paparazzi-devel
[Top][All Lists]
Advanced

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

Re: [Paparazzi-devel] (no subject)


From: Antoine Drouin
Subject: Re: [Paparazzi-devel] (no subject)
Date: 30 Apr 2004 20:48:12 +0200

Hi Chris

On Fri, 2004-04-30 at 14:58, address@hidden wrote:
> Are you interrested in enhanching your project using
> a HMC1055 (tree axis electronic compass + pitch + roll)
> in combination with a active distance measurement and barometric /
> pressure sensors. 
> 
> Regreads
> Chris
>  
> 
> 

That is exactly what we are trying to do.

Today, our autopilot is very simple: we rely on infrared only for
attitude (20Hz loop) and on GPS only for navigation (1Hz loop, hopefully
faster soon). We also use the GPS to tune infrared calibration during
flight. We rely on GPS only for altitude control. This solution works
surprisingly well, but has several drawbacks which we are trying to
fix.  

The problem we have noticed are:

  - when there is "alot" (7-8m/s) wind (our plane flies 10-12m/s), 
the ground speed changes quickly when going upwind to downwind and this
confuse the GPS (its filter lags...) this can lead to oscillations.
We hope to fix this using compass informations. we also want to try a
4Hz tim-lp on which we have more control on the dynamic of the filter.

 - when the sun is low, it give a false horizon to the infrared (tilted
in the sun axe). This leads to trajectories like the one of the trophee
(http://www.nongnu.org/paparazzi/images/screen_shot_trophee.png). This
is very obvious on the map (the sun was in the upper right corner - it
was early). We are working on detecting/fixing this effect. We also
think the compass data will improve things (at least allow calibration).


We have tested or are about to test the folloing sensors :

Magnetique:
we have a vector v2xe 2 axis compass. It is wired and almost ready to
fly. We still need to calibrate it. We also have a rotomotion 3 axis
unit. It is still in its bag. We will try it after the v2xe. For the
v2xe we use the hardware spi. For the rotomotion unit, I think we need
to use a software spi because the interface is 16bits.

We intend to fly the v2xe, get some measure and see how we can use then.
We will have to hybridize this data with the other we have (GPS +
infrared), so probably some sort of kalman filter. I think we don't have
the need nor processing power to run a full fledged one like in the
autopilot project.

For the calibration we will try hanging the plane level an rotating it.
We also want to try inflight calibration while doing constant bank
turns. This would also give precious information for estimating wind.


Airspeed:
We have build a sensor with a tiny propeller driving a cell phone motor.
We amplify and measure the analog voltage produced. The sensor works
fine for the small speeds we fly (8-15m/s).
We have done nothing yet with these data.


Pressure:
We have a MPX2100 sensor hooked to a 24bits ADC with integrated 128 gain
(ad7714). This ADC doesn't work very well and is in a bit dip case.
We are building a new unit with a MPXA6115 sensor (preamplified to 5V).
We intend to reamplify it a little and feed a 16bits AD7927 SPI ADC. 

Distance:
For distance measurements we have a MSU08
(http://www.lextronic.fr/Capteurs/PP1.htm). It's a 6 meter range unit
with i2c interface. We'll have to to i2c by software as the mega128
doesn't have such hardware. 
It is still in its bag...

Inertial: 
we have hooked the autopilot(TOKIN gyro + ADXL accel) inertial unit to
our v1.0 controller. We had data, but didn't manage to use them. We
tried the infrared solution just after and as it worked... we keeped it.
We want to try again with inertial data, probably when we have a 32
processor onboard.

I have looked at the HMC1055. Seems to be exactly what we need....
except there is some work to do with the analog part. Plus the result of
the accelerometer won't be directly usable for attitude when the plane
is accelerating (like when flying curves). So this brings us again to
some kind of filter...

But it would allow us to have full control on sensor calibration than
ready build units like the v2xe (plus add 3rd axis).

We have the requested connector to hook this sensor to the 1.2 board
(SPI+analog). We just need to build a board and write the code :)


We can have alot of data, I think the most difficult part is how to use
them. We will plug a cerfboard (or gumstix or whatever) to allow for
more processing power, but i think it's important for robustness and
popularity that the microcontroller board alone is able to achieve
autonomous flight. The mega128 still has alot of place and processing
power.

What do you think ?

Antoine








reply via email to

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