manual control selection and cleanup #17404
Commits on Nov 8, 2021
-
-
-
manual_control: add selector class [WIP]
This adds a selector class with unit tests. The idea is to have a valid flag in manual_control_septoint and set that according to the selection and/or timeout of manual_control_inputs.
-
rc_update: rename update method
This reflects better what it actually does.
-
-
manual_control: add instance to selector
This instance is then used to schedule the callback for the subscription.
-
-
manual_control: move override detection
This also removes the option to ignore throttle for the override detection as it's not really required anymore.
-
-
-
-
-
commander: move main_state_transition out
This is a first step to having a desired main state.
-
commander: only move to POSCTL when armed
I don't think we want to switch to POSCTL when the sticks are moved when armed.
-
manual_control: send initial mode in the beginning
In order for commander to know the desired mode we need to send the initial position of the mode slot.
-
commander: move valid mission check inside
I think it makes more sense to move this check inside of main_state_transition.
-
commander: always reset posvel validity
I assume reseting posvel validity is ok for any case, not just all cases except manual, acro, stabilized.
-
commander: add desired main state
This is an intermediate solution to carry forward the initial state of the mode slot. Basically, it allows that we start up in Stabilized but switch to POSCTL as soon we have the required GPS.
-
commander: move manual_control and switches out
This moves the remaining handling of the manual control stuff out of commander. All communication between manual control now goes through vehicle commands, and the landing gear topic.
-
manual_control: use filter to check user override
Instead of only looking at the diff between two RC samples to decide whether a user wants to override, we now look at the filtered diff over one second. This should be more robust to RC sent at various or varying rates.
-
commander: only override according to param
This got lost in the refactor earlier.
-
manual_control: ORBIT and PRECLAND not supported
ORBIT and PRECLAND are not actually possible on the flight mode switch.
-
manual_control: support arming button
The arming button required some refactoring in order to support to toggle arm/disarm using the vehicle_command. Otherwise manual_control would have to subscribe to the arming topic and we would spread out the logic again, and risk race conditions.
-
-
-
-
msg: re-use manual_control_input in setpoint
This way we avoid duplication between manual_control_input and manual_control_setpoint.
-
-
-
manual_control: add reset to diff
This should prevent triggering user override right after RC re-appears.
-
manual_control: update first, get diff after
Not sure why I had this the wrong way round.
-
-
-
Better checking once too much than once too little.
-
-
ManualControlSelectorTest: add constantly outdated samples case
Reviewing the code I saw this case arising from the order things get checked.
-
-
-
manual_control: don't switch gear on first init
This will ignore the first transition from NONE to something. This matches the behaviour that existed before.
-
-
-
-
-
-
-
-
mc_rate_control: always allow landing gear control
Except of course when landed and we try to put it up.
-
commander: remove unused mode/main_state methods
With the previous changes these seem no longer required.
-
commander/manual_control: use msg enum for params
Instead of using a private enum class we should define the enum in the vehicle_command message and then use it consistently.
-
-
-
-
state_machine_helper: rever explicitly handling unavaialble mission
It was already handled by the existing condition.
-
-
ManualControlSelector: copy manual_control_input struct instead of ea…
…ch individual field
-
Commander: don't publish RC_IN_MODE to vehicle_status
This just contains the content of the parameter which is redundant and results in multiple sources of truth.
-
-
ManualControl: fix arm button logic during testing
The latching logic has to be different than for the arm switch this difference got lost in the rewrite of the old obfuscated arming trigger logic.
-
ManualControlSelector: simplification refactoring
- clang tidy did't like {bool} == true in a condition - setpoint_from_input() got small enough ot just put in line -
-
-
-
Replace arm_request and mode_request with combined action_request
Which saves flash space, log size and is extensible to handle e.g. the VTOL transition and future actions.
-
Commander: fix "Manual control lost" showing double in QGC
because of the missing tab in the deprecated message.
-
-
-
-
-
-
-
-
-
-
-
-
