This course will cover broadly the many aspects of constructing a model predictive controller for a real application: model formulation, constraints, integrating disturbance models, moving horizon state estimation, setpoint tracking with unreachable setpoints, and offset-free control with nonzero disturbances.

Model predictive control (MPC) has become the leading advanced control method in many industries. MPC is the method of choice for controlling the large, multi-variable, and economically important parts of most chemical processes and plants, but it is also now finding application in many electrical, mechanical and aerospace applications. MPC is especially effective in handling processes that run up against constraints. Examples of constraints are: valves saturating fully open or closed, threshold limits on product quality variables, and safety limits on temperatures and pressures. Operating at constraints is recognized as an everyday occurrence in chemical and biological processes.

This course will cover broadly the many aspects of constructing a model predictive controller for a real application: model formulation, constraints, integrating disturbance models, moving horizon state estimation, setpoint tracking with un-reachable setpoints, and offset-free control with nonzero disturbances. The course will present in addition the theoretical analysis tools that are used to establish the closed-loop control properties and the statistical properties of moving horizon estimation: Lyapunov stability theory, the theory of random variables, and conditional probability.
In addition to giving the students a set of design tools for applying MPC to challenging control applications, the course will also supply sufficient analytical background to enable the student to read and understand the theoretical MPC research literature.