Before I’d even finished my previous post, I’d found a solution that meant I didn’t need a Kalman filter just yet
It all started with this, describing how to merge accelerometer Euler angles with gyro angular rate transformed from body frame to the 3 Euler frames. But as per usual, it wasn’t enough; while I understand the complementary filter well, I didn’t understand the conversions of gyro rotation rates in the body frame to the Euler frames. Until I found this.
That now allows a complementary filter to track Euler angles short term via gyro and long term with the accelerometer. I know I’ve done this before, but this time, I’m merging measurements in the same same set of Euler frames so the results should be correct.
A bit of manic coding last night, combined with a repeat of yesterday’s tests showed this.
Compare this with the graph from yesterday’s equivalent tests:
I hope you agree this is much better – as expected there are roughly sinusoidal lines about 90° out of phase as Phoebe rolled around the floor in circles.
No option now but to test this new code live – need to find where I left my bravery first though!