaller). They contacted the network owners, listing the two clusters and asked "what is the difference between these and those satellites?".

The owners were a little confused. Something like "Well, inside each satellite, the cable from the antenna to the processing unit is a high performance transmission cable of tightly controlled specification, one inside each satellite. They happened to have been supplied in two batches with supposedly strictly identical behaviour up to the precision available at the time. And you sent us the exact listing of satellites matching batches A and B. But how on earth have you got this list? Were you spying into our internal archives?"

Answer: "no, we just have a model so precise that we could calibrate difference in transmission delays in your satellites from our receptor network on the ground".

Here's the startup Web site : http://www.geoflex.fr/?lang=en

4cm accuracy is almost enough for precision agriculture. You usually want a year-over-year positioning accuracy of better than 2.5cm for your combine harvesters to mange issues of soil compactification. This is about the same order of magnitude as continental drift speeds, which can be up to 10cm/year. The usual solution to this is to use a correction signal generated by a nearby reference station with a well known position that can be distributed over the internet: https://en.wikipedia.org/wiki/Real-time_kinematic

Compared with this, the new approach you describe may have the disadvantage (at least for applications that care about positions relative to the surface and not the core of the earth) that it does not use a reference station that is fixed to the continental plate it is on (so that continental drift cancels out implicitly), and users have to handle it explicitly. Or do they offer a drift corrected service?

That's similar performance to NASA's Gipsy solver and a couple other PPP solutions.

The practical problem is PPP falls apart when there's any sort of multipath or noise, with will be the case for any receiver not in a completely open field. A single nearby tree is often the difference between 4cm accuracy and a meter.

Trimble has a similar service, called PP-RTX.[1] There are various online PPP services that can give up to CM accuracy in post-processing, but near-real-time CM accuracy is still a new (and unexpected, to me) thing.

As for updates, the NGS have plate models that allow you to get the shift in coordinates at any given location[2] between two moments in time. Those models are in large part based on GNSS signal recorders called continuously operating reference stations[3] (CORS) which give actual movement at locations throughout the states.

For example, Bar Harbor, Maine, is moving at -0.0013 m/y north, 0.0020 m/y east, and -0.0009 m/y in elevation relative to the North American plate, and 0.0071 m/y north, -0.0153 m/y east, and 0.0001 m/y in elevation relative to the world[4].

The NGS is a remarkably good agency. See https://www.ngs.noaa.gov/TOOLS/ for more of their work.

[1] https://trimblertx.com/ [2] https://www.ngs.noaa.gov/TOOLS/Htdp/Htdp.shtml [3] https://www.ngs.noaa.gov/CORS/ [4] https://www.ngs.noaa.gov/cgi-cors/CorsSidebarSelect.prl?site...

That's incredible technology - I wasn't aware it was even theoretically possible to obtain such a resolution with data available to civilians.

But what are some use cases that would benefit from that level of service? For example the highest tier is a 4 cm resolution with a convergence interval of 5 minutes.

The anecdote was told by a speaker of GeoFlex during this meetup: https://www.meetup.com/fr-FR/MeetupVoitureConnecteeAutonome/... though I cannot see it on the published slides.

The slides of the meetup are on https://fr.slideshare.net/LaurentDunys/meetup-voiture-connec... . Geoflex slides are in English, pages 22-53.

How well can it do when travelling through a city with tall buildings all around?

Reply on news.ycombinator.com