Nonparametric recursive identification and control of a flexible joint robot manipulator

In the paper a linear quadratic Gaussian (LQG) dynamic regulator is applied with an extended Kalman filter (EKF), a LQG with fuzzy logic adaptive EKF (FLAEKF), LQG with an EKF and a FLAEKF combined with time delays in the feedback loop for modeling non-minimum phase (NMP) response for control of the end effector with non-collocated sensor and in the feed-forward loop for corrective control of a two-link flexible robot in the tracking of square trajectory task. The system is compared in simulations with a fuzzy logic system (FLS) vibration suppression control system. Results show that FLS adaptive vibration suppression yields higher tracking precision than FLAEKF, EKF or corrective time delays. It is also more effective while maintaining tracking accuracy and reasonable time efficiency than the classical PID controller or advanced adaptive controllers. Recursive nonparametric identification procedure for nonlinear friction in the joints of flexible robot is introduced. Its asymptotic properties are investigated.