IMU or GPS for drift?

as we all know, wind drift is a problem when trying to get to our GPS coordinates. While in physics, for constant velocity drift, the equation is pretty easy, programming it to accomodate is a different story. while i haven't checked out, does the autopilot software being used by the geocrawlers adjust for that? Also, which is better to use to figure out the drift? An IMU is great, yes, but it isn't as accurate at predicting course errors. a gps has a 1sec refresh rate, but it atleast knows where you're going. which do you people use?

Views: 221

Replies to This Discussion

Excellent point. We do not have special wind correction in our ArduPilot code (or any of the others), so wind can cause us to deviate considerably from a straight line to waypoint path. But the code is open source and being constantly improved, so I imagine one of us will put it in soon enough. I'd be interested to hear what algorithm people favor for this function.
cool beans. To me, the easiest thing to do is a constantly shift- adjustment algorithm. I will get the equations for you in a little bit, but, for now, i'll label them A, B, C, etc for reference ease.

GPS way- USes vector based equations
You tell it to go to GPS waypoint. The plane should be under constant velocity, as it IS autopilot. Initial X,Y,Z coodindates are logged and a check XYZ coodinates is plotted. This Check XYZ is where the plane expects to be, and should be placed around 5 seconds after the Initial XYZ. Using a Vector Triangle **EQA**, you can find out the velocity of the the wind at that time in each axis. Using the compensation equation **EQB** you then replot the course, flying with, or against the wind. Once you attain the Corrected Heading, which will be a vector that will account for the wind, you restart the algorithm. This SHOULD accomodate for sudden gusts, as well as constant wind without really pressing any micro controller.

IMU way-acceleration based equations (A/V/S eqs)
This requires a timer and is a bit more processor intensive. Calculus is neccessary. From the starting point, there should be no acceleration. Constant velocity= 0 accleration. Any additional movement in the XYZ direction will have an acceleration associated with it. You must save and time every acceleration in the last 5 seconds. You can then find the velocity in all three axis by ading them together in their component form **EQC**. Create a vector triangle, then follow the process for the GPS way. However, the only change i would suggest is that the z direction be a bit mulled.

Overall, the IMU will be the most precise, but also the one that you have to be the most careful in your adjusments. the plane is small and light so almost anything will make the sucker move.

All algorithms assume you have a Co Pilot on board :-P. adding in a tilt is going to be annoying as all hell.
Easy, you use a 5Hz refresh rate GPS :), but a combination of both would give better accuracy at the cost of more complex code (and \$).
\$100 Gps...wow. At that price, i'd want the military precision accuracy instead of a faster refresh rate of a nice large diameter of where i should be in...

And yeah, the code gets harder if you do both

I don't think the speed of the GPS refresh rate is necessarily a need, unless this was for dogfighting or another task of high maneuverability or in tight places. I guess what i'm doing with the code is something close to taking a derivative of the path curve (and it WILL be a curve), and making sure it has the same end point. The larger the chunk fo resolution, the less you take the derivative and the more you just make a sectional average, which is more the value that i think should be aimed for.

lol, and i only ask, because, i have a 5 axis IMU, but i wasn't sure i wanted to use it. Based on some long thought, the GPS and the altimeter should be sufficient to plot the average wind velocity without being too sensitive and accounting too much for gusts, although correcting for it:)

BTW, these are just my thoughts, if you disagree, share your thoughts, believe me, i'm listening with an open mind:)
"\$100 Gps...wow. At that price, i'd want the military precision accuracy instead of a faster refresh rate of a nice large diameter of where i should be in..."

Till you can organise us some of those military GPS' I suppose well have to be content to know more or less were we are. I only have 1Hz gps' for now, but I think 5hz should help to hold heading without snaking too much to the target zone. I'm only using 8bit PIC's for my autopilot so can't put too heavy maths on it.

I have 3 GPS' and if I compare them against each other on different days they report the same location to within about 1.5 meters when they get 8 or more sats... that's good enough for me. Altitude is not so good though.

What altitude sensor are you using?
haha c'mon, i just did the on board DVR ;)

I am using the VTI SCP1000 as an altimeter.

I did some intensive research on GPS.

you don't want a military gps lol. instead of WAAS, they use, SAASM, which has been discontinued, so now they will use M-code. SAASM is fine, its just...well read...

http://en.wikipedia.org/wiki/SAASM

Wait till LAAM comes out.

i'm using a GM11. its the lightest gps that i have, with the largest antenna. Hell i don't even know if it still works lol. anyways, yeah, i can only hope that it has the 1.5m accuracy, although i think the most i've gotten is 4 confirmed sats.

My only reason for such a long wait time was so that it would reduce errors. Also, if your plane travels at 30mph, its moving a 44.4 feet per second. At 5hz, that's only 9 feet. at 2-5 meters accuracy, if you don't get 8 sats, OR aren't in the US (which makes your accuracy drop to some larger number).

While my 5 seconds was pretty arbitray, turns out WAAS correction signals are sent out every 5 seconds. I thought that was kinda funny, there was no point to it :)

Now this is where i've been getting confused through...
Apparently WAAS was intended to allow GPS to be used as a category 1 ILS system. Why can't we get an accurate altitude reading?
GM11 doesn't have WAAS but has SA. Interesting. Anywyas, i just completed Psuedo code for the GPS Windage correction algorithm. Maybe someone will be nice enough to write it out fully before i seek out a friend. if you do, message me!

1

2

3

4

5

6

7

8

9

10

Groups

3133 members

24 members

496 members

1045 members

• Cheerson CX-20 Big Shark

9 members

Season Two of the Trust Time Trial (T3) Contest
A list of all T3 contests is here. The current round, the Vertical Horizontal one, is here