Gyros- Gimbal Lock

[RCvertt]

Does MultiWii still have gyro gimbal lock? I think Timecop said that his code doesn't have it so I was wondering if a lighter version that the 328p can handle has filtered into the Arduino based code yet or if it's even possible?

I would like to use the Accel to automatically detect if my tail sitter plane has pitched 90 degrees, so that it knows when it is flying forward or taking off vertically but I can't do that with gimbal lock present.

[timecop]

Of course it still has gimbal lock, as long as your calculation isn't in quaternions.
I didnt touch control/attitude code from mwc, so same problems exist.
Also for some insane reason mwc code does pitch in -180+180 range where everything else sane does -90+90.
I'm sure patches to fix this would be very welcome.

[RCvertt]

Thanks for the reply. Wasn't sure if quaternions had been implimented yet or not.

I think it will allow for some cool now features and projects when it goes away. Hopefully someone can replace the Euler calculations for us.

[JBear]

There is an open source IMU algorithm using quartenions available that has been used for Multirotors (in the stalled project Pipo by Erez Raviv). According to the report the sensor fusion algorithm is slightly better than a Kalman filter implementation and the use of quartenions prevent gimbal lock. The C-code implementation of the filter is provided and as it was used in the Pipo, even an Atmega328 managed to calculate it all without bogging down.

[RCvertt]

There is an open source IMU algorithm using quartenions available that has been used for Multirotors...

Very nice. So did anyone every get around to implementing this for multiwii 2.1 or 2.2? This stuff is way over my head to implement or I would gladly have done it.

[timecop]

You don't want this calculation on 8bit/16mhz platform.

[brm]

nada - tc, you are WRONG!
seb's code on a stupid 8 bitter makes sense.

[Alexinparis]

If you use v2.2 + a 9DOF sensor, there is no more gimbal lock problem.
The angle calculation changed.
You can see it in the java GUI model representation.
I think the algo is way faster than any quaternion approach.