Multiple model-based control: development and initial evaluation

For a large class of physical systems the structure of the plant dynamics is known, but the parameters in the system model vary with time and/or configuration. A new form of adaptive model-based control has been developed for this class of control problem. The multiple model-based control (MMBC) technique utilizes knowledge of nominal plant dynamics and principles of Bayesian estimation to provide controller performance independent of plant parameter variations. The MMBC is formed by augmenting a model-based controller with a new closed-loop form of multiple model adaptive estimation (ΔMMAE). The ΔMMAE uses perturbation models of the combined plant and feedback control system, along with measurements of tracking error, to provide an estimate of the plant parameters. The model-based controller combines the a priori knowledge of plant structure with the parameter estimates to produce the multiple models of the plant dynamics required to maintain controller accuracy. Extensive simulation studies have demonstrated the MMBC´s ability to adapt quickly and accurately adapt to the parameter variations of a plant described by complex nonlinear coupled dynamics