[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Paparazzi-devel] hff filter in ins subsystem

From: Loic Drumettaz
Subject: Re: [Paparazzi-devel] hff filter in ins subsystem
Date: Tue, 11 Jun 2013 16:49:22 +0200

"I think it may be related with the speed and position estimates of the hff filter because i see there is a huge noise on the speed and position signals."

Did you actually measure the speed and positions estimates on ground, with motors running? (quad fixed on ground) Are you sure oscillations in flight come from HFF horizontal position estimates?
From my experience, there are actually several reasons to experience position or speed oscillations in flight:

1- Bad tuning of horizontal control loop (this is not your case if you have already tested several gains)

2- Attitude estimation dynamic errors. AHRS filters use accelerometers readings, which provide an accurate angle only when there are no horizontal accelerations. When the quad reaches a waypoint, it decelerates, the attitude angle estimation gets wrong, leading to position and speed oscillations.
If you use int_cmpl_quat filter, you may be interested to try the new version of the filter:
This filter is being validated at the moment and gave good flight results, and to be more specific, it solved most of our rotocraft speed oscillations. Please be aware that it's experimental and not tested yet with high level of vibration or gyro biases.

3 - Horizontal acceleration can be reduced using the reference generator. This also reduce attitude estimation dynamic errors, which reduce speed oscillations. You may try:
REF_MAX_ACCEL = 1 m/s2
REF_OMEGA = 18 deg/s
as an extreme smooth navigation setting...

4: GPS speed and position estimates oscillations. If GPS accuracy is bad, you won't get stable flight. Less than 1 m/s speed accurracy (SACC in GPS_INT message) is good enough.

5: HFF filter: so far I haven't been able to find tuning that filter efficiently GPS errors. Note: if you remove it, the position and speed estimates will be sampled at 4 Hz (GPS frequency). This causes spikes in position error (reference - estimates) as a result, the horizontal loop output, the attitude setpoint show the same spikes... So the HFF filter helps...

Sorry, I may have been completly off topic, but your problem reminded me something that i already experienced, so i thought it would be nice to share...


reply via email to

[Prev in Thread] Current Thread [Next in Thread]