Author Topic: GNSS trustworthiness in mountainous and forested areas  (Read 6196 times)

captain paranoia

  • Sr. Member
  • ****
  • Posts: 384
    • View Profile
GNSS trustworthiness in mountainous and forested areas
« on: May 15, 2014, 05:26:55 PM »
A friend recently asked my advice following an experience with an MRT call-out in the mountains.

The team members are equipped with radios that add a GPS fix to each transmission, allowing the users to be tracked.  In the event in question, the radio wasn't used until the team member arrived at the casualty.  The fix placed the incident on one side of a mountain ridge, and a helicopter was called in to this location.

As it happens, the GPS fix was wrong, and the incident was actually on the other side of the ridge.

Since I'd discussed the effect of multipath and reflections on a UKC thread, the question to me was "was this a multipath incident?"  And my conclusion, looking at the traces of the incident, was that it almost certainly was (strictly, it was a shadowing & reflection problem).  Quite where the reflections were coming from that fooled the GPS receivers is hard to determine, but the incident occurred in a gully that gave at least 270 degree blockage, and in some parts, formed an almost complete bowl.

The problem is associated with canyons, and occurs in mountainous and urban environments, where the direct path between satellite and receiver is blocked (or 'shadowed'), but a signal is received from a reflecting surface, such as a cliff face or tower block. These reflections add additional path lengths to the measured pseudoranges, and so fool the triangulation calculations, pulling the fix away from the actual position. The problem is exacerbated by the increasing sensitivity of modern receiver chipsets, which can lock onto, and track the (generally) weaker reflected signals.  As the satellite constellation moves about, the reflections come and go, so a static user can appear to jump about.  If a satellite moves into direct view, this will often cause the position fix to jump back to the 'real position'.

The next obvious question was: "how can we tell when this effect is happening?"

Sadly, the answer to that is "tricky..."  Here are a few things that should start to ring alarm bells, and let you consider the possibility that the fix may be in error:

You're in a gulley, or a steep-sided valley.  There are slab-sided rock faces around you (although even more rounded or broken-up surfaces will cause problems).

If you're tracking users, and have regular updates*, then look out for sudden, large changes of position, which will look like long, straight track sections (you need to consider the time interval between fixes, too, as lack of intermediate position updates will also look like this).  Also watch out for zig-zagging paths, since these are often caused by changing reflections.  Of course, they can also be caused by zig-zagging up a steep face...  It might be possible to provide a graphical indication of large jumps, by colour-coding the traces with apparent velocity, so that jumps are highlighted in red, for instance, and 'normal progress' in green.

* that might be another recommendation: make sure users report their positions regularly, and don't simply wait until they're 'in position'; that way, you can track their progress, and probably detect large position jumps.  If you're using radios that prepend a GNSS fix to all voice transmissions, instigate a regular 'comms check' procedure (this might simply be a pressel/PTT, rather than an actual voice message), or, if the radio supports it, enable a regular, autonomous position transmission.  You'd need to consider the effect on battery life, of course.

Keep an eye on the receiver's accuracy estimate, since the receiver computation should be able to determine that the signals "don't add up"; the longer reflected path length is likely to cause a large error circle (cf a conventional 2D resection on a map; think about the size of the 'cocked hat' if you deliberately include a bearing error).  A poor accuracy estimate should warn you that something may be amiss.

If you can display the satellite constellation and C/No values for the receiver, you might be able to see if the geomorphology is blocking a number of satellites.  If your control station is fairly close to the incident (say 50km), then, if it has a clear view of the sky, you should be able to know what satellites should be visible, and compare them to the satellites received by a team member.  Since the satellite orbits are known to the system by almanac and ephemeris, it ought to be possible to display the predicted satellites (although I cannot tell you a system that will do this; GPS Test, perhaps?).

For a receiver equipped with mapping of some sort, and a user who can navigate without GPS, it should be obvious that the receiver is giving an incorrect fix (e.g. in this case, obviously on the wrong side of the ridge), and this might be reported to control.  Assuming voice comms is available...

Oh, and SBAS (WAAS/EGNOS etc) won't help, since it corrects (mostly) for clock and ionospheric errors, and knows nothing about the local multipath & reflection environment, so cannot correct these errors.

For the future, one might envisage a system that used ephemeris and a terrain map to predict possible reflection hot-spots, and provide a warning to users and control that GNSS fixes might be being distorted, either in real time (using a constellation model), or statically (using only the terrain map).  There is quite a lot of work going on at the moment addressing this problem of shadowing and reflections in urban environments (since urban users now dominate GNSS use).  There are a couple of interesting figures in the following paper that illustrate the sort of problems that occur, and the kind of traces that result from reflections distorting position fixes:

http://www.cs.bgu.ac.il/~eranfrie/gnss_signals_in_urban_canyons.pdf

The yellow trace is the actual path, and the red trace is the GNSS computed path.

My understanding is that SARLOC retrieves more data from the receiver than a simple position fix, and that data might usefully be displayed to a control operator to help identify possible reflection issues (estimated error radius, visible constellation, C/No).  The coloured trace & error radius ideas would be applicable to MRMap.  If any of the MRT/SAR types on here have contacts with SARLOC or MRMap developers, feel free to pass the suggestions on.  I'm also more than happy to hear feedback on the discussion above; it was pointed out that my naďve, laboratory-based suggestion of a way to reduce reflection effects might be the last thing on the mind of an MRT member on a steep, wet, scree slope...
« Last Edit: July 28, 2014, 06:13:42 PM by captain paranoia »

boogyman

  • Full Member
  • ***
  • Posts: 192
    • View Profile
Re: GNSS trustworthiness in mountainous areas
« Reply #1 on: May 16, 2014, 01:19:48 PM »
CP, if you are in such a location where GNSS reception is not trustworthy, would the averaged location-calculation be close to the real location? If so, than the functionality "average waypoint" (available on most recent satnavs) would be of much help, wouldn't it?

krenaud

  • Jr. Member
  • **
  • Posts: 70
    • View Profile
Re: GNSS trustworthiness in mountainous areas
« Reply #2 on: May 16, 2014, 10:14:46 PM »
In the Army we spent a lot of time training CASEVAC requests in order to get the guys to provide us with proper coordinates. One rule was to always plot the GPS position on the map and verify the location before calling in CASEVAC. It is a good procedure to follow and with modern GPS hardware with built in maps it is really easy to see if the position on the GPS map seems to match the actual location.

Using automatic GPS-positions without requesting confirmation from the crew in the field is a certain receipe for delayed extraction.



Lyle Brotherton

  • Administrator
  • Hero Member
  • *****
  • Posts: 939
  • Competent and safe navigation sets you free.
    • View Profile
Re: GNSS trustworthiness in mountainous areas
« Reply #3 on: May 17, 2014, 08:14:22 AM »
Kernaud, I totally agree with you and would go one stage further, any critical decision, dependent upon an accurate location, always confirm your location on your map, preferably your paper/plastic map; I use these in preference as you can view a much larger area on the map and if as CP described in his scenario you are on the other side of the ridge it will soon become apparent.

In poor visibility it is a good idea to reverse the procedure: transfer the grid reference of your current location to your handheld satnav/GPS and using my GRT makes this quick and easy.
“Opinion is the medium between knowledge and ignorance” - Plato

Lyle Brotherton

  • Administrator
  • Hero Member
  • *****
  • Posts: 939
  • Competent and safe navigation sets you free.
    • View Profile
Re: GNSS trustworthiness in mountainous areas
« Reply #4 on: May 17, 2014, 08:38:16 AM »
I wonder if this recent gross-inaccuracy has anything to do with firmware update for handheld satnavs/GPS?

There have been numerous notifications of inaccurate readings, which were traced to a specific satellite that is broadcasting data unsuitable for navigation – SVN64 (Space Vehicle Number) which GPS receiver chipsets should not be using.

 SVN64 actually broadcasts a data message that clearly indicates it is unusable for PNT (Positioning, Navigation, and Timing).
Nevertheless, the U.S. government has confirmed that certain GPS receivers are using data from SVN 64, in violation of GPS interface specifications, resulting in outages or corrupted, inaccurate position calculations.

I have emailed Garmin to ask if this is the case with their handsets.
“Opinion is the medium between knowledge and ignorance” - Plato

Callum

  • Administrator
  • Hero Member
  • *****
  • Posts: 512
    • View Profile
Re: GNSS trustworthiness in mountainous areas
« Reply #5 on: May 17, 2014, 02:16:21 PM »
Everything leads back to a map and compass ;)

captain paranoia

  • Sr. Member
  • ****
  • Posts: 384
    • View Profile
Re: GNSS trustworthiness in mountainous areas
« Reply #6 on: May 21, 2014, 02:47:02 PM »
Boogyman,

Averaging might help a little, but, actually, this is unlikely, since the error is systematic rather than noise-like.

In fact, in the incident in question, the position fixes did 'snap back' into place after some tens of minutes.

Lyle's theory of SVN64 is interesting; I've certainly seen one instance of a SiRFstar III chip set receiver giving me a 400m fixed offset for no apparent reason the I could see; in the rolling hills of the Ridgeway, although that was long before the launch of SVN64 (launched in Feb this year, and still being commissioned, i.e. not yet in operational use).  Reading the InsideGNSS article suggests 'outages' rather than errors, but I'd prefer a better report of the nature of the 'outages'; pretty poor reporting, IMHO.

captain paranoia

  • Sr. Member
  • ****
  • Posts: 384
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #7 on: July 28, 2014, 06:16:10 PM »
I've added 'and forested areas' to the title of this thread, since those areas are also subject to GNSS errors in a similar manner.

Hugh posted a thread about his experiences (although I suspect they may have been due to receiver startup, rather than multipath and non-line-of-sight reflection issues).

Lyle Brotherton

  • Administrator
  • Hero Member
  • *****
  • Posts: 939
  • Competent and safe navigation sets you free.
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #8 on: July 28, 2014, 09:37:38 PM »
CP would predictive ephemeris help negate this situation?
“Opinion is the medium between knowledge and ignorance” - Plato

captain paranoia

  • Sr. Member
  • ****
  • Posts: 384
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #9 on: July 29, 2014, 11:10:39 AM »
> CP would predictive ephemeris help negate this situation?

No.  Predictive ephemeris is merely a way of bypassing the cold start, by extrapolating the almanac to predict the ephemeris, so you know where the satellites are.  Once you know where the satellites are, you can determine the pseudoranges from them to the receiver, and thus compute the navigation solution.

The problem we have here is simply an RF propagation issue, due to ground-level objects that reflect, scatter, block and attenuate the signal.  And, since those effects are local in scope, the wide-area SBAS  correction systems cannot correct the errors; SBAS corrects for clock and ionospheric errors, and is best suited to open-sky applications such as aircraft.

It's a tricky one, and there's a lot of work going on to try to mitigate it.

Lyle Brotherton

  • Administrator
  • Hero Member
  • *****
  • Posts: 939
  • Competent and safe navigation sets you free.
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #10 on: July 30, 2014, 04:10:49 PM »
Ths CP :) You have helped me get my head around this topic, in part ;)
“Opinion is the medium between knowledge and ignorance” - Plato

Callum

  • Administrator
  • Hero Member
  • *****
  • Posts: 512
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #11 on: July 30, 2014, 09:43:04 PM »
Other than reporting a false location, is there another way you can tell if this is happening?

captain paranoia

  • Sr. Member
  • ****
  • Posts: 384
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #12 on: July 31, 2014, 06:22:18 PM »
Lyle, I'm still struggling to get my head around the causes and implications of this problem, and I'm aware that I'll never understand the theory any more than superficially.  Fortunately, I have theoretical gurus to call upon...

Hugh, I'm only making a guess at the effect you saw, based on your description.  But it's worth remembering that a GNSS receiver, as it does a warm start, will acquire satellites, and when it has their ephemeris, can start to use the signals in the navigation solution.  To get a basic 2D fix, it needs three satellites.  At this point, the receiver might say it is 'ready to use', but the chances are that the solution will be poor, and subject to considerable error.  As the receiver gets valid signals from additional satellites, the quality of the fix improves, and the error falls.  Thus, from a cold or warm start, it might take a while for the receiver to give an accurate position fix, and this can often be observed by watching the reported position jump about after it first says it's ready.

It's worth keeping an eye on the number of satellites being used in the solution (different to the number in view; it may not have acquired ephemeris for some of the in-view satellites, so their data cannot be used in the solution), and the estimated error radius.

captain paranoia

  • Sr. Member
  • ****
  • Posts: 384
    • View Profile
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #13 on: July 31, 2014, 06:39:45 PM »
Callum,

I think if there were an easy way of recognising that this situation is happening, receivers would report it, or modify their solution in some way to eliminate the rogue signals.  Unfortunately, the way that a GNSS receiver works means that it is susceptible to local reflections.  The receiver pulls out each satellite signal from the others using Code Division Multiplexing, and uses a correlator to find the individual pseudo-random bit sequence (PRBS).  This correlator finds the best match between the incoming signal and the PRBS, and, to some extent, assumes that the line-of-sight (LOS) signal is the strongest.  The local multipath reflections have a slightly longer path length, and so arrive a little later.  The averaging effect of the correlator can cause the LOS signal be be apparently delayed a little (the pulse is smeared 'to the right').

If there's not LOS signal, but a strong non-LOS signal (i.e. a reflection off a building or rock face), this signal will be found by the correlator, and there's nothing to distinguish it from a valid LOS signal.  With a bit of luck, there may be some statistical properties of this NLOS signal that might allow the receiver to eliminate it from the solution.

The essential problem is that a reflected signal looks very like a direct signal...

You might think that things would be made easier for walkers, who are moving slowly, compared to cars and other vehicles, but, actually, the opposite is true; a faster-moving receiver has a much more dynamic multipath environment, and that variation often falls outside the filters used in the navigation solution.  Whereas the variation for the slow-moving walker falls within the filter bandwidth, and is thus included in the nav solution...

GNSS receivers look simple; a little chip that fits into every smartphone.  But the signal processing involved is rather complicated...  Not only that, but even some receiver manufacturers don't properly understand how best to process the data, and when it's appropriate to apply certain filtering techniques, and when it isn't...

I've been giving this problem some thought, but I haven't yet checked the patent prior art, so I'm afraid I'm going to have to keep quiet about possible solutions I might have thought of...  They're probably well-established techniques, but my employer would be a little miffed if I gave the game away...

Hugh Westacott

  • Sr. Member
  • ****
  • Posts: 306
    • View Profile
    • Walk with Westacott
Re: GNSS trustworthiness in mountainous and forested areas
« Reply #14 on: August 01, 2014, 09:07:43 PM »
Thanks again, CP, for your detailed and helpful replies to what is a far more complicated subject than I'd ever imagined. In future, I shall not consult my satnav with the casual insouciance that I have heretofore!

Hugh
I grow old...I grow old, I shall wear the bottom of my trousers rolled. T.S.Eliot