Investigate what default should be used for SBAS - NONE or AUTO

iforce2d commented 8 years ago

Attempting to use ground assistance in regions with poor SBAS coverage can cause many satellites to be ignored (all GLONASS sats and even many GPS sats too). In the worst cases, this can cause GPS fix to be abandoned completely.

http://www.rcgroups.com/forums/showthread.php?t=2495732&page=384#post34725614

Currently there is no option in the configurator to disable the ground assistance type used by the GPS receiver. It's either auto, or a specific SBAS region.

By doing set gps_auto_config=OFF in the CLI, it is possible to skip all GPS module settings, but this is not ideal for two reasons:

Skipping GPS config does not disable SBAS, it just does not enable it (ie. if it was already enabled it will remain enabled)
Skipping GPS config also skips all the setup necessary to use UBX protocol for most modules which use NMEA by default.

I think it would be great to have a 'Disable' option in the 'Ground assistance type' combobox in the configurator.

DzikuVx commented 8 years ago

Did some testing with 1.1 and SBAS Auto. Sat count 18-19, most of the time 19. No jumps, no losses, solid sat cout and pretty stable HDOP.

theArchLadder commented 8 years ago

I have never really seen any position drift with SBAS=EGNOS either. For now i have been using SBAS=NONE, and i have been surprised a few times over how much the home position have drifted, like several meters.

I will try to switch between EGNOS and NONE every other day or so, to try to get a feel for how much my home position drifts.

iforce2d commented 8 years ago

I would vote to have NONE as the default. Usually I don't like to mess with defaults because backward compatibility is a nice thing, but it's a question of having a setting that works well for 100% of users, versus one that works very well for 90%+ of users and can be dangerous for the other 10%. These percentages I just made up, I suppose there may actually be more than 90% of users with good SBAS coverage, but I think the principle is still valid.

Even when SBAS does work well, the advantages for the typical user don't seem that huge. Some situations I can think of where having a good absolute positioning would be helpful:

I want to send my quad to hover within a few meters of a building/tree etc, just by using mission planner and giving no manual input while flying
I want my quad to land autonomously at a remote location where I will not be able to observe or control it while it lands
I want the quad to go to an exact position and drop something accurately, and I will not be able to adjust the position before dropping

So in general, SBAS helps in situations where I want to position the quad using only the map, and then give no manual input or adjustment, and I cannot tolerate more than a few meters of error. Quite frankly I don't see myself ever trusting my quad that much, no matter how good iNav gets :) And we should also keep in mind that the google maps we're using to specify the absolute locations may not be that accurate in the first place.

I would expect that most users will be flying in fairly open areas where their landing area has at least some meters of leeway for RTH, and they don't need to hover in close proximity to buildings/trees, and they will be present to observe the craft and make adjustments manually anyway. I would also think that having mission waypoints a few meters off the true absolute position is not a huge deal.

From what I've read about SBAS, the real strength of it will be seen if the flight starts and ends in different locations (so the measured home location cannot be used) and it spans hundreds of kilometers and many hours. In those cases the local conditions (atmospheric/ionospheric) at the start and end points can be quite different, compared to our flights which start and end in pretty the same conditions, and finish in less than an hour.

digitalentity commented 8 years ago

Well written! I totally agree with you.

theArchLadder commented 8 years ago

@iforce2d "can be dangerous", do you mean that sudden 2-3m shifts you talked about earlier?

Regarding that 90% vs 10%, it looks like 10% can have sudden shifts in position, and the 90% could maybe have improved position accuracy, but we don't really know that. I have been suprised with the home-shift with SBAS=NONE, but that could have been a coincidence. More testing needed here.

So yeah, i guess with the data we have now, i would agree with you that SBAS should default to NONE.

stinkydiver73 commented 8 years ago

I'am lost, i thought the problem was not the whole SBAS, but only the integrity information in it. And we disabled that (true ?).

theArchLadder commented 8 years ago

@stinkydiver73 Yes, true, that was the big problem and it is solved now!

Now we are just trying to figure out if SBAS should default to AUTO or NONE, it doesn't really make a huge difference it seems. @iforce2d had position shifts of 2-3 meters because he has crappy SBAS coverage, SBAS sounds great but we have not yet been able confirmed that it actually helps us.

iforce2d commented 8 years ago

@theArchLadder Sorry, when I said "dangerous" I was thinking back to the original problem where SBAS would drop enough satellites that GPS functions would be abandoned altogether and the quad would drift with the wind. So that problem is not a problem anymore, correct. At least now it will not fly away, but I wouldn't say the 2-3m shifts are ideal.

In most typical cases we should not be flying that close to obstructions, although in my testing I have sometimes been hovering quite low and it almost got me once. So perhaps it's not as dangerous as I made out, but I feel that the safest available option is better for a default, especially when the gains to be had from SBAS are quite small for most users. People seeking better performance can delve a little deeper into their settings, and will hopefully have more awareness of what those settings are doing as a result. I think it would be great to have a link in the configurator next to the SBAS option, to explain a bit about what it does.

theArchLadder commented 8 years ago

@iforce2d I agree with everything you said! How confident are you that those 2-3m shifts are because of SBAS=AUTO? How many times has it happen and have you flown a lot with SBAS=NONE without having any of those shifts? (not saying you're wrong, I'm just curious!)

SBAS=NONE as default and info in the wiki that SBAS could be helpful if they live in areas with good SBAS coverage sounds like the way to solve this issue IMHO!

digitalentity commented 8 years ago

Ok to conclude: SBAS doesn't really improve PosHold and RTH performance (that's what a general users wants) and it may actually reduce performance in areas with crappy SBAS coverage.

I suggest we default SBAS=NONE as it's the safest.

theArchLadder commented 8 years ago

Well, we don't know if SBAS helps imho, i will keep using SBAS NONE/EGNOS every other week or so and see if i can notice any difference, would be nice if @stronnag and others with good sbas coverage could do the same!

Still, i agree SBAS=NONE seems like the best default for now :+1:

aster94 commented 4 years ago

did you noticed any improvement in these 3 years? is still advisable to leave SBAS to NONE or try one region? do you know where i could find a map of EGNOS? i didn 't find any

teckel12 commented 4 years ago

@aster94 https://en.m.wikipedia.org/wiki/European_Geostationary_Navigation_Overlay_Service

aster94 commented 4 years ago

Thanks @teckel12 I happen to be near a RIMS station, I guess I turn it on 😁

iNavFlight / inav

Investigate what default should be used for SBAS - NONE or AUTO #193