The paper "Distributed Model Predictive Control" by Gabriele Pannocchia from the University of Pisa, Italy, discusses the principles and applications of distributed model predictive control (DMPC) in large-scale systems. DMPC involves dividing a complex system into subsystems, each controlled by a local controller, with interactions between subsystems. The paper covers various levels of communication and cooperation among these controllers, emphasizing the benefits of cooperative schemes where all controllers optimize a common objective function. Key topics include: - **Introduction and Motivations**: Explains the need for decentralized control in large-scale systems due to computational limitations and organizational reasons. - **Definitions and Architectures for Constrained Linear Systems**: Describes the dynamics, constraints, and objectives of subsystems, and introduces different control architectures such as decentralized, noncooperative, and cooperative MPC. - **Cooperative Distributed MPC**: Focuses on cooperative schemes, which are preferred for their theoretical guarantees and computational efficiency. The paper presents a basic cooperative MPC algorithm and discusses its properties, including feasibility, cost decrease, and convergence to the centralized optimum. - **Complementary Aspects**: Discusses issues like coupled input and state constraints, and proposes solutions to ensure convergence and stability. - **Parsimonious Local System Representations**: Suggests simplifying the state dynamics to reduce computational complexity while maintaining global optimality. - **Output Feedback and Offset-Free Tracking**: Introduces methods for using local state estimators and integrating disturbance models to achieve offset-free control. - **Distributed Control for Nonlinear Systems**: Reviews recent advancements in DMPC for nonlinear systems, including cooperative architectures and challenges in non-convex optimization. The paper concludes by highlighting future research directions, such as nonlinear DMPC, economic and tracking DMPC, reconfigurability, constrained distributed estimation, and specialized optimization algorithms for DMPC problems.The paper "Distributed Model Predictive Control" by Gabriele Pannocchia from the University of Pisa, Italy, discusses the principles and applications of distributed model predictive control (DMPC) in large-scale systems. DMPC involves dividing a complex system into subsystems, each controlled by a local controller, with interactions between subsystems. The paper covers various levels of communication and cooperation among these controllers, emphasizing the benefits of cooperative schemes where all controllers optimize a common objective function. Key topics include: - **Introduction and Motivations**: Explains the need for decentralized control in large-scale systems due to computational limitations and organizational reasons. - **Definitions and Architectures for Constrained Linear Systems**: Describes the dynamics, constraints, and objectives of subsystems, and introduces different control architectures such as decentralized, noncooperative, and cooperative MPC. - **Cooperative Distributed MPC**: Focuses on cooperative schemes, which are preferred for their theoretical guarantees and computational efficiency. The paper presents a basic cooperative MPC algorithm and discusses its properties, including feasibility, cost decrease, and convergence to the centralized optimum. - **Complementary Aspects**: Discusses issues like coupled input and state constraints, and proposes solutions to ensure convergence and stability. - **Parsimonious Local System Representations**: Suggests simplifying the state dynamics to reduce computational complexity while maintaining global optimality. - **Output Feedback and Offset-Free Tracking**: Introduces methods for using local state estimators and integrating disturbance models to achieve offset-free control. - **Distributed Control for Nonlinear Systems**: Reviews recent advancements in DMPC for nonlinear systems, including cooperative architectures and challenges in non-convex optimization. The paper concludes by highlighting future research directions, such as nonlinear DMPC, economic and tracking DMPC, reconfigurability, constrained distributed estimation, and specialized optimization algorithms for DMPC problems.