bartoszkledzik / ardupilot-mega

Automatically exported from code.google.com/p/ardupilot-mega
0 stars 0 forks source link

Implement Dead Reckoning code to account for GPS lag #245

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago
Implement dead reckoning algorithm to account for lag in GPS vector information 
using UDB code as discussed here:

http://diydrones.com/profiles/blog/show?id=705844%3ABlogPost%3A232352&commentId=
705844%3AComment%3A232212&xg_source=activity

Original issue reported on code.google.com by analogue...@gmail.com on 6 Dec 2010 at 4:17

GoogleCodeExporter commented 8 years ago

Original comment by analogue...@gmail.com on 6 Dec 2010 at 4:17

GoogleCodeExporter commented 8 years ago

Original comment by dewei...@gmail.com on 10 Jan 2011 at 4:34

GoogleCodeExporter commented 8 years ago
I am not sure this is worth the hassle.

Basing this on the what information I can find that on average the GPS fixes 
arrive (on average) timestamped 0.25 seconds after a NTP type syncronized UTC 
clock. This is a fairly constant, and smallest lag for the lat/long type GPS 
replies. It would be easy to test this is true.

In 0.25 seconds, a plane especially one moving at high speed, is not going to 
significantly change its direction: so dead reckoning based on IMU forces are 
pretty useless (unless you want to dead reckon a crash!). As I understand that 
code, it is adjusting GPS fixes from G-forces since the fix?

Secondly, in 0.25 seconds this hypothetical 200kmh autopilot RC jet has 
travelled 55 meters, in a quarter of a second, it has travelled ~15 meters, in 
almost exactly a straight line.

The GPS position is used for calculating of desired bearing and distance to 
desired waypoint. Neither of these calculations could be much influenced by a 
15 meter inaccuracy in position, especially not for a plane flying at 200kmh 
that would neither be required or able to do pinpoint waypoint or track 
following.

Perhaps I misunderstand something critical here, but I don't predict any 
visible improvement in autopilot skills from adjusting GPS for lag.

Now if GPS is marginal, if fixes fade in and out for seconds at a time, or drop 
significantly in quality based on the number of satellites metric, then dead 
reckoning is useful. Typically though I think such a state of affairs is 
unusual for a plane where the sky is entirely visible?

Original comment by justinbe...@gmail.com on 6 Sep 2011 at 1:25

GoogleCodeExporter commented 8 years ago
I think this will be a great addition. Especially because I don't trust GPS 
data that much. Has anyone tried estimating the position with a Kalman filter?

www.hardiepienaar.co.za

Original comment by HARDIEPI...@gmail.com on 26 Sep 2012 at 12:56