A linear model of a one-link flexible robot arm is only anapproximation to the real system, whose parameters are subject to changedue to working conditions. The paper presents a simple experiment todemonstrate the necessity of robust control to industrial robots. Twocontrol techniques, linear quadratic Gaussian (LQG) and loop shaping,are used to design controllers for our laboratory flexible robot arm.The LQG controller yields an optimal closed-loop performance for thenominal system, but when plant parameters are perturbed, the performancesignificantly degrades. On the other hand, under the same perturbations,a loop shaping controller can robustly maintain the system stability andresult in an acceptable performance. The agreement between simulationresults and experimental results provides an initial step to ourlaboratory setup which is presently planned for the introduction of arobust control class
展开▼