0

Rather than accurate latitude/longitude/elevation data, I need accurate change-in-latitude/longitude/altitude data. If I use the same satellites over a 10 minute time period, then I expect that two computed positions, p1 and p2, will have somewhat consistent error vectors, e1 and e2, such that (p2+e2) - (p1+e1) ≈ p2-p1, i.e. e1 ≈ e2. What level of accuracy can I expect in such (p2+e2) - (p1+e1) data?

popham
  • 101
  • 1
  • Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. – Community Nov 22 '23 at 17:58

1 Answers1

0

I don't think that using the same set of satellites would give an approximately fixed position, even if you are operating in nice open-sky conditions.

Keep in mind that most GNSS satellites are travelling at high speeds relative to the ground (around 11000 km/h for GPS satellites, but some Beidou satellites are Geostationary) which means that the calculation is changing rapidly even if you're using the same satellites and staying in the same position on the ground.

There's also been significant changes in the positioning algorithms - GPS receivers used to use methods like weighted least squares, where every position was calculated independently and positions jumped around almost at random. Nowadays, most GNSS receivers appear to use a Kalman Filter where there is a dependence on previous states, but there's huge opportunities to change parameters which change the behaviour to make positioning more suitable for pedestrian use or for vehicle use etc.

Also take into account that smartphones typically use the cellular radio antennas for GNSS, and these follow the general antenna rule that smaller is worse.

Trams
  • 749
  • 3
  • 13
  • I saw that GPS satellites have an orbital period of 12 hours, so 10 minutes time correlates to 5 degrees of arc. Reading up on DGPS, I saw that improved accuracy back in the 15m accuracy days involved adjusting for atmospheric interference, satellite location inaccuracy, etc. Extrapolating, I anticipated that the 5 degree satellite position shifts would preserve many of these errors over the short timeframe. Is today's 5m accuracy already nullifying the sources of error that I anticipate nullifying with the difference computation? – popham Nov 22 '23 at 21:33
  • One of the graphs at https://gis.stackexchange.com/questions/25798/how-much-sense-does-it-make-to-average-lat-lon-samples-in-order-to-increase-2d?rq=1 demonstrates the smooth drift that looks compatible with my thinking. – popham Nov 22 '23 at 21:36
  • I suggest you capture some data using GPSTest by BarbeauDev from Google Play Store (assuming you're using Android, no suggestions if you're an iPhone user) then look at it in Google Earth or similar to see how it behaves. I suspect that the smooth drift shown in the other topic might look just the same as walking from one point to another, how do you plan to tell the difference between drift and actual movement? – Trams Nov 23 '23 at 01:35