To tune, to switch, or to switch and tune: adapting to constantly changing environments

Systems theory as taught in academia is methodology driven. Systems practice, on the other hand, is invariably problem driven and extremely diverse. Control problems that arise in industry are characterized by poor models, high dimensional decision spaces, nonlinearities, and rapid time variations in system dynamics due to external disturbances, parameter variations, or sensor and/or actuator failures. Conventional controllers, designed using only one of the theories mentioned earlier, do not possess the attributes needed to control systems satisfactorily in the presence of large uncertainties and time variations. The objective therefore is to determine flexible architectures for control that are computationally efficient, yield fast and accurate response, and at the same time can be analyzed within the framework of mathematical systems theory. The article examines some of the issues involved, in general and in particular types of system.