Nonlinear model based/model predictive control with constraints and with/without nonlinear observer

A literature survey from the area of nonlinear model predictive control (MPC) is presented. After a brief review of the main characteristics of linear MPC algorithms various classes of nonlinear MPC algorithms based on state-space models, such as nonlinear extensions of dynamic matrix control (DMC), Newton-type algorithms and nonlinear programming (NLP) based algorithms are discussed. These algorithms are often combined with algorithms for on-line state and parameter estimation. Recursive and moving horizon-based approaches are briefly reviewed. Various input/output models for nonlinear MPC, such as neural network models, fuzzy models, Volterra models, Hammerstein models and polynomial NARX models are also discussed. The same optimization strategies as is used for state-space models can usually be used for nonlinear input/output models. Finally, some features with various constraint handling strategies are discussed.