Copter: fix compilation when drift mode is disabled

[peterbarker]

I have only validated that this compiles.

This should be NFC (apart from actually compiling...).

I think there's a wider discussion to be had on when we should allow these mode switches. Depending on how aggressively it is tuned, popping the vehicle into several other modes may cause it to leap off the ground here.

e.g.:

STABILIZE> mode guided
STABILIZE> Got MAVLink msg: COMMAND_ACK {command : 11, result : 0}
GUIDED> Mode GUIDED
rc 3 2000
GUIDED> param set FS_GCS_ENABLE 0
GUIDED> arm throttle
GUIDED> APM: Arming motors
Got MAVLink msg: COMMAND_ACK {command : 400, result : 0}
ARMED

GUIDED> mode loiter
GUIDED> Got MAVLink msg: COMMAND_ACK {command : 11, result : 0}
LOITER> Mode LOITER
APM: EKF2 IMU0 in-flight yaw alignment complete
APM: EKF2 IMU1 in-flight yaw alignment complete
height 15
height 25
SIM_VEHICLE: Keyboard Interrupt received ...
SIM_VEHICLE: Killing tasks

[peterbarker]

... I should note that our autotest suite covers getting a failure back in the case this is checking.

[rmackay9]

maybe we can just change drift mode to return true to the "has_manual_throttle" method?

EDIT: to answer my own question, it's tricky to just change Drift's has-manual-throttle to return true because we have these lines in mode.cpp's exit_mode() method. I think this will not calculate the I-term correctly. So I guess in short my suggestion would need to be checked very carefully.

// smooth throttle transition when switching from manual to automatic flight modes
if (old_flightmode->has_manual_throttle() && !new_flightmode->has_manual_throttle() && motors->armed() && !ap.land_complete) {
    // this assumes all manual flight modes use get_pilot_desired_throttle to translate pilot input to output throttle
    set_accel_throttle_I_from_pilot_throttle();
}

[OXINARF]

LGTM.

I think there's a wider discussion to be had on when we should allow these mode switches. Depending on how aggressively it is tuned, popping the vehicle into several other modes may cause it to leap off the ground here.

@peterbarker I want your ok_to_enter/enter methods idea back! 🙂

[peterbarker]

On Wed, 21 Aug 2019, Randy Mackay wrote: maybe we can just change drift mode to return true to the "has_manual_throttle" method?

I did consider that - but it really isn't the same thing as stab/acro as drift mode is much more like loiter that stab or acro (a velocity-based controller). My initial cut of this patch added a "has_user_throttle()" method on each mode which returned true if the user's input contributed to the throttle output (so auto mode and guided mode would both return false (not counting auto-launch stuff)). It worked, but I decided that that wasn't my fight at this point, and just wanted the build fixed with drift disabled :-) Also, greping for has_manual_throttle shows a rather large number of points where we would have a behavioural change for drift - that would be a lot of testing and thinking to merge IMO.