I have read here and elsewhere of different techniques of estimating tilt angle, rate of change of tilt angle.
Many of these use a combination of an accelerometer and rate gyroscope. There is also extensive use of Kalman filtering with both of these sensor types.
I may be a little naive, but I think this could be over complicated.
Surely a single accelerometer can be used to derive the instantaneous angle of a body and also the rate of change of angle?
And, surely by limiting the bandwidth to the minimum that the system (i.e. balancing robot) will exhibit and, low pass filtering the output from the acceleromoter will remove the need for complex Kalman filtering?
I must be missing something horribly obvious here but, this is something I want to investigate during the development of this robot.
Parts arrive tomorrow.
Schematic capture in progress.