[Canberrauav] flight helis at NAAS yesterday
andrew at tridgell.net
Mon Jun 15 08:26:20 AEST 2015
We flew Gregs Trex500 heli at NAAS yesterday, flying 5 flights. The main
aim of the initial flights was to fix a yaw control issue from previous
flights using changes that came out of log analysis by Rob and myself.
The yaw tuning changes worked extremely well. Within a minute or two of
takeoff on the first flight we had very good yaw control. The heli
reacted to rudder control precisely and quickly. It was great to finally
have yaw under control.
The next 3 flights were uneventful. Darrell and myself both had a go on
the sticks (using a buddy box setup with Greg on the main controller),
and we did a short auto demo flight for the visitors at NAAS. The heli
flew extremely well, with precise control and very neat landings.
The final flight was not so good! Greg and I wanted to fix an issue with
pitch control. The copter was flying missions very well, but only at low
speed (5 m/s). The logs showed that the heli was compensating for poor
pitch tuning using the higher level velocity controller. That is fine at
low speed, but we were concerned that if we tried to fly at higher
speeds that it could become unstable.
We had also noticed that there was tail vibration when the battery was
fresh just as we took off on each flight, but a minute or so into the
flight the tail vibration vanished. We thought this was happening due to
a resonance at a particular main rotor speed, so we decided to try
changing the main rotor speed requested of the governor by a few percent
to see if it solved the issue.
After takeoff I first lowered and then raised the governing speed, and
we found that by raising the governing speed by 5% the vibration
stopped. We would need more flights to confirm it, but it did seem we
were on the right track.
Next we tried raising the pitch tuning gain to see if we could get
better pitch tracking. About 3 seconds after raising the gain in 3 steps
the copter became unstable, oscillating rapidly. I lowered the gain
immediately, and the oscillation stopped, but the heli started flying
off in the wrong direction quickly. The logs show that the oscillation
had caused the accelerometers to clip (recording peaks of 24g on the
lsm303d) and this caused the EKF to diverge. The heli had lost its
position and attitude solution and was out of control. Despite best
efforts by Greg it crashed hard into a tree and was destroyed.
Full logs here:
I am now analysing the logs in more detail with the following aims:
1) see if anything can be done to prevent the oscillation we saw
2) change the EKF to detect the accelerometer clipping and to stop
fusing the inaccurate accelerometer data
3) possibly fall back to GPS velocity and position, ignoring the EKF,
if the EKF diverges
More information about the Canberrauav