So we have a 5 axis IMU on one of our robots. Robot drives around on a flat field, so we only care about translation on the xy place and rotation about the yaw. Onboard the IMU consists of a 2 axis Accelerometer, and a 3 axis Gyroscope. So from my research I have come to understand that we can not accurately have a calculation of the yaw angle due to the fact that the Accelerometer sees nothing during yaw movement. However, if the Accelerometer is offset from the rotation axis and the gyroscope would it not be possible to calculate a more accurate yaw angle, since it actually sees a change.
So my questions are:
-Is this idea even plausible?
Is it possible to do with just the 2 axis Accelerometer and 3 axis Gyroscope, or would you really need another separate Acceloremeter for the offset?
-If so does the math become that much more complicated that it is almost impossible to do, and is that why everyone just goes for a 9-axis imu?
You will only get an approximate yaw with either an accelerometer or gyro. Any error in the calculation of how much you've turned will continually compound and eventually you'll have no idea what direction you're facing.
You need either a magnetometer (electronic compass) to tell you what direction you're facing or GPS to tell you what direction you're moving.
I read that some use an accelerometer on the arm of a robot to calculate the position. So you also could use the accelerometer to measure the yaw.
But using the gyro (in the middle position of the rotation) would be better. The added accelerometer with offset would probably not increase the accuracy.
I suppose the accelerometer can give a reason value for the magnitude of yaw rotation rate (but not the sign), if
that can be used to refine the value from the gyro you might gain some signal/noise ratio, but you'd have to carefully
correct for every other type of motion in order to get a clean signal - and you have to make the assumption that
rotation is always about the same axis (is this true?).
Adding a magnetometer for drift-elimination would perhaps be less hassle. The performance of a 3-axis gyro in the
short term should be fairly good if tuned carefully.