proper in_flight detection for rotorcrafts

[flixr]

The "detection" when you are actually in flight needs some serious improvements.
This is currently done in autopilot.c autopilot_check_in_flight(), just based on if the motors are on and the throttle stick is over a certain threshold for some time.

While you are not in_flight, the yaw control is disabled and the integrators are not running.

The in_flight detection could maybe be done on a combination of velocity, acceleration and throttle...
It would be nice to have a basic state machine with e.g. landed, takeoff, flying, landing or so...

[dewagter]

Same for the landing detection. I know of at least 2 crashes due to killing
inflight.

-Christophe

On Sat, May 26, 2012 at 2:39 AM, Felix Ruess <
reply@reply.github.com

wrote:

The "detection" when you are actually in flight needs some serious
improvements.
This is currently done in autopilot.c autopilot_check_in_flight(), just
based on if the motors are on and the throttle stick is over a certain
threshold for some time.

While you are not in_flight, the yaw control is disabled and the
integrators are not running.

The in_flight detection could maybe be done on a combination of velocity,
acceleration and throttle...
It would be nice to have a basic state machine with e.g. landed, takeoff,
flying, landing or so...

---

Reply to this email directly or view it on GitHub:
#201

[softsr]

Same story. Landing detection should be improved.
I've made actually some small improvements of the present in_flight detection.
https://github.com/softsr/paparazzi/blob/krooz_port/sw/airborne/firmwares/rotorcraft/autopilot.c#L228
It is indeed should be a combination of the throttle stick position, vertical speed and acceleration. The yaw control should be always enabled, IMHO. It would probably be useful, to trim the yaw PID in a case of descent to prevent possible roll-overs during touching ground.

[flixr]

Same story. Landing detection should be improved.
I've made actually some small improvements of the present in_flight detection.
https://github.com/softsr/paparazzi/blob/krooz_port/sw/airborne/firmwares/rotorcraft/autopilot.c#L228

Could you make a pull request with this please? Although not the final solution, that should at least be better than what we have now...

The yaw control should be always enabled, IMHO.

I don't quite understand why you want to have yaw control enabled on the ground... could you elaborate?

It would probably be useful, to trim the yaw PID in a case of descent to prevent possible roll-overs during touching ground.

Sorry, I'm not quite following. What do you mean by that?

[softsr]

Yes, right, there's no need in a yaw control on the ground. But before we detect ground, it should be on, that's what i mean.
While touching ground with a high trimmed yaw PID the rotorcraft can become very unstable and roll over eventually. So getting P and D terms down while descending would be reasonable, IMHO.

[flixr]

Closing this issue, since it was improved quite a lot already with the heuristic using thrust, speed and acceleration.
Of course further improvement are still possible (e.g. full state machine)...