Model predictive control (MPC) has proven highly effective for control of systems subject to constraints, but is usually considered in a time-invariant setting. On the other hand, periodic control has been considered for many decades, but usually in an unconstrained setting.
In this talk recent advances in linear-periodic MPC are presented. The presented methods are extensions of the well-known methods for MPC of linear time-invariant (LTI) systems, both in the sense that the MPC concepts are extended from the LTI to the periodic setting, but importantly also in that when the periodic system has a period length of unity (i.e., LTI system) they reduce to the well-known LTI-MPC methods.
There are three reasons why periodic MPC is useful. First, systems with periodic dynamics exist in practice, for example turbines, engines and walking robots. Second, systems with time-invariant dynamics may lead to periodic control problems, because the factors affecting the system are periodic. An example of this is building climate control. Third, periodic systems are convenient models of systems that may be time-invariant, but are subject to asynchronous timing constraints on the inputs.
The talk consists of three parts. Part I covers the modeling of systems with asynchronous inputs by periodic systems, and subsequently develops the MPC synthesis generalizations of well-known LTI-MPC methods. In Part II a brief outline of the application of periodic MPC to building climate control is presented. Part III covers the performance analysis of constrained periodic MPC laws.
Ravi Gondhalekar was born in Boston, USA, in 1979. He received B.A. and M.Eng. degrees in Engineering in 2002 from the University of Cambridge, UK, and a Ph.D. degree in Informatics in 2008 from the Tokyo Institute of Technology, Japan. He is currently an assistant professor at Osaka University, Japan. His research interests include model predictive control, constrained control, set invariance and hybrid systems. Ravi has been employed at the Massachusetts Institute of Technology, the University of Cambridge, Princeton University, Pi Technology, the Rutherford Appleton Laboratory and the United Kingdom Atomic Energy Authority.