Still dropping out of Fixed mode when resolving occasionally about piksi_firmware HOT 5 OPEN

1. is getting log files containing the problem
2. run through SITL

from piksi_firmware.

I'm keeping a journal of this issue, which I'm going to link here later today. The current plan is:

1. Find a reproducible test case for using external inputs, since it seems like a control flow problem.
2. Add a HITL test for inappropriate fixed mode transitions. Collect overnight instances of logs where that happens. Might want to see how many fixed=>float transitions result from this, versus other situations (like loss of tracking).
3. Do some trace debugging for the dgnss_update methods to eliminate the current hypothesis.
4. SITL checking of the filter with the existing log.
5. Add some mocked filter testing to the dgnss_update based on results of 2 and 3.

from piksi_firmware.

I think (5) should probably be in the form of a harness to automatically run a set of log files through SITL and testing of the results. Lets chat about it when you have a free moment.

from piksi_firmware.

Current status: fixed -> float transitions have been reduced a lot but not entirely eliminated, the remainder are mostly due to false locks and satellite signal loss (rather than bugs) but this needs to be properly confirmed

from piksi_firmware.

What fergus said. I think the remaining cases would need to be categorized.

from piksi_firmware.