[Canberrauav] notes from today's flight test

Chris Gough christopher.d.gough at gmail.com
Wed Dec 14 13:45:17 EST 2011


Flight-test report from today.

The primary purpose was to test the new "multiple data link" features
that Tridge added to mavproxy yesterday. This allows mavlink data
(telemetry and telecommand) to travel across two or more data links at
the same time in parallel (i.e. multiple mavproxy "masters"),
providing redundancy. The secondary purpose was to continue
observing/documenting procedures, and I have added some notes to the
wiki (bring-up procedure for the telemaster at CMAC):

http://wiki.aestabjoo.net/wiki/index.php/CMAC_pre-flight_procedure

The flight test was a success, however we discovered some bugs and
captured some issues.

Prior to having multiple links in mavproxy, the operator would hear
the occasional "no heartbeat" warning vocalised from the GCS. This
indicates that the data link has been lost; it's usually only for a
brief moment, e.g.  when banking into a turn some distance away. With
this new feature we would expect instead that one or the other data
link might drop out for a moment but almost never both at once, so
instead of "no heartbeat" messages we would get "data link x down",
"data link x up" messages from time to time, but a solid composite
link. During the flight, there was some "data link 1 down" / "data
link 1 up" messages and none "no heartbeat" messages. We'll know more
when Tridge has analysed the logs, but basically it seems to be
working.

We uncovered a few issues and a bug in the new feature, here are my
notes (TODO/issues):

 * a mavproxy module for waking an operator through ground-based tests.

 * realtime ground-based image viewing solution, for flight observers
/ supervisors

 * mavproxy module for logging signal strength

 * solve smearing problems with chameleon cameras. What settings are
wrong? how to reproduce on the bench? how to detect smearing from the
air?

 * update message logger - use last 3 bits of timestamp (super-fine
resolution, just noise) to identify the data link

 * BUG - debayer program not garbage collecting properly

 * make "play/stop/RFWD/RWD" buttons for preview program (no more
click click click to browse through images)

 * blast the control key on "sunny" (GCS laptop) with compressed air,
it tends to get a bit stuck and this might fix it.

 * ubiquity modem has a dodgy connection somewhere, intermittent power
failure (and it takes 30 seconds to reboot, after the cable-jiggling
makes the light come back on). While this was useful for testing
multiple data link features, we've done that now so it should probably
be fixed :) One option would be to replace the CAT cable (and perhaps
use braid to isolate the socket from mechanical stress), or replace
the POE injector with a direct power connection with a high
reliability connector (eliminate one possible source).

 * BUG: using the multi-link features, when sending new parameters to
the plane from the GCS, the state model of each link is not updated
with the new parameter value (unless parameter explicitly "fetched").

 * BUG: episodic latency (stutters ~3s duration) on one link but not
the other is causing some grief. May be a buffering issue (where;
xbee? ubiquity? software blocking?). May need to find and fix
buffering issue, and also handle message asynchronicity differently.

Chris Gough

-- 
.



More information about the Canberrauav mailing list