Plane: ignore invalid pilot throttle

[IamPete1]

Extract from #25416 superseding #25631

This now always ignores pilot throttle if there is invalid RC. Currently on master you can get stuck at your last throttle in some cases with guided or suppressed passthrough and manual mode. This should be covered by the set_control_in call here:

ardupilot/ArduPlane/radio.cpp

Line 278 in bf6bd3a

channel_throttle->set_control_in(0);

However that does not fix all cases.

If we call get_throttle_input(true) to get the no deadzone value we call get_control_in_zero_dz:

ardupilot/ArduPlane/reverse_thrust.cpp

Lines 123 to 139 in bf6bd3a

	/*
	return control in from the radio throttle channel.
	*/
	float Plane::get_throttle_input(bool no_deadzone) const
	{
	float ret;
	if (no_deadzone) {
	ret = channel_throttle->get_control_in_zero_dz();
	} else {
	ret = channel_throttle->get_control_in();
	}
	if (reversed_throttle) {
	// RC option for reverse throttle has been set
	ret = -ret;
	}
	return ret;
	}

However, get_control_in_zero_dz does not use control_in it uses radio_in so our set_control_in call is bypassed.

ardupilot/libraries/RC_Channel/RC_Channel.cpp

Lines 391 to 397 in bf6bd3a

	int16_t RC_Channel::get_control_in_zero_dz(void) const
	{
	if (type_in == ControlType::RANGE) {
	return pwm_to_range_dz(0);
	}
	return pwm_to_angle_dz(0);
	}

ardupilot/libraries/RC_Channel/RC_Channel.cpp

Lines 361 to 379 in bf6bd3a

	/*
	convert a pulse width modulation value to a value in the configured
	range, using the specified deadzone
	*/
	int16_t RC_Channel::pwm_to_range_dz(uint16_t _dead_zone) const
	{
	int16_t r_in = constrain_int16(radio_in, radio_min.get(), radio_max.get());
	if (reversed) {
	r_in = radio_max.get() - (r_in - radio_min.get());
	}
	int16_t radio_trim_low = radio_min + _dead_zone;
	if (r_in > radio_trim_low) {
	return (((int32_t)(high_in) * (int32_t)(r_in - radio_trim_low)) / (int32_t)(radio_max - radio_trim_low));
	}
	return 0;
	}

The problem with get_adjusted_throttle_input is that is uses norm_input with again bypasses control_in

Recreation:

Turn off short fail-safe (FS_SHORT_ACTN 3)
Arm and go to full throttle in manual.
Have a RC fail-safe, (SIM_RC_FAIL 1)

Master:
Throttle remains at 100% until long action.

PR:

Throttle goes to 0 as soon as RC is lost, throttle is output again once in RTL.

The same issue also affects roll/pitch/yaw inputs in some modes....

[IamPete1]

Now were handling it in the get_throttle_input and get_adjusted_throttle_input calls we don't need this check.

[tridge]

it would be good to test QSTAB, QACRO, QHOVER, QLOITER, make sure they don't lose height during the short rc failsafe period

[IamPete1]

Added a commit to be sure that quadplane behaves the same, it is a re-implementation of the no deadzone case of get_throttle_input returning the set control in value rather than 0 in failsafe. These are exclusively used when comparing throttle to some threshold for things like throttle wait and suppression.

Also re-checked with SIM_RC_FAIL 2. Behavior is the same.

Tested Q modes to make sure we don't fall out of the sky.

Qstab and Qacro we get hover throttle.

Qhover and Qloiter a slow descent.

[IamPete1]

This is almost certainly wrong for Quadplanes, but it is the current behavior.