I would not mind testing that, now that
my airspeed is working. (Ok, theoretically working. I never flew
it yet.)
Anything I should look out for? Mitchell or Christophe, can you
perhaps share a airframe file for 'inspiration' (cargo cult
configuration)
Thanks
Tilman
On 19/09/12 20:01, Mitchell SLOWER wrote:
Dear all,
Regarding the recent problems I had with the the standard
control loops, and from Gautier and Christophe's advices, I
tried today the total energy control loops.
I pulled from tudelftv4.0; it was straightforward: I had a
really successful flight on the first try.
I have made only a few flights in calm weather, but all I can
say for now is that this controller is really good.
No more diving after steep climbs, strong and precise
altitude hold, even at high climbrates (+- 15m/s).
The UAV has clearly better flight performances with a very
quick setup with the total energy control loops than with the
old standard control loops with hours of tuning.
Basic setup was very easy, I haven't tuned yet the extra
gains (auto throttle to airspeed, and auto pitch to airspeed).
Thanks a lot Christophe (and TU Delft).
The only "problems" I had were the following:
Error: subsystems/navigation/gls.c:107: error:
'V_CTL_AUTO_AIRSPEED_SETPOINT' undeclared (first use in this
function)
Code in gls.c #define APP_TARGET_SPEED
V_CTL_AUTO_AIRSPEED_SETPOINT;
=> V_CTL_AUTO_AIRSPEED_SETPOINT is unused with the total
energy control loops; but is required for the landing procedures
(gls.c)
The other problem is that the climb rate is not bounded
during flying (maybe I did something wrong).
In energy_ctrl.c me max climb rate is set to 2m/s
(v_ctl_max_climb), but the plane was climbing at +/- 15m/s.
Thanks again,
Mitchell
2012/9/19 Christophe De Wagter
<address@hidden>
1) -> Felix?
2) one for me: energy_ctrl apparently needs an
upgrade for the state interface. thanks for telling
me.
3) all your flightplans must be updated to the new
state interface too. replace all estimator_ stuff with
the new: GetPosAlt(), ground_alt
and PowerVoltage(), GetPosX() etc. Or maybe
start with an default flightplan.
We did all tests in the tudelft/paparazzi branch
which is a v4.0 clone with several extra features like
energy, rate limited flaps, dual-lisa-classix, etc...
All these features are being pushed to
paparazzi/master but do not necessarly work there yet.
So it is probably way less work to try tudelft repo
directly:
git fetch tudelft
in your v4.0 based branch:
git pull tudelft tudelft4.0