Identification and gain scheduled vibration control of an experimental two-link flexible manipulator

In this paper experimental results are presented for the identification and control of a very flexible two-link robot manipulator. A parametric frequency domain identification algorithm is applied to input-output data to identify linear SISO models at several elbow angle configurations which are subsequently used for LQG compensator synthesis. It is demonstrated that the variation in structural dynamics with elbow angle is sufficiently large to destabilise a fixed gain vibration controller during large angle manoeuvres. A gain scheduled compensator is developed from the identified models and successfully implemented to maintain stable control despite large configuration induced parameter variations