Zoe is pointing towards me. Vertical drift is good – you can tell because she’s at ground level when the flight completes. But leftwards drift (her point of view) is terrible. Why? Here’s my best guess and explanation why it’s unfixable as it stands.
It’s all to do with offset errors and integration to velocity. Imagine the perfect world with Zoe hovering above the ground; her X and Y accelerometers read zero and her Z accelerometer reads 1g.
But the world isn’t perfect and all three readings will be ever so slightly out by a fixed offset. For the Z-axis at start of data we measure gravity including the offset error; during flight takeoff acceleration, we measure the acceleration including the offset error. Subtract the two (remove gravity from total acceleration) and integrate and the offset errors cancel each other out – no vertical drift.
But it’s not the same with X and Y; there is no gravity to be subtracted, so their’s no cancellation of the X and Y offset errors as they are integrated for the horizontal velocity, and that means drift.
I’ve looked at X and Y axes offset measurement per-flight, and I might play some more with the concept, but fundamentally, the horizontal drift is unfixable and must be compensated for with other sensors. More on these later when I have progress to report on my laser altitude and drift tracker.