It's enough that the setpoints and the unallocated values are logged, from these
the allocated values can be calculated if required.
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
It's very important that the in descend detection is always
at a strictly higher velocity than the vertical movement check.
This combination is basically used to check for vertical downwards
velocity tracking. Desired descend, no movement -> ground
If in descend threshold is lower than vertical movement it is
by definition even with perfect tracking the case that with
any velocity between the two thresholds there is no movement
even though a descend is commanded.
See first fix of this problem #7831
e39b38ba96971245aaf6d2b1c249868c8717665e
I don't see where this is necessary. During takeoff, the maybe landed flag should
only get cleared once system is about to takeoff, and thus well after the spool up
is complete (for which the higher thresholds are meant in this case).
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
Don't consider these params for vertical speed threshold,
and instead increase the default for LNDMC_Z_VEL_MAX and
use solely that one. Makes the land detector outcome more
predictable and its interal logic simpler, while the new
default tuning is resulting in about the same vz threshold
as before.
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
before there was a corner case where if in an auto mode that sets a reset command on the attitude setpoint message (e.g. auto takeoff), if the mode was then switched stabilized, this reset bool would never be changed back to false and the integrators would reset every cycle
This was previously required to reset the flight speed after a VTOL transition,
but is now no longer required as the DO_CHANGE_SPEED commands are handles directly
in the controllers.
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
This fixes the issue where the init happended in the initializer list, at which point
the params were not yet initialized and thus resulted in random values for the init
values of _mission_item.loiter_radius and .acceptance_radius.
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
Refactor
- Require reboot for PD_GRIPPER_EN parameter change
- Define gripper ACTION_NONE for readability. This makes implicit assumption that -1 equals no-action commanded more explicit
- Tidy the scattered vcmd_ack struct handling cases into a single function
- Refactor to remove return in the middle of function: avoids future complications where a programmer may expect the logic at the end of the function to be executed, but isn't
Vehicle Command Handling
- Cancel the previous running vehicle command if we receive a different vehicle command
- Reject vehicle command if we get a same one that is getting executed
- Save the source system & component of currently running vehicle command
- Added note about new discovered edge case of having same entity sending different gripper commands consequently, where an unexpected ack result may be received
Gripper:
- Don't command gripper (via uORB `gripper` topic, which maps into an
actuator via Control Allocation) if we are already at the state we want
(e.g. Grabbed / Released) or in the intermediate state to the state we
want -> This prevents spamming on `gripper` topic
Payload Deliverer:
- Add read-once function for Gripper's released / grabbed state
- Send vehicle_command_ack for both release/grab actions.
TODO: target_system & target_component for the released/grabbed vcmd_ack
is incomplete, since we are not keeping track of the vehicle_command
that corresponds to this. This needs to be dealt with in the future, not
sure what the best solution it is for now.
Possible solutions:
- Queue-ing the vehicle command?
- Tying the gripper's action to specific vehicle command one-on-one, to make sure if we send multiple vehicle commands, we know
which command resulted in the action exactly?)
Only command Gripper grab when we are actually initializing gripper
- Previously, on every parameter update, gripper grab was being
commanded
- This commit narrows that scope to only when we are actually
initializing the gripper
Handle gripper de-initialization upon parameter change
- Also added some local state initialization code to init() function of
Gripper
- This will now make init / de-init more graceful & controlled compared
to before
- This hopefully then alerts the GCS that the command is getting
processed
- Referenced commander's `handle_command` function to implement this. As
it seems like GCS needs the acknowledgement of the command being
processed to execute such commands properly
- Also send FAILED command ack if we can't actuate the gripper
Fix wrong GRIPPER_ACTION conversion from floating point to int32_t
- Due to the MAVLink spec, we actually just convert enums into floating
point, so in PX4 we need to convert the float directly into integer as
well (although there can be precision issues on large numbers)
- This is a limitation in MAVLink spec, and should hopefully be
changed in MAVLink v2
Affects the states AUTO_ALTITUDE and AUTO_CLIMBRATE. Those modes should only be entered
if armed (as they are pure failsafe modes). Also allow though to enter them even if
the position setpoint(s) are invalid, as they are not needed.
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
Instead of checking the .valid flag of position_setpoint, check for ISFINITE() of lat, lon, alt
when pulling the position_setpoint triplet. This fixes problems where the .valid flag didn't
reflect the proper state of the setpoint (e.g. .valid was true, .lat though NAN)
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
- vehicle_angular_velocity: ESC RPM notch filters minimize filter resets
- only allow one filter init per axis per cycle
- "park" ESC notch filters at min frequency instead of full disable
- relax timeout before a notch filter is disabled
- add new parameter IMU_GYRO_DNF_MIN for configuring the minimum notch filter frequency
- avoid the possibility of unsigned underflow from subtracting two HRT timestamps (uint64_t)
- most of these aren't problematic, but people tend to replicate the pattern, so it's better to be safe
- likely wasn't a problem when people were using hrt_absolute_time() in place, but if using an existing timestamp there's the possibility it's older than a more recent topic update
- update all msgs to be directly compatible with ROS2
- microdds_client improvements
- timesync
- reduced code size
- add to most default builds if we can afford it
- lots of other little changes
- purge fastrtps (I tried to save this multiple times, but kept hitting roadblocks)
- working towards keeping all height source (baro/ev/gnss/rng) handling as consistent as possible, possibly refactoring these out into separate classes later
