Advances on Distributed Model Predictive Control

Model Predictive Control is a class of advanced control techniques, widely used especially in the process industries, and it has its fundamentals in optimal control. Further, several class of predictive controllers were developed in the last two decades. Nevertheless, conventional feedback controllers (e.g., PID) are, so far, the de-facto standard for most industrial applications. This is due to the fact that no system model is necessary for these approaches. On the other hand, in case of large-scale applications, conventional feedback controllers are no longer appropriate since industrial control systems are often decentralized (i.e., interactions among subsystems are not considered). Due to dynamic coupling it is well known that performance may be poor, and stability properties may be even lost. MPC provides several form of distributed approaches that guarantee nominal closed-loop stability and convergence to the centralized optimal performance. This thesis shows recent research activities on Distributed MPC to demonstrate how is getting a mature technology, suitable to be applied to different application areas and large-scale systems, with computational as well as organizational advantages. This could lead to use MPC beyond its control aspects, in order to exploit its management capabilities. This work shows how Distributed Model Predictive Control, not only seems to fit the new technologies that are entering the global market, by creating a new interesting opportunities, but also could become a keyword for the emerging "smart factory".