IMU? Aerobatics?

[shedao]

I'm curious, have any of you here has either used an existing IMU device like the 3DM from microstrain, or built their own device and then rolled their own kalman filtering software?

For the UAV I'm building I need more precision that just a GPS/wing-leveler can provide, it will be aerobatic and precise (as can be) state-control will be needed. Is it's worth the time/effort trying to roll my own versus just buying an existing IMU?

Also, have any of you ever attempted to create an aerobatic UAV (while autonomous)? I posted this on another forum as some might already know but I figured why not ask in here also.

[kartoffel]

How much processing power do you have onboard, and what kind of weight limits are you working within? You might consider a 3 axis solid state gyro setup.

I've got a small wheeled robot that uses a PNI TCM2-20 magnetometer. It takes 16 samples per second and talks RS232 to an onboard computer, and it works just great for what we need to do. Check out hobbyist gyros for aircraft and helos, but also check out magnetometers. PNI magnetometers (http://www.pnicorp.com/supportGroup?nodeId=c4b)

[shedao]

I'm still deciding on the processing power (this is all "in the works") but I'll most likely go with a 32-bit CPU like the Intel PXA 255 for outer loop and 8-bit MCUs like the AVR ATMega series for servo control and sensor data gathering.

I'm still designing the whole thing so what I'm trying to find out is the best way to determine aircraft state given the relatively low tolerance for error when doing aerobatics. I want to determine what i'll need computer wise then find an airframe that can carry it.

This is just a hobby UAV though, just to make a few people smile. Nothing serious. Never out of eye sight.


What concerns me about the PNI magnetometers is their tilt ranges. They say that the TCM 5 has a title range of 90 degrees pitch and 180 degrees roll but they also say:

"Accurate orientation determination of the module using Euler angles. This method is the same used in aircraft orientation. The output is Heading (Yaw), Pitch and Roll so the module can be used in systems requiring full 360° rotation."

What would this sensor do if it was oriented outside of the specified tilt angles (say the aircraft is upside down)? Hrm, I'm gonna have to dig further to find out.


At $1600 it's close in price to the high end IMU from Microstrain, the 3DM-GX1
http://www.microstrain.com/3dm-gx1.aspx



Thanks kartoffel!

[shedao]

Oh one more thing, I did see that the winner of this year's IAUV competition, the subjugator (http://subjugator.org/) uses the TCM5 for orientation while in the water so it may be enough.

[Bg~]

The pitch and roll limits they mention might just be due to the numerical limitations in using Euler angles. The Euler angle convention commonly used with aircraft has a singularity in pitch at 90°. I'd guess the output angles would just roll over from +limit to -limit. So 91° pitch would come out as 180° roll,
+89° pitch, and 180° yaw if I did it right. (just a thought though, i didn't read too much of the manual...)

[LukeZ]

shedao, you might check out the 6-degree of freedom unit (http://www.sparkfun.com/shop/index.php?shop=1&cart=346965&cat=71&) sold by Spark Fun (you'll have to scroll down a bit). It incorporates Analog Devices solid-state gyros and accelerometers, and there are several configurations you can choose for your needs. Prices vary but the most you'd spend is about $400 - you can even get it in a wireless version. But more practical would be to just use the serial output.

This is probably what I'll use for my UAV. It won't be aerobatic because I don't have a need for it, but I don't think this IMU would be a limitation if that's what you wanted.

The real challenge, it seems to me, is coding your Kalman Filters and what not. There are not a whole lot of other people who have done this that are sharing. The Autopilot project (http://autopilot.sourceforge.net) is one exception; however, their filters and other software were developed for a heli platform, and I understand this is quite a bit different from a fixed wing project. The Paparazzi (http://savannah.nongnu.org/projects/paparazzi/) project was a fixed-wing spinoff of the Autopilot project, but they have not used state estimation thus far, prefering rather the now-ubiquitous horizon detection method - which has its limitations, and may not be the best approach for an aerobatic ship: though I really wouldn't know for sure.

Anyways, most of this you probably already know. If you do have success with your Kalman/state estimation efforts, and want to share them, you'll certainly find quite a few open ears here in this forum...


Luke

[shedao]

Thanks guys for the input. LukeZ, yeah I do plan to make it all open source. I'm gonna try and pull as much as I can from those projects listed.

Even if the IMU has a built in filter I will probably still need to write my own to filter the IMU data with the GPS/alti/airspeed data and then get my approximate state from that.

Since it's just for fun I don't have much on the line, just my time and apparently tons of my money. But don't expect much soon, I see this as a 3-5 year project at the very least. But anything I come up with I'll share. Even if it's just my "here's what _doesn't_ work" information.

[shedao]

The pitch and roll limits they mention might just be due to the numerical limitations in using Euler angles. The Euler angle convention commonly used with aircraft has a singularity in pitch at 90°. I'd guess the output angles would just roll over from +limit to -limit. So 91° pitch would come out as 180° roll,
+89° pitch, and 180° yaw if I did it right. (just a thought though, i didn't read too much of the manual...)


Ahh ok, so the 3 angles up to 90 degrees can be used to determine 3D state, XYZ. I'm still learning a lot of this so bear with me. But I did find a few good links to Euler angles in case somebody comes along:

http://www.euclideanspace.com/maths/geometry/rotations/euler/
http://www.euclideanspace.com/maths/geometry/rotations/conversions/angleToEuler/
http://mathworld.wolfram.com/EulerAngles.html

[typicalaimster]

What your working on sounds alot like the GW-UAV project. Just wonder if it's the same project?

[shedao]

No, this is just a hobby project of mine, but if you point me to the GW project I'd like to check it out. Or any info for that matter.

[typicalaimster]

GWU's UAV program is doing this thing where one UAV chases another UAV. The tailing UAV is supposed to mimic the leading UAV. This involves various manuvers, loops, rolls, etc. That's why I was curious. I wish I had a link to their project.

[shedao]

Wow that sounds pretty advanced. I'm gonna have to see what I can dig up from GWU. I'd love to see that in action or any research they've done.