Weekly Dev Call of October 10, 2016

Agenda for next week’s call:

  • FMUv5 / STM32F7 support: #5681, pinout proposal
  • Additional HW support: Looking into NXP and Atmel offerings
  • Quick update on Parrot Bebop 2
  • DSM parsing investigation results

Re: changes to sensor pipeline (issue https://github.com/PX4/Firmware/issues/5453)

There is an example log posted here: http://logs.uaventure.com/view/x46eDbXxMAHzyWV6bPPBdj which shows the “normal vibration levels for a standard vehicle”.
What is the size and type of airframe?
Also what kind of vibration isolation and which flight controller was used?

Its a 1.5 kg hexacopter with a FMUv1 with no isolation. Any F440 sized frame with proper mounting should come close to that.

6DOF ACRO mode
Re: https://github.com/PX4/Firmware/pull/5498
This PR adds an “ACRO” mode to the MC position controller to remove the restriction on roll and pitch angles imposed by the current STABILIZE mode. This will improve “lock-in” in all orientations (including inverted) assuming the IMU has good dynamic accuracy. If that is not the case, it will at least make the errors easy to demonstrate.

In addition, it adds what I’m calling an “attitude sequencing” capability which allows one to program flight sequences such as 4-point rolls. This would obviously be more useful for a collective-pitch copter/multicopter, and it would be nice to fully support those airframes.

Third, it adds logging of the simulated attitude to aid in testing changes to the attitude estimators and controllers. Assuming good simulation fidelity, this is useful for quantifying dynamic accuracy of the attitude estimate.

RC Loss Failsafe
Re: https://github.com/PX4/Firmware/pull/5388

This issue results in a fly-away on RC loss. The PR has been tested on PixRacer, but not FMUv2, since there are open questions about what would be a correct implementation.

att_estimator_q improvements
Re: https://github.com/PX4/Firmware/pull/5275

This change was demonstrated (in actual flight) to improve dynamic attitude estimation accuracy in a branch incorporating multiple additional changes but has not been validated on latest master. The test branch was running 2 instances of att_estimator_q and logging both attitude estimates while manually flying coordinated turns. The new attitude sequencing feature in the 6DOF PR along with simulator ground-truth logging will enable quantitative testing in SITL also.