MPU6050

[idhorst]

Have you guys been able to get the Digital Motion Processor (DMP) of the MPU6050 to work in order to enable sensor fusion? In other words do you have the licence that allows open source projects to use the built-in sensor fusion processor in the MPU 6050? What kind of output can we expect besides angles? Are we able to use quaternions?

Keep up the good work!

[tobias]

We have not had any time to investigate the possibility of using the DMP in the MPU6050. Would be fun to do, maybe you are up for it . However I'm not so sure the DMP algorithm will be the best choice for the Crazyflie. Only a test would reveal that but we don't have a license so it might be tricky. Jeff Rowberg is doing some great work though: https://github.com/jrowberg/i2cdevlib

We have a little block diagram of our stabilization implementation in the wiki: http://wiki.bitcraze.se/projects:crazyf ... mware:arch. It is not 100% accurate but good enough to get the picture.

[idhorst]

The advantage of getting the DMP to work is that it will use all sensor information and obtain a more precise output. It could even filter out the problems you guys have with the compass. All this without using the processor of the Crazyflie itself. AFAIK the license for the DMP is free for open source projects. At least that is the impression I get from reading the Arduinopilot fora. Coding your own Kalman filter to do the same is a lot of work and it would use some processor power and memory.

Forum
http://diydrones.com/profiles/blogs/inv ... 0-dmp-proc

Invensense website
http://www.invensense.com/developers/in ... _r=default

I hope you guys are able to get the license.

[tobias]

Yes, as you point out there might be many benefits for using the DMP, freeing processing power is one. Another one is that we kind of expect the sensor fusion algorithm to be top notch, however we would loose the incite to the algorithm with is a big drawback.

We will look into the possibility to obtain a license and thanks for the links!

[TheFrog4u]

Do you guys know the angular accuracy of your sensor fusion algorithm and the DMP using the MPU6050?
For another project I am searching for a motion sensor witch high angular accuracy (<0.5 deg). The result can be avaraged of a long time (several seconds), though - therefore I thing the gyroscope is less of interest than the accelerometer, which should have an accuracy (not resolution) of better than 0.008 g -> avaraged.

[tobias]

Not really but we know there is a big difference whether the Crazyflie is affected by vibrations (flying) or not. If you calibrate it I think it is possible to get <0.5 deg accuracy (with no vibrations). Currently we have not focused on calibration routines so there are plenty of work to be done here. My guess is that the current accuracy is about 2 deg.

Fabiano (FreeIMU) where almost finished with a python tool to calibrate the FreeIMU. It might be possible to build upon his work to quickly get some calibration working. http://www.varesano.net/blog/fabio/free ... ersion-out