On-line generalized predictive control combined with recursive least squares system identification.

摘要

Motivated by the development and application of techniques for controlling optical jitter, an adaptive control algorithm is developed and demonstrated to structural vibration control, acoustic control, tracking control, and flutter control. The proposed adaptive control algorithm combines the RLS system identification algorithm and the GPC design algorithm in a single process, referred to as the Recursive Generalized Predictive Control (RGPC) algorithm.; The RLS algorithm is preferable due to the fast convergence rate. The GPC design algorithm estimates the future outputs and designs a controller to make the predicted output be as close as possible to the desired future output. In the design process, the prediction horizon and control horizon are the constants to be chosen. Two new parameters are defined to describe the effects of the prediction and control horizons and those parameters provide the effective ranges of the horizons.; The RGPC algorithm is a discrete-time, indirect, self-tuning algorithm. The adaptive control algorithm is designed for discrete-time application and can be implemented in real-time. The indirect algorithm estimates model parameters from measured and known system data and designs a controller from the estimated model parameters, while the direct algorithm designs a controller directly from the system data. The algorithm adjusts the control penalty in real-time based on the stability of a closed-loop system model. The RGPC algorithm is developed for MIMO systems, non-zero reference positions, and feedforward of disturbance measurements.; RGPC is applied to nine different systems to demonstrate its feasibility. First, it is applied to structural vibration control: a cantilevered beam and a simply supported plate. Second, it is applied to an acoustic enclosure with a time-varying configuration. Two different jitter testbeds are used for jitter suppression. In addition, RGPC is applied to a rocket fairing and acoustic cylinder to demonstrate acoustic disturbance rejection. As a further application, RGPC is extended to tracking control—the servo control of a piezoelectric device in an atomic force microscope,—and to flutter control. The algorithm is applied to a wing model to stabilize the unstable open-loop system. In all experiments, the RGPC algorithm improves the system performance.; The primary contributions are the development of the adaptive RGPC algorithm and the real-time implementation of the algorithm. The experimental results show the potential of the RGPC algorithm. The proposed adaptive control algorithm requires no prior system/disturbance model since models are estimated from real-time data, and controllers are updated adaptively in the presence of a changing operating environment.