Recent Advances in Model Predictive Control: From Theory to Practice
Model Predictive Control (MPC) is a promising approach for high performance multivariable control applications, since it systematically handles constraints on system inputs, states, and outputs, and it shapes the transient response through the optimization of a user-defined performance criterion. Even though MPC can be applied to any type of dynamics, efficient computational algorithms have been proposed for MPC of linear systems and of hybrid dynamical systems. The use of multiparametric programming allows to synthesize the MPC controller in the form of a piecewise linear state-feedback. Thus, MPC that was historically limited to controlling slow dynamics with abundant computing resources, such as in the chemical process industry, can now be applied to fast processes with limited computing power, for instance, the ones in automotive applications.
In this talk we discuss recent results in linear and hybrid MPC, including stabilization techniques, controller matching designs, stochastic MPC, and (wireless) networked MPC. We also present applications of linear, switched, and hybrid MPC in different domains. In particular, we describe the application of linear and switched model predictive control to idle speed regulation developed at Ford Motor Company, and the network hybrid MPC strategy applied for controlling a process where feedback measurements are obtained via a wireless T-Motes sensor network. For these applications, experimental results are reported.