I’ve noticed that some of the Intel RealSense cameras and the ZED-F9R module have their own IMU or are designed to take IMU input. But the flight controller has its own IMU.
From a sensor fusion perspective, which is better:
- Relying on the flight controller’s IMU alone.
- Using the IMU built into these sensors to improve their position, velocity, acceleration, etc. estimate, then feeding that data to the flight controller to be fused with the flight controller’s IMU data?
Does using multiple IMUs, even if they’re fused at lower sensor levels (like the IMU in a T265 camera, for example) improve or degrade overall position, velocity, acceleration, etc. estimates?
It seems to me (a layman with no knowledge of how the EKF works) that using multiple IMU data from different sources, baking it into the intermediate sensor output of a T265 tracking camera, for example, and then fusing it with the flight controller’s IMU data could introduce additional error.
Is one IMU better than three IMUs that all have some amount of covariance (if that’s even the right way to use that word)?