Nonlinear tip-position tracking control of a flexible-link manipulator: Theory and experiments (Q5950257)

This paper addresses the end effector trajectory tracking problem for flexible link manipulators. Its main result consists in the design of an inverse dynamics controller assisted by a link deflection speed observer. The equations of the manipulator's dynamics take on the form NEWLINE\[NEWLINEM(q,\delta) {\ddot q\brack \delta}+ {f_1(q,\dot q)+ g_1(q,\dot q,\delta, \dot \delta)\brack f_2(q,\dot q)+ g_2(q,\dot q,\delta, \dot\delta) +K\delta}= {u \brack 0}, \tag{*}NEWLINE\]NEWLINE where \(q\in\mathbb{R}^n\), \(\delta\in \mathbb{R}^m\) denote, respectively, joint variables and deflection variables, \(f_1,f_2, g_1,g_2\) stand for Coriolis, centripetal and gravity torques, \(K\) is the stiffness matrix, \(u \in \mathbb{R}^n\) denotes a vector of input torques. The outputs of this system have been chosen as NEWLINE\[NEWLINE(\%)\;y=q+\Psi (\alpha)\delta,NEWLINE\]NEWLINE where the matrix \(\Psi (\alpha)\) depends on modal shape functions, and \(\alpha\in\mathbb{R}^n\) serves to achieve stability of the zero dynamics of the system (*) with outputs \((\%)\). It is shown that this system has an input-output map NEWLINE\[NEWLINE\ddot y=a(\alpha,q,\dot q, \delta, \dot\delta)+ B(\alpha,q, \delta)u,NEWLINE\]NEWLINE and is locally input-output linearizable by a static feedback, which results in the tracking algorithm of the form NEWLINE\[NEWLINEu=B^{-1} (\alpha,q, \delta)\bigl(\nu -a(\alpha,q,\dot q,\delta,\dot \delta) \bigr)+K_1 (q)\delta+ K_2(q)\dot \delta.NEWLINE\]NEWLINE The last two terms in the above expression increase robustness of the algorithm, \(\nu\) incorporates the 2nd derivative of a desirable output trajectory as well as a PD controller depending on the position and estimated velocity tracking errors. Since the speed of deflection variables is usually unknown, a full order and a reduced order Luenberger-type observer have been proposed in order to estimate \(\dot \delta\). The main result proved in this paper states that the tracking observation errors in the closed loop system composed of the tracking controller and the observer can be made arbitrarily small. Theoretical developments have been illustrated with results of experiments.