Meso-scale coalitional control in large-scale networks☆

Network systems are usually composed of many interacting nodes. A large number of nodes and their possibly high-dimensional dynamics may result in complex system designs, which pose challenges to control of such systems in real-time. The traditional methods generate reduced-order models that well approximate the behavior of a high-dimensional system. However, when applied to large-scale network systems, these approaches face the problem of structure preservation. The control of complex networks is highly dependent on their interconnection structure, which is usually lost in the process of the conventional model reduction. Frequently, large-scale systems possess structural features that divide the system into subsystems or clusters. In this paper, we propose a meso-scale control approach that employs structural features of the system to reduce computational complexity. This approach is integrated into the model predictive control (MPC) synthesis framework. Our theoretical analysis highlights the significant reduction in computations at the expense of a little decrease in accuracy. The simulation results demonstrate the benefits and limits of our proposed control method. (c) 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.