Model predictive control. Classical, robust and stochastic (Q892968)

The book provides a comprehensive, formal, and up-to-date description of the theory and design of Model Predictive Control (MPC) algorithms for discrete-time linear systems subject to model uncertainties. State-space linear prediction models in closed-loop with full state feedback laws are mostly employed throughout the book, and corresponding designs are also worked out for input-output Auto-Regressive Moving Average in most chapters. This feature makes the book appealing to both academic researchers and practitioners interested in a thorough understanding of the model predictive control fundamentals. A distinguishing feature of the book is the exhaustive treatment of additive, multiplicative, deterministic and in particular, stochastic uncertainties. The overall material is organized based on technical content, as follows. After presenting an introduction to MPC in Chapter 1, the relation with unconstrained optimal control and the basic ingredients for parameterizing prediction dynamics, guaranteeing recursive feasibility and establishing stability are presented in Chapter 2. The authors make use of the standard terminal cost and set method, and they put emphasis on invariant terminal sets, accompanied by corresponding computational methods for polyhedral or ellipsoidal terminal sets. This approach is followed throughout the book, for most uncertainties types. Chapters 3 and 4 offer an exhaustive treatment of tube-based MPC for linear systems with additive uncertainties, including recent advances provided by the authors in the design of parameterized feedback strategies. Chapter 5 presents robust MPC schemes for multiplicative and mixed uncertainties, which make use of linear matrix inequalities formulations or, alternatively, of low complexity invariant polytopes to deal with the computational complexity due to the propagation of uncertainty. The remaining Chapters 6, 7 and 8 are dedicated to an introduction to stochastic MPC, establishing feasibility, stability and convergence in a stochastic setting, algorithms for Markov chains as prediction models, and an explicit use of probability distributions in stochastic MPC, respectively. Chapter 9 concludes the book with summary and further recommendations. Regarding readability, the mathematically rigorous treatment of MPC is rendered accessible and enjoyable by many helpful examples and colorful illustrations present in all chapters. The chapter-based, selective referencing policy makes it very easy to find relevant works, while the exercises with solutions at the end make the book very suitable as a study material for students. Overall, this book manages to provide complete and mathematically rigorous solutions to all the raised problems, under the considered assumptions. In conclusion, the reviewed book is highly recommended to all students (and in particular starting PhD students), researchers and practitioners seeking for a self-standing, clear and mathematically rigorous exposition of the theory and design of classical, robust and stochastic MPC with a linear prediction model structure.