We have been running into an issue quite often when flying at certain altitudes and 'near' structures. We are doing autonomous waypoint navigation and all of a sudden the GPS position jumps anywhere from several feet up to a few hundred feet. Obviously this isn't cool and luckily we have enough people to notice the issue and regain manual control before it does anything too crazy. From what I have seen via some quick research, this could very easily be what is known as 'multipath interference.' Has anyone seen similar issues and if so, have you been able to mitigate it?
Replies
It certainly sounds like multipath. It can be mitigated by better antenna design and/or receiver processing which tend to add cost, size, and weight. You don't say what you are flying or the type of GPS and antenna used (and budget) so I don't know how realistic a different GPS setup is. There is a huge difference between a ~$30 GPS receiver and a $1000+ one when it comes to things like multipath. Buying a higher-end GPS receiver with specific multipath rejection technology will probably do the trick. You may not have to throw that much money at it though. It really depends on your exact situation.
As for the antenna, a simple thing that can help is a ground plane. Assuming you are flying above the reflecting surface a ground plane of sufficient size will block signals from below (multipath) though getting something large enough, say 10 cm in diameter, on a small aircraft might not be practical. It doesn't have to be a metal disk. It could be part of the aircraft structure. It doesn't even have to be metal, it just needs to block the L1 frequency.
If you are flying below the tops of buildings or other objects then you may be dealing with multipath AND periods of poor satellite geometry (blocked signals). At this point you may not be able to rely on GPS no matter what you spend though a higher-end GPS/GLONASS board with multipath rejection technology would give you your best chance at success.
I'd bet cash money on multipath interference, too. I've dealt with that on my ground vehicle. I ran into those issues in the 2011 Sparkfun AVC where my SirfIII just couldn't get a decent position fix when driving so close to the large, multistory Sparkfun building. In 2012, in the few test runs I did, using a Venus 638flpx, the issue seemed reduced. The Venus is stated to be less sensitive to multipath.
Some GPS antenna types are more sensitive to multipath interference than others.
My robot didn't actually use GPS position but instead GPS heading combined with dead reckoning from gyro and encoders. It worked very well. It was very precise in test runs and the 3rd place run around the building. That's with just a 2d autopilot. The point being is that something similar could help your autopilot.
Position estimate would be based on some sort of fusion of GPS reported position and dead reckoning position estimate based on groundspeed (from airspeed and prior wind estimates) and AHRS heading.
The idea being that if one of the sensors goes whack, the algorithm's estimate isn't wildly affected. It could be as simple as programmatically switching estimation algorithms when error between them grows quickly. Or it could be a Kalman Filter.
There's a lot of detail that would have to be explored and discussed. I'm just thinking out loud here but my intuition is that coming up with an estimate from multiple sources could help.
That is, if simply changing GPS modules / antennas doesn't.