[FOSS-GPS] foxtrotgps: Tracking weather probes in addition to the own position

[Tilman Glötzner]

Dear Joshua, Dear all

> These days I think libgps/gpsd actually has the ability to use multiple
> `GPS' devices
> (whether or not they're actually GPS receivers or something else entirely)
> and even report them to us with distinct ID-strings. So it might
> actually be
> most *straightforward* to just try to do things that way: maybe have your
> RS decoder feed into gpsd, and modify foxtrotgps to differentiate
> between multiple devices as reported by libgps?
> 
> This is of course just a suggestion, and is based on things that I've
> read--
> if someone has actually used the multi-device support and can comment
> on its usability (or lack thereof), I'd love to hear it.
> 
> I'm assuming that it would be easy enough to produce gpsd's JSON protocol,
> but again I don't really have first-hand experience doing so....
> 
> Maybe you've already considered it and rejected it though?
a) Something like that has been done:
https://destevez.net/2017/11/tracking-an-rs41-sgp-radiosonde-and-reporting-to-aprs/

The procedure  works fine at home. When going after the probe by a car
(i.e. hunting the probe), this setup with all its pipes and scripts
proved  to be difficult to handle in  "field conditions". Usability and
reliability become issues in cramped conditions of a car and operating
under time pressure (other are after the probes as well -- only the
first one at the probe  wins :-)).

b) I have not really rejected using gpsd and have not looked deeply into
it.

 The choice to try follow the path of writing an additional module for
foxtrotps had several drivers:

-  I was not really sure how to extend foxtrotgps to use 2 gpsd sources
and found it easier to implement an additional module for foxtrotgps
that reads the datagram of the probe decoder.

- I would probably need to write a gpsd driver in addition as the setup
using the pseudo device as described above works not reliably: I needed
to restart gpsd  and the scripts/programs feeding into the pseudo device
after I woke up the laptop from sleep mode for instance.

- The gpsd's JSON protocol seems  not provide separate data fields for
the vertical and horizontal velocity.

Ideally, the full setup  (receiver, probe decoder, gpsd for gps mouse,
and foxtrotgps) would be running on a raspberry pi. The used programs
would open the sockets themselves (rather than using nc ) and
automatically reconnect in case something fails. And fewer programs
along the processing chain should increase reliability.

c) Why did I opt for foxtrotgps ?
- Code is available and in principle good to read/understand
- Maps can be downloaded for offline use
- Can be compiled and runs on rasperrypi. The the gui elements are big
enough to be used with a touch display.
- The routing function is already there. It should not be difficult to
enhance this with function "calculate route to additional target, i.e.
the position of the weather probe.

> 
>> - Once a GPS position is received, it is drawn. If a datagram from a
>> weather probe is received, everything i.e. the map window is redrawn. So
>> if a valid gps position is received and a datagram of the probe is
>> received, redrawing takes place twice per second. I wonder if it is
>> possible to only locally draw the probe position rather than redrawoing
>> everything.
> 
> Hm. How are we doing it right now for the GPS marker that we already have?
Not entirely sure. The key seems to be gps_function.c:cb_gps_timer.
gdk_window_process_all_updates() updates all pending event of all
windows. The GPS marker seems then to be drawn directly with
gdk_draw_arc and subsequent gdk_draw_line commands on the canvas. I
however do not understand when the events drawing the GPS marker are
actually processed, i.e. if they are synchronously visualized with the
execution of the gdk_draw_* commands on the canvas. And I do not really
understand how the outdated GPS marker is removed when a new fixed is to
be drawn. I assume that first the map is redrawn and then the new fix on
top of it.

When I draw the probe icon in rs_functions.c:paint_rs_position with
gdk_draw_pixbuf, it is not drawn until I click on the canvas with the
mouse or zoom thereby issuing an event. When I insert an additional
gtk_widget_queue_draw_area at the same position, the icon is shortly
drawn and then vanishes...

rs_functions:paint_rs_position
....
gdk_draw_pixbuf (map_drawable->window,	gc_map,	rsposition_icon,
					0,0,x-12,y-32,
					24,32,
					GDK_RGB_DITHER_NONE, 0, 0);
gtk_widget_queue_draw_area (map_drawable,x-12, y-32,24,32);

Thanks
Tilman