With years of work on control systems, lots of control tools have been created for automation in different situations. Adaptive Control was developed for the models with unidentified but slow-varying parameters. Aiming to guarantee the stability of control systems in presence of modeling uncertainties or disturbance, Robust Control originated from methods to optimizing both performance and robustness of systems. Nevertheless, transit performance of systems remains a big problem, no matter in Adaptive Control or Robust Control. Switching/Supervisory Control was proposed to improve transit performance by adopting the most appropriate among multiple candidate controllers.

After A. S. Morse established a relatively general switching scheme by introducing a concept called “weighte...[ Read more ]

With years of work on control systems, lots of control tools have been created for automation in different situations. Adaptive Control was developed for the models with unidentified but slow-varying parameters. Aiming to guarantee the stability of control systems in presence of modeling uncertainties or disturbance, Robust Control originated from methods to optimizing both performance and robustness of systems. Nevertheless, transit performance of systems remains a big problem, no matter in Adaptive Control or Robust Control. Switching/Supervisory Control was proposed to improve transit performance by adopting the most appropriate among multiple candidate controllers.

After A. S. Morse established a relatively general switching scheme by introducing a concept called “weighted signals”, people tried many ways to explore more effective switching logics. The first contribution in this dissertation follows the mainstream of switching control, which covers existence and construction of multi-controllers, design of supervisors and proof of input-to-state stability based on the general theory of average dwell-time. Compared with previous work, two points deserve emphases: 1) A more general contructive method for multi-controllers, which generalizes the parametric uncertainty from scalar variables to vectors; 2) A structure of supervisor directly based on tracking errors, instead of “weighted signals”, and its input-to-state stability is guaranteed and proven. Nevertheless, logic-based switching control with finite controller coverings is only capable to partially improve performance. Such effects of improvement are not fully under our control due to potentially complicated theoretical issues. Instead of emphasizing stability in control design, the second contribution of this dissertation is the “logic-event based switching control”, which heavily relies on controllability to drag states to the positions we expected. And with a new supervisory architecture developed, switching control with on-line controller updates becomes possible to be implemented and satisfactory transit performance follows. Moreover, in order to show the advantages of the core principles, simulation and experimental results were presented in this dissertation.

In a sum, the work in this dissertation presented advancement in both current area and extensions of switching control. Considering the “logic-event based switching control”, it would prove to be a promising direction in the future studies.