Team-BHP > In-Car Entertainment


Reply
  Search this Thread
5,834 views
Old 12th April 2019, 12:50   #16
Team-BHP Support
 
tsk1979's Avatar
 
Join Date: Feb 2005
Location: San Jose, CA
Posts: 23,717
Thanked: 22,769 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Jeroen View Post
thanks, some very interesting reading. I wonder how this would work for speed readings, as the underlying technique on GPS tends to be a doppler measurement on some of the signals. Not sure how GLONASS or the others measures/calculates speed.

With different orbits, different reference clocks etc, it will need quite some corrections to get a speed using satellites from different constellations.

Jeroen
Check this out for doppler speed measurement and why its accurate. Using multiple systems (Galileo, Glonass, GPS....) is not difficult

https://racelogic.support/01VBOX_Aut...es_GPS_work%3F

Also check this out. This is one such algorithm employing multiple systems at the same time. See how they say "Algorithm", because its the software which is using multiple systems. The hardware is just relaying the message. What that message means is the coders job. In this case there is a base station, providing another signal to the roaming VBOX. Its like the black hole image


https://www.vboxautomotive.co.uk/ind...al-antenna-rtk

Last edited by tsk1979 : 12th April 2019 at 12:51.
tsk1979 is offline   (4) Thanks
Old 13th April 2019, 18:03   #17
Distinguished - BHPian
 
Dr.AD's Avatar
 
Join Date: Apr 2010
Location: Bangalore/Pune
Posts: 1,791
Thanked: 18,436 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Jeroen View Post
Thanks.
This paper shows that it can be done, not that it is being done. I am interested so I have searched around a bit. I have found quite a bit, but I haven't seen anything yet that suggest that there are chipsets available and being used to provide proper integrated positioning. Rather than switching between two different constellations.
Jeroen
Not a technical paper, but a somewhat direct proof that this is already being done is the screenshot below. This is a screenshot from "GPS Status" app, which shows the satellites used in the current fix. The satellites shown by green circles are GPS satellites, whereas those those shown by green rectangles are GLONASS satellites. The app shows that it has used 14 satellites (out of 23 visible satellites) for obtaining the current fix, and as you can see, those 14 are a mix of GPS and GLONASS satellites.

At many times, in areas of poor visibility to sky (such as inside my office), I have seen fix obtained with only 4 satellites, and even there it can do it with a mix of GPS and GLONASS satellites.

GPS epoch rollover-screenshot_201904131301111.jpg
Dr.AD is online now   (6) Thanks
Old 14th April 2019, 12:20   #18
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,036
Thanked: 49,768 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Dr.AD View Post
The app shows that it has used 14 satellites (out of 23 visible satellites) for obtaining the current fix, and as you can see, those 14 are a mix of GPS and GLONASS satellites.

At many times, in areas of poor visibility to sky (such as inside my office), I have seen fix obtained with only 4 satellites, and even there it can do it with a mix of GPS and GLONASS satellites.
Thanks, that is an interesting app too. I need to find something like that for my iPhone.

with so many satellites from both systems available it does beg the question how it would calculate a position. It has sufficient satellites to calculate a 100% GPS and a 100% GLONASS position.

As far as I know GPS tends to be more accurate than GLONASS except for high latitudes. Adding more than 5 satellites into a position calculation does not increase the accuracy in a meaningful way I believe (at least up to some years ago it did not)

So if you have sufficient GPS satellites in view, why would you use GLONASS satellites as well. Would that not decrease the accuracy? I think it might be to do with the dilution of precision (DOP) as well, or rather that is where using different constellations might have a considerable effect.

For tracking purposes, the more satellites you can track from more system the better it is obviously, but not necessarily for precision.

Jeroen
Jeroen is online now   (2) Thanks
Old 14th April 2019, 19:42   #19
Distinguished - BHPian
 
Dr.AD's Avatar
 
Join Date: Apr 2010
Location: Bangalore/Pune
Posts: 1,791
Thanked: 18,436 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Jeroen View Post
Thanks, that is an interesting app too. I need to find something like that for my iPhone.
This is a very useful Android app called "GPS Status & Toolbox". It gives a fairly detailed technical information about the GPS system. Unfortunately it looks like this is not available for iPhone. On iTunes, I found another app which is a bit basic, but nonetheless can give you some information: https://itunes.apple.com/lr/app/ulti...ps/id403066634

Quote:
Originally Posted by Jeroen View Post
As far as I know GPS tends to be more accurate than GLONASS except for high latitudes. Adding more than 5 satellites into a position calculation does not increase the accuracy in a meaningful way I believe (at least up to some years ago it did not)
I am not an expert in GPS technology, but from what I have read and from what I have observed, more number of satellites used in the fix generally (not always) gives more accurate position. The Dilution of Precision (DOP) that you mentioned is one factor. However, in general, with uncertainties in calculations from individual satellites, there is always some overall uncertainty in calculation of the net position. With 4 satellites, theoretically it is enough to calculate the position (i.e. to get the fix), but the uncertainty will be high. From my observations, whenever I get fix with only 4 or 5 satellites, the errors are about 50m.

As you get data from more and more satellites, it looks like (from what I read) the positioning algorithm uses a least square based averaging, which can reduce errors by using all the data, and the accuracy of net position increases.

Mathematically speaking, with 4 unknowns, and 4 equations (i.e. fix from 4 satellites), you can solve this system of equation and get a solution. However, given that there are uncertainties in the 4 equations, your answer will have uncertainties. Further, if these 4 equations are not independent (i.e. the 4 satellites are not sufficiently away from each other), then you will have an "ill conditioned" system of equations. In such cases, it will help to use more number of equations. However, now you have more number of equations than unknowns. Therefore, instead of a "direct solver" (which would solve this as system of equations using only four equations, and will not be able to make use of extra equations), one could use a "least squares solver" which would try to minimize the errors averaged over all the equations. In simpler words, one would use an averaging method that would find a position using many satellites such that average error is very low.

Again, I am not an expert in specifics of GPS, and I hope the experts would pardon me and correct me if I am wrong. But here are some links which give similar answers:

https://www.reddit.com/r/askscience/...ed_satellites/

https://wiki.openstreetmap.org/wiki/...cy_of_GPS_data

https://electronics.stackexchange.co...els-make-sense

As an empirical proof of this, I did a quick experiment where I measured the error in location with different number of satellites used. See below the screenshots of my observations:

GPS epoch rollover-gpsaccuracieswithsatellitenumbers.png

As you can see, the error decreases (i.e. the accuracy increases) with more number of satellites used in the fix. However, please note that the above is a general case I chose for illustration. It is usually this way, but not necessarily always true. In some cases, depending on which particular satellites are used, sometimes you may get higher errors even with more number of satellites (and this is because of the degenerate system of equations - some of the equations will not be linearly independent).

Last edited by Dr.AD : 14th April 2019 at 19:57.
Dr.AD is online now   (4) Thanks
Old 14th April 2019, 19:47   #20
Senior - BHPian
 
ruzbehxyz's Avatar
 
Join Date: Oct 2014
Location: MH02 to MH46
Posts: 1,612
Thanked: 6,591 Times
Re: GPS epoch rollover

Being a Merchant navy captain, the Navstar GPS is a very important piece of equipment onboard which is connected to many navigational equipment on the bridge / wheelhouse of the vessel.

The discussion here has covered all major aspects, so there is nothing much for me to write about.

Thanks to all on this thread to share such valuable info.
ruzbehxyz is online now   (3) Thanks
Old 14th April 2019, 22:10   #21
Senior - BHPian
 
Join Date: Jul 2009
Location: Calcutta
Posts: 4,668
Thanked: 6,213 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Dr.AD View Post
This is a very useful Android app called "GPS Status & Toolbox".
Just a note - 'GPS Status' tracks you whenever you turn on location (irrespective of whether you are using it). And sells data collected. (I even paid for the pro version! Uninstalled now).

Regards
Sutripta

Last edited by Sutripta : 14th April 2019 at 22:11.
Sutripta is offline   (3) Thanks
Old 15th April 2019, 14:15   #22
BHPian
 
Join Date: Jan 2010
Location: Bangalore
Posts: 473
Thanked: 472 Times
Re: GPS epoch rollover

Thanks for all the extraordinary posts here. I learnt a lot over the past few days. Added to the epoch rollover, I read about the following hacking of GNSS as below. This is all becoming downright scary about our dependence on things that are vulnerable to manipulation by state powers.

https://www.businessinsider.in/The-R...w/68872749.cms
bsdbsd is offline   (1) Thanks
Old 15th April 2019, 15:34   #23
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,036
Thanked: 49,768 Times
Re: GPS epoch rollover

Quote:
Originally Posted by bsdbsd View Post
Thanks for all the extraordinary posts here. I learnt a lot over the past few days. Added to the epoch rollover, I read about the following hacking of GNSS as below. This is all becoming downright scary about our dependence on things that are vulnerable to manipulation by state powers.
]
It is beyond me why anybody would rely on Russian technology for anything as critical as this. It’s ok in an iPhone or a car navigation. But on a plane or a ship?

I don’t know why you would want to have GPS and GLONASS out at sea. No problems with getting satellites I would think?

By the way, also GPS gets regularly jammed due to military exercises. Contrary to the Russians the Americans will publish exactly where and when you can expect problems with GPS.

Note jamming is something very different from the hacking/interfering as suggested in the previous post.

Jeroen
Jeroen is online now   (1) Thanks
Old 15th April 2019, 18:09   #24
Senior - BHPian
 
Join Date: Jul 2009
Location: Calcutta
Posts: 4,668
Thanked: 6,213 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Jeroen View Post
GPS needs 3 satellites to calculate a position, fourth satellite will make it more accurate.
4 satellites to get a fix in 3D space, I would think. (Though sometimes one set of possible positions can be discarded as being absurd)

Quote:
Originally Posted by Dr.AD View Post
As you get data from more and more satellites, it looks like (from what I read) the positioning algorithm uses a least square based averaging, which can reduce errors by using all the data, and the accuracy of net position increases.

Mathematically speaking, with 4 unknowns, and 4 equations (i.e. fix from 4 satellites), you can solve this system of equation and get a solution.
An overdetermined solution, handled properly will certainly increase confidence in the result. But I don't think that is how commercial consumer systems work. Essentially because these lack the computational chops necessary to do so, and for very little incremental user benefit. Normally commercial chipsets can simultaneously track only a fixed number of satellites. Firmware then decides which satellites to use for best results. Importance is paid to HDOP, rather than VDOP (which in any case is not very good with GPS)

Quote:
Originally Posted by Jeroen View Post
It is beyond me why anybody would rely on Russian technology for anything as critical as this. It’s ok in an iPhone or a car navigation. But on a plane or a ship?

I don’t know why you would want to have GPS and GLONASS out at sea. No problems with getting satellites I would think?
Backup/ fallback/ reliance on a foreign government/ reliance on technology not in your control.

Quote:
By the way, also GPS gets regularly jammed due to military exercises. Contrary to the Russians the Americans will publish exactly where and when you can expect problems with GPS.

Note jamming is something very different from the hacking/interfering as suggested in the previous post.
The owner/ controller of the system does not have to jam the system. eg the US can just switch off the civilian part, or switch SA back on.

Jamming/ spoofing/ hacking all are possible. And are different from each other.

@DrAD what's your ELO rating? Why don't you start a thread on say your favourite games, with comments.

Regards
Sutripta
Sutripta is offline   (2) Thanks
Old 15th April 2019, 18:28   #25
Distinguished - BHPian
 
Join Date: Oct 2012
Location: Delhi
Posts: 8,036
Thanked: 49,768 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Sutripta View Post
The owner/ controller of the system does not have to jam the system. eg the US can just switch off the civilian part, or switch SA back on.

Jamming/ spoofing/ hacking all are possible. And are different from each other.
Yes they can, but more often they actually jam.Or at least they call it jamming. It is brought to attention of pilots, very civil, in an orderly fashion:

https://www.aopa.org/news-and-media/...rcises-planned

Jeroen
Jeroen is online now   (1) Thanks
Old 16th April 2019, 21:51   #26
Distinguished - BHPian
 
Dr.AD's Avatar
 
Join Date: Apr 2010
Location: Bangalore/Pune
Posts: 1,791
Thanked: 18,436 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Sutripta View Post
Just a note - 'GPS Status' tracks you whenever you turn on location (irrespective of whether you are using it). And sells data collected. (I even paid for the pro version! Uninstalled now).
Thanks for pointing this out. I did not know this. I will be careful. However, when I checked the permissions for this app, it only has permission for location. I have disabled every other permission (including storage) to this app, and it still works well. So it can get only anonymous location data. Sadly, there are so many other apps that have access to location data (food delivery apps, shopping apps for example), and from what I have heard, they also use/sell/share anonymous (or rather specifically anonymized in their case, since they actually have the identifiable data) location data. So looks like there is not much we can do about anonymous location data.

Quote:
Originally Posted by Sutripta View Post
4 satellites to get a fix in 3D space, I would think. (Though sometimes one set of possible positions can be discarded as being absurd)
Yes, 4 satellites are minimum required for a 3-D fix (including altitude), and 3 satellites are required for a 2-D fix (longitude and latitude only), based on the assumption that you are at mean sea level.

Quote:
An overdetermined solution, handled properly will certainly increase confidence in the result. But I don't think that is how commercial consumer systems work. Essentially because these lack the computational chops necessary to do so, and for very little incremental user benefit. Normally commercial chipsets can simultaneously track only a fixed number of satellites. Firmware then decides which satellites to use for best results. Importance is paid to HDOP, rather than VDOP (which in any case is not very good with GPS)
Interesting. With modern chips, I don't think solving a least squares is much of a problem. Anyways, I am not sure exactly which algorithms are used today. However, as I mentioned earlier, one can clearly observe that the position accuracy increases with more number of satellites. So I assume some form of overdetermined solution is used in reality. Else we would have seen more or less constant error with more than 4 satellites. In reality, what I have observed is that the error becomes constant only when more than 12 satellites are used in the fix. Upto 12 satellites, the error decreases monotonically with number of satellites. And there is definitely additional benefits is using more satellites upto 12, and after that there is no additional benefit.

Quote:
@DrAD what's your ELO rating? Why don't you start a thread on say your favourite games, with comments.
I had a decent ELO rating about a decade ago when I was an active player. Now I am not, and only play some casual games on the internet. I did not realize there is no thread about chess games. Surely we should start one. Are you a chess player too?

I have many favorite games and moves. Off the top of my head, a couple of games that comes to mind are Kasparov-Topalov (Wijk aan Zee, 1999) and Topalov-Shirov (Linares 1998). The 47. ... Bh3 by Shirov was a masterstoke! A move that I vividly remember everytime I think of Shirov :-)
Dr.AD is online now   (1) Thanks
Old 17th April 2019, 20:43   #27
Senior - BHPian
 
Join Date: Jul 2009
Location: Calcutta
Posts: 4,668
Thanked: 6,213 Times
Re: GPS epoch rollover

Quote:
Originally Posted by Dr.AD View Post
Thanks for pointing this out. I did not know this. I will be careful.
...
So looks like there is not much we can do about anonymous location data.
People knowing where your home is is different from knowing your travel patterns. Both are bad, in different ways.
Data is anonymized, but does not take much to fit back the pieces.

Quote:
based on the assumption that you are at mean sea level.
That is an added operation - 'clamp to ground'. (Google Earth/ KML calls it tesselate, though that is not the dictionary meaning of the term).

Quote:
Interesting. With modern chips, I don't think solving a least squares is much of a problem. Anyways, I am not sure exactly which algorithms are used today. However, as I mentioned earlier, one can clearly observe that the position accuracy increases with more number of satellites. So I assume some form of overdetermined solution is used in reality. Else we would have seen more or less constant error with more than 4 satellites. In reality, what I have observed is that the error becomes constant only when more than 12 satellites are used in the fix. Upto 12 satellites, the error decreases monotonically with number of satellites. And there is definitely additional benefits is using more satellites upto 12, and after that there is no additional benefit.
The Sirf Star3 was the first of the commercial consumer GPS chip with acceptable performance. In that DOP values were calculated on the chip, and read in by external circuit. Along with other values like lat/ long. Not calculated externally. Don't know, but I think the philosophy would be the same on other modern implementations, including Qualcomms.
The VLSI experts here will be better able to explain (or contradict) it, but digital circuits (gates) on a silicon wafer essentially fabricated to handle analog signals is costly, and so kept to a bare minimum.

If one is seeing only a limited number of satellites, in all probability one is seeing only a small quadrant of the sky. Tunnel vision. One is unlikely to get good accuracy in that case.

I think something a bit more sophisticated than least sum of squares is used. Kalman filters!

Quote:
Are you a chess player too?
Now? No. Lack of sparring partners, and time, after college. (A few years back, a fairly simple program like Chessmaster 7000 beat the pants off me! And 7000 was vintage by then! I take solace in the fact that I could at one time beat CM2000!)

Regards
Sutripta

Last edited by Samurai : 18th April 2019 at 11:22. Reason: typo
Sutripta is offline   (2) Thanks
Reply

Most Viewed
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Copyright ©2000 - 2024, Team-BHP.com
Proudly powered by E2E Networks