Re: GPSD For Real Time Application

[Gary E. Miller]

Yo Kashyap!

On Sat, 28 Dec 2019 00:01:24 +0530
Kashyap Gada <address@hidden> wrote:

> My application -
> Record some events/sensor readings in a moving vehicle 100MPH using a
> Ublox Neo-M8U GPS. Accuracy required is better that 5 meters.

Good luck with that.  100mph is 44.70 meters per second.  GPS receivers
do not do well at that speed.

> Until Now -
> Have built GPSD/CGPS/GPSMON from source on Raspberry Pi. Used UCentre
> to Fix 10Hz output from GPS. Tested a python and C++ source code.
> Tested them for more than 12 Hours continuous operation.

u-center is winblows, we don't do windows.  Use ubxtool instead.

gpsd is lightly tested at 10Hz.  Any feedback you have is welcome.

> Developing stable client software that can handle 10hz update rates
> which can operate for 24x7 all round the year.

I assume you ask because gpsd has not done this for you yet?  If not,
why not?

> 1. As per other posts on the mailing-list, 'gpsd accepts all what gps
> sends to it and forwards it even at higher frequency'.

I would say forwards is at lower frequency.

> What does the
> gpsctl -c command which has minimum value of 0.25 actually do.

That is very driver/device dependent.  For u-blox, use ubxtool instead.

> Do its
> setting matter when GPS is sending data at 10hz or more?

Nope.

> What cycle
> times mean in gpsd context.

The rate at which a GNSS receiver send new unique fixes.  Most plain
receivers send a new fix every second.  So the cycle time is 1 second.

A 10Hz receiver outputs 10 fixes a second, for a 0.10 sec cycle time.

> 2. For 100MPH and requirement of 5 meter accuracy what is ideal GPS
> frequency to be selected. I feel 10Hz is minimum that is required.

Maybe.  We don't ahve that data.

> But some where on the mailing-list i read replies where it was said
> that 1hz was more that enough in major applications and any speed
> more that that was not helpful.

So do a search of the mailing list.  I vaguely remember that and the
important part was supplementing the GNSS data with IMU and wheel
tick data.  But we never got any good details.

> I didn't understand why higher rates
> are not required, even in such cases where there is high speed
> movement in vehicles.

Let us know when you figure it out.

> 3. I have set 10hz frequency in NEOM8U GPS through Ucenter and saved
> it in flash with a baud rate of 38400.

38400 seems pretty slow for 10Hz data.  That adds a lot to your latency.

> The gpspipe log file
> 'gpspip-timestamp.log' is attached.

I'm not sure what to do with that.  Did you see something in the data
we should look at?  Also, the timestamp makes the file not work with
gpsfake.

> Is this an optimum setting for my
> application? I don't mind occasional missing of locations btw.

Dunno.  The u-blox has so many settings that interact in odd ways,
One of the important ones is the "Dynamic Platform Model".  You'll need
to run a lot of tests.

> 4. Will python/c++ clients be able to receive or handle higher update
> rate outputs?

People have reported success before.

> 5. I have attached a python code I modified from a sample I found
> online. This code has a thread where next() is called with some
> sleep.

You will want to use the new aiogps module to get async performance.

> And in the main loop i do some processing. This code, after
> operating satisfactorily for 5-6 hours, gives only one set of
> Lat/Long, even through cgps/gpsmon Lat/Long are changing.

So, since the lat/lon does not change your code is broken?

> Read
> somewhere that you don't need to sleep in thread. I dont know what to
> do here. I need 10 hz latest location data in my python script how
> should i go about it. file name is 'test_gps.py'

Not having a 10Hz GNSS receiver, nothing I can do with it.  Maybe someone
else has one.

> 6. I need gps data only when the vehicle is moving above 10MPH. Is
> there any way to pause gpsd?

Why would you care to pause it?  To save 100 mA?

> I have  tried to list all questions I had in my mind not to miss out
> any. Apologies for the long list.

Well, apologies for the lack of answers, you are going where few have
gone before, and fewer sent us back postcards from the edge.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        address@hidden  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin

pgpXFh96lJD0P.pgp
Description: OpenPGP digital signature