Problem: The drone started taking off in LOITER (HOLD) mode even if on
ground and did not do a proper takeoff jump. The effect was mostly a
hovering close above ground or tilting/flipping because of integrator
wind-up.
Solution: Stay at minimum throttle until a takeoff setpoint is issued.
This adds the possibility to use offboard velocity control in the body
frame and not just the NED (world) frame.
The frame is set in the set_position_target_local_ned message and passed
on to mc_pos_control in the position_setpoint topic.
If velocity offboard control is used, it makes sense to lock/hold
position if the velocity input is 0. If this is not done, we will slowly
drift because nothing is integrating to keep the UAV at its position.
- fixed a bug where tilt compensation was done also when the horizontal
velocity was controlled. This is not needed because in this case
the controller outputs a 3D thrust vector.
Signed-off-by: Roman <bapstroman@gmail.com>
This resolves a bug where a takeoff would go sideways instead of
straight up. What happened was that the position setpoint got shifted
around even though there was actually no real setpoint set but only a
setpoint of type IDLE. This then lead to a position setpoint far away
from the takeoff point and therefore scary takeoffs.
This fix prevents the part of the position controller which
moves/integrates the position setpoint from running in the case of an
idle setpoint.
This bug could be reproduced by switching the vehicle to mission mode
without a valid mission, then switch to hold mode, and then send the arm
and takeoff command.
* Fix jmavsim HITL simulation of MAV_CMD_DO_REPOSITION in the case where you have no radio attached to the PX4 and so you have disabled RC link loss for that reason (set NAV_RCL_ACT = 0) but you still want the jmavsimulation to work. The line of code changed here causes failsafe RTL to kick in without this fix.
* Add altitude hold option using Z position control, while doing velocity control on vx and vy.
* Fix style and rebase issues
Since orb_copy is guarded with a orb_check, we ended up not having the
corrent landed state in the beginning which could lead to some
unexpected behaviour on takeoff.
It makes more sense to set the optimization flags on a platform basis
instead of individually for each module. This allows for different
optimization options for SITL, NuttX, Snapdragon, etc.
By running the loop in mc_pos_control without successfully polling on
position data, we still copy the manual input over so that
mc_att_control can consume it. This allows to fly in manual mode without
GPS and no position estimate.
Julian Oes