ComplementaryFilter computes pitch but not roll about imu_tools

[mintar]

Great that you figured it out!

The complementary filter uses an NWU (north-west-up) reference frame. The device orientation is reported wrt. to that reference frame. This means that the neutral orientation is when the accelerometer is reading xyz=(0, 0, +) and the magnetometer is reading xyz=(+, 0, *). In the neutral orientation, the device frame and the reference frame align, so the orientation is rpy=(0, 0, 0). If you don't use the magnetometer, the yaw angle depends on the startup orientation of the device instead of magnetic north.

If you have any more questions, just ask. Otherwise you can close this issue.

[oliverdain]

I think I may have figured it out. The issue appears to be that my roll computation is actually a yaw computation.

[mintar]

By the way, NWU isn't really a standard. It should be either ENU (east-north-up) or NED (north-east-down). The imu_filter_madgwick supports those, but I haven't gotten around to porting these changes to imu_complementary_filter. Without a magnetometer, NWU and ENU are equivalent anyway.

[oliverdain]

All good. Thanks!