[Top][All Lists]

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

Re: [Paparazzi-devel] Change in altitude handling?

From: Felix Ruess
Subject: Re: [Paparazzi-devel] Change in altitude handling?
Date: Fri, 16 May 2014 15:25:18 +0200

In master I set a default INS_SONAR_MIN_RANGE with 1mm.

@Gautier, if you think this variance based option is not really useful, remove it...

On Sun, Mar 30, 2014 at 11:58 PM, Gautier Hattenberger <address@hidden> wrote:

Le 29/03/2014 15:26, Ben Laurie a écrit :

On 29 March 2014 14:03, Gautier Hattenberger
<address@hiddencom> wrote:
It looks like the data coming from the navboard of the ardrone contain a few
more information about the raw sonar measurements. Perhaps this can be use
to know if the distance measurement itself is valid or not. In the meantime,
it seems that removing too high and too low values is sufficient.
Another couple of things:

a) The variance is computed even if the value is out of range. Seems wrong!
I think it is better to have the input at constant rate.
In the end, it just makes the filtering a bit conservative since it will keep track of bad measurements during 10 samples in our case (update rate of sonar is around 10 to 50 Hz, so not a big deal).

Actually, I would like to remove this variance based filtering option:
- it is really not easy to tune
- it assumes that baro is "better" (less accurate, more noise but no outliers)
- it was design for a specific config (small custom quad, noisy maxbotix sensor) where issues were finally solved by changing the sonar (newer version of the sonar)

So, unless someone really can't live (or at least fly) without this 'feature', it should be removed.
Is anyone actually using this ?

b) Since it uses the baro measurement in the sonar calculations, it
would be helpful if it also sent the baro to the downlink.
see above

Also, if the sonar is in massive disagreement with the GPS, maybe it
would be a good idea to trust the GPS? :-)
Mixing GPS and sonar is actually not easy, unless your playground is flat. The current filter is using baro + sonar where the sonar is computing the baro offset. It means that if you fly at constant altitude, it is in fact constant AGL (if inside sonar detection range). Mixing the GPS, would not work unless we introduce a 'sonar offset' (which correspond to the ground elevation above reference point) or we have an accurate elevation map (SRTM is not good enough for that).

It is still possible to use GPS data instead of sonar in the same filter, and use the sonar 'directly' in the control as an AGL protection (like "don't go below x meters").

@Freek, did you ever looked at those values (debut_echo, fin_echo,
nb_echo,...) ?


Le 29/03/2014 11:42, Ben Laurie a écrit :

On 28 March 2014 10:33, Gautier Hattenberger
<address@hiddencom> wrote:
The telemetry mode can be changed during flight in the settings tab, but
define has to be added to the airframe file indeed (in the firmware
for instance).
The observation I made from your log are matching the observation I made
during my own flight, so we definitely need to correctly reject outliers
the sonar measurements.
I was looking at old logs, and they don't have sonar in them - I guess
this has been added relatively recently? Given how unreliable the data
is, not sure how useful it is.

Also, how sure are we that if I apply the thresholds it'll behave itself?

Le 28/03/2014 11:07, Ben Laurie a écrit :
On 27 March 2014 20:58, Gautier Hattenberger
<address@hiddencom> wrote:
baro_qfe is the reference pressure, so it only moves when the ins is
You should have the baro pressure in BARO_RAW message if you change
telemetry mode to raw_sensors. You can also add <define
name="DEBUG_VFF_EXTENDED"/> to enable VFF_EXTENDED message. 'measure1'
the sonar and 'measure2' is the baro.
You mean change it before flying, I presume? So this observation is
from your own ardrone2. not from my log?

Paparazzi-devel mailing list

Paparazzi-devel mailing list
Paparazzi-devel mailing list

Paparazzi-devel mailing list
Paparazzi-devel mailing list

Paparazzi-devel mailing list

reply via email to

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