Intelligent control of flexible/nonlinear systems

The dissertation focused on two topics: control of flexible-link robotic systems with known nonlinear rigid and known distributed flexible dynamics and intelligent control of flexible-link robotic systems and a class of nonlinear systems both having partially unknown dynamics.;For the case of flexible-link robotic systems with known dynamics, an improved reduced-order (i.e. modified assumed-mode model) is derived that does not simply truncate high-order modes, but incorporates quasi-steady state information about the neglected modes. Experimental results are given for the above model. Further, a nonlinear tracking controller for link-tip positions and velocities of a multi-link flexible robot arm is designed such that it gives guaranteed performance. First, a feedback linearization is performed with respect to the link tip position. Then, relaxing the tracking requirement allows the application of singular perturbation theory so that a boundary layer correction term (fast control) can be derived and used to stabilize the internal vibratory dynamics.;The intelligent control applications and theory addressed in this work are neural and fuzzy systems. Neural-network and fuzzy-logic systems are very powerful control tools because they allow model-free closed-loop control of complex nonlinear systems. A problem is that for said systems it is difficult to obtain stability theorems without a plant model. That is, generally the use of ad hoc controller structures results in the inability to guarantee adequate performance in terms of small tracking error and bounded NN weights. In this work, using both a neural network and singular perturbation theory, a tracking controller is designed for a partially known flexible-link robot arm. Here, a closed-loop control stabilizes the internal dynamics of the flexible arm because we choose a physically meaningful modified output, the slow portion of the link-tip motions. The final contribution is an adaptive fuzzy logic controller that uses basis vectors based on the fuzzy system to form a feedback controller that achieves tracking control of an entire class nonlinear systems with no plant model. Stability is proven and a systematic repeatable design algorithm is given for this fuzzy logic controller.