IMU question

[EddieWeeks]

I have used one of these for over a year... Works great..
I only use it for heading but it also outputs pitch, roll..

http://www.ocean-server.com/compass.html

Can someone explain the difference between this and a full IMU...

Thanks a bunch..

Eddie Weeks

attached... one of my older planes..
video_www.youtube.com/watch?v=kRI_VuptnGY

[rbeall]

I didn't look at it too closely but this looks like its just a magnetometer with an accelerometer (both 3 axis of course). The underlying problem here is that this unit will only work when the vehicle is not linearly accelerating. The unit does some trig to correct the magnetic flux measurement with respect to the 1-G down gravity vector. Heading is always horizontal to ground so in short the accelerometers find the ground reference and then back out the heading from the three flux measurements.

That being said this unit will get all confused in turns due to the added centripetal acceleration. A traditional IMU has gyros and some filtering to back most of this out and continue to lock onto pitch roll and yaw even when accelerating.

I hope this helps feel free to ask more if need be

[EddieWeeks]

So the reason it will not work is because its just a fancy tilt sensor with a compass ???

I ran this thing (Ocean Server Compass) for over a year in an Unmanned boat...
Hundred of hours watching a computer screen of the Nav program I wrote...
and it just seemed like it would work for an airplane because it worked so well in a boat.. I mean the boat would pitch and roll a lot and at high rates...

But since the pitch and roll would allways avg zero, is that why it worked so well.. ?

Eddie

Thanks for your help

[mikel]

I'm not an expert in IMUs, but...

Averaging to zero is only one side of why it worked so well.
You probably didn't use this "compass" for pitch and roll estimation since you don't usually control pitch or roll of a surface boat - it levels itself when it stops turning.
If you limit the speed and turn rate of the boat it will never pitch or roll as much as a plane could. It stays level most of the time and your compass can determine the heading accurately (on average).

If you have a plane that always levels itself like a boat, this "compass" would probably work fine for HEADING estimation. Some relatively simple UAV designs use IR-sensors to stabilize roll and pitch, i.e. keep them at 0. These UAVs are somewhat similar to boats - they always fly level and navigate between waypoints using only one control surface (rudder) to change heading. Assuming that UAV is almost always leveled, "compass" can be used for heading estimation and accuracy should be pretty good. However, IR-sensors get confused by obstacles (trees, houses, etc) and bad weather (fog, rain, snow).

If you want to control the UAV in bad weather and be able to precisely control pitch and roll of the plane, you need a full IMU that provides good pitch/roll/heading estimation regardless of the plane position, attitude and acceleration. Such IMUs usually use 3-axis gyroscopes, accelerometers and some filtering. GPS data can also be used to compensate for accelerations in long turns.
Combination of compass (magnetometer) and accelerometers is usually not enough to get accurate and timely pitch/roll estimation when plane is not leveled by other means.

Don't get me wrong, magnetometers and accelerometers alone can be used for pitch/roll estimation, but results are usually not accurate and fast enough to control a plane. If there is no wind, your plane has a strong self-leveling tendency (high wing, dihedral, etc) and very slow pitch/roll rates, you might get away with "compass" only.

[rbeall]

This is all true. It worked well in the boat because the acceleration caused by turning or centripetal acceleration was low. Airplanes can achieve significantly higher latteral g's which significantly would affect the estimation. The basics of how tilt compensated magnetometers work is as follows.

Three axis of acceleration fall into some trig equations which rotate the measured magnetic flux on three axis. The math all assume that gravity is 1 G and always pointing down. Big problems arise clearly when the aircraft is in a two G turn. The equations can no longer determine the 1 G down vector because it has some arbitrary 2 extra Gs affecting on some other unknown axis.

The magic here is that the centripetal acceleration can be subtracted out if you know your body rates and velocities with respect to the inertial frame. Basically a couple of gyro readings multiplied by the speed over ground from gps.

[EddieWeeks]

ok...very good explation... rbeall and mikel.... Thanks a bunch

This told me what I need to know... so... can you point me to
a full IMU that does not cost $4000... hahah

Thanks again..

Eddie Weeks
http://aerialphotobatonrouge.com/

[mikel]

I don't think there are any cheap IMUs that can be simply plugged in and used for attitude estimation without advanced filtering.
There are some hardware 6DOF IMU platforms that are essentially 3 gyroscopes and 3 accelerometers in one package. They might have additional sensors for temperature, pressure, wireless interfaces, etc. The price is about $300 - $600. Sparkfun.com has some in IMU/Gyro section. Analog Devices has new ADIS1635x chips - 3 gyro and 3 accelerometers in one package plus temperature calibration. Simple evaluation board is about $500.

However, all of them are just sets of independent sensors that cannot be used directly to measure pitch/roll/heading. Some advanced digital filtering/processing is required to combine data from these sensors and get rid of the errors. Without filtering they will likely work worse than IR-sensors. People use relatively simple complementary filters, more advanced Kalman filters (EKF, UKF) and other. These filters might get really complex and often require fast MCUs with floating point support. I'm planning to buy ADIS16354 6DOF sensor and implement a UKF filter using ARM9 MCU. It will probably take more than 6 months and I'm not even sure I can get it to work right.

IMUs with built-in filtering would cost much more than $500. I don't know if anybody sells them.

[EddieWeeks]

I found this one for 1k...

http://www.motionnode.com/

is it a full IMU... I think it is..

Eddie

[mikel]

I found this one for 1k...

http://www.motionnode.com/

is it a full IMU... I think it is..

Eddie

It is a combination of 3-axis gyro, accelerometer and magnetometer with usb output and software library.
They mention some calibration and "filtering pipeline" used in one of the modes of operation, but there is no details on what kind of filtering it is.
On another page they say that magnetometers and accelerometers are used to estimate full 3D position which tells me it's not a full IMU we were talking about earlier.
They don't say what sensors they use.

This device doesn't look any better than other $500 hardware IMU platforms. I doubt there is anything like Kalman filtering in it.
It should be easy to use under Windows since it has a USB interface and software library. Most likely you would need to use raw sensor outputs and implement digital filtering (Kalman, etc) yourself.
I might be wrong, but it doesn't look like a "full IMU".

[Tuner]

Atair Aerospace has a nice INU witch is an IMU with GPS and airpressure sensor and all the data is integrated using a Kalman Filter.
This cost me $2500 the price may be less with their newer version??

I have yet to use it but it was the best bang for the buck I could find and still be lightweight.

http://www.atairaerospace.com/circinus/