Currently I am working on my first UAV so I am only a beginner. I understand the basics of IMUs and some of the challenges but do not yet fully understand the code. It occurred to me that we all wanted good IMUs in order to get rid of thermopiles and make everything nice and neat but what about people who just want to replace their old FMA copilot with something nice and neat? Has anyone made an IMU device you can just stick in line with your current RC gear and have it keep your wings level. Something you only have to read the manual once on, you know, so the average RC guy could understand and use it.
You need to be a member of diydrones to add comments!
We had originally thrown this module together: Analog Node
To use with our CHIMU AHRS to do what you are talking about. Its a little pricey for this group (the unit uses magnetometers instead of GPS for yaw) and no open source, but...
Here is a picture of it next to an FMA module:
As for the speed question, we allow an analog input of voltage to accomodate corrections for this if you have a pitot tube speed measurement signal available. However, both experience (see Chris's reply) and the math say that you can get by without speed input if you aren't planning on threading any needles :)
The quick math version is that your attitude estimate will be corrupted by centripetal acceleration if not accounted for. Centripetal acceleration is dominated by your forward velocity, and it only couples into Y and Z accelerometer errors via yaw and pitch rates (respectively). Since (typically) body z and y velocities for a plane are centered about zero, your y accelerometer will see just your yaw rate* body velocity as an error.
So quick, high rate motion could produce high errors (for a constant speed), but those typically don't last long, and if they do then the pilot is probably pretty good and doesn't need a wing leveler. Further mitigating the error build up is what Brian said basically, which is that your autopilot / IMU filter incorporates your accelerometer measurements thru its filter gains anyway, so its a minimized effect.
Sustained slower motion (say, a long loiter), could be worse as it lets a constant build up in the error states occur. If you are doing a 20 meter radius loiter at a speed of 10 m/s for, say 1 or 2 minutes (depending on your IMU filter), you'll still only build up 30 degrees of attitude error durng that time (say one minute), which the pilot should be able to handle, especially if he recognizes the possibilty.
Analog Node for CHIMU - IR Sensor Replacment Module
This is impossible as the knowledge of absolute pitch angle in most IMUs is not stable during acceleration, so it cannot work with out good airspeed sensor. You end with a very complex system that has order of magnitude higher manufacturing costs (both tuning, calibrating and sensors).
There was a German system for this purpose that ended around 1000EUR+, don't remember now.
We haven't released such a product yet, but you're right that there is a clearly a place for it. I expect that this community will create one in the next few months. It is, as you say, an obvious choice and pretty easy to do. The only catch is that it would probably require a new PCB to integrate the RC (ArduPilot) and ArduIMU components into one small package, with an enclosure.
We should be able to duplicate all the functionality (and more) of the FMA CoPilot II for about half the price, with the additional advantages of being smaller and self contained (and open source!)
Replies
We had originally thrown this module together: Analog Node
To use with our CHIMU AHRS to do what you are talking about. Its a little pricey for this group (the unit uses magnetometers instead of GPS for yaw) and no open source, but...
Here is a picture of it next to an FMA module:
As for the speed question, we allow an analog input of voltage to accomodate corrections for this if you have a pitot tube speed measurement signal available. However, both experience (see Chris's reply) and the math say that you can get by without speed input if you aren't planning on threading any needles :)
The quick math version is that your attitude estimate will be corrupted by centripetal acceleration if not accounted for. Centripetal acceleration is dominated by your forward velocity, and it only couples into Y and Z accelerometer errors via yaw and pitch rates (respectively). Since (typically) body z and y velocities for a plane are centered about zero, your y accelerometer will see just your yaw rate* body velocity as an error.
So quick, high rate motion could produce high errors (for a constant speed), but those typically don't last long, and if they do then the pilot is probably pretty good and doesn't need a wing leveler. Further mitigating the error build up is what Brian said basically, which is that your autopilot / IMU filter incorporates your accelerometer measurements thru its filter gains anyway, so its a minimized effect.
Sustained slower motion (say, a long loiter), could be worse as it lets a constant build up in the error states occur. If you are doing a 20 meter radius loiter at a speed of 10 m/s for, say 1 or 2 minutes (depending on your IMU filter), you'll still only build up 30 degrees of attitude error durng that time (say one minute), which the pilot should be able to handle, especially if he recognizes the possibilty.
There was a German system for this purpose that ended around 1000EUR+, don't remember now.
We should be able to duplicate all the functionality (and more) of the FMA CoPilot II for about half the price, with the additional advantages of being smaller and self contained (and open source!)