3D Robotics Comment by Chris Anderson on February 16, 2009 at 10:02am

That is awesome on so many levels! How close do you think this is the actual data you'd be getting from the NXT sensors? Aside from the refresh rate,are you more or less testing a flyable autopilot?

Comment by Sami Finnila on February 16, 2009 at 10:33am

I'm quite confident with the IMU I now have in my NXT AutoPilot.. While my Lego IMU is missing one degree of freedom I'd say that in level flight the IMU data is never off more than 3-4 degrees and even that only momentarily (probably only hundreds of ms).

The tricky part is when the aircraft turns; I had a nasty bug in my gyro integrator that caused the pitch data to go nuts when the aircraft was turning to steep but, I believe, that I have now fixed that issue and it remains to be seen how accurate the pitch data will prove to be. If I had to quess, I'd say that I'll be able to get the error of the IMU data to be less than 5 degrees in the worst case scenario which is far less than that of the X-Plane simulator when you take the 10hz refresh rate into consideration.

All in all, I'd say that this baby is ready for the skies..

Comment by Sami Finnila on February 16, 2009 at 11:56am

Just a clarification: the simulator code for NXT is otherwise identical to the real deal aside from the IMU part so now it all really boils down to the quality of the IMU..

Comment by Sami Finnila on February 20, 2009 at 5:11am

Here's the new source code of the NXT AutoPilot: NXT AutoPilot v0.2b.zip

// -Changes in 0.2 BETA:
// -Fully rewritten IMU and stabilization code.
// -Altitude control (with PID) given that your GPS module sends GSA.
// -Moved all the settings into one place for easy configuration.
//
// -Support for compass and barometer removed.

Sami F.

Comment by will on March 24, 2010 at 11:48am

Hey could you make one for Microsoft flight simulator please and can it also be connected via usb instead of bluetooth

Comment by Sami Finnila on March 24, 2010 at 12:22pm

@will, I've abandoned the project since I sold my NXT to afford a tried and true autopilot (Paparazzi v3). There are so many advantages a purpose built ap has over systems like NXT. Besides the code isn't flyable in real world. The simulator flies perfectly but the IMU/AHRS can't handle turns; it doesn't properly compensate centripetal forces. As long as it flew in straight lines it worked fine (i.e. the stabilization worked) but once it started steep turns it didn't stop. Who knows it might've work with proper correction for centripetal acceleration (and maybe one more DOF). Also because the HW can't over sample the sensors often enough (accelerometer is the main problem) and the HW doesn't have filters what so ever, I experiensed some pretty severe problems with vibration and had to overcome those problems by means of installing the sensors in foam with lead plates to make them physically "slower".

The concept of making a UAV with NXT is exciting, granted, but the fact is that NXT isn't very well suited for such a task.

And because I don't have my NXT anymore I'm sorry but I'm unable to help you with the programming.

Comment