Decentralized wireless networked model predictive control design for large and complex systems

Abstract

Decentralized Networked Model Predictive Control system (DNMPC) is a decentralized control system that involves the exchange of information between subsystems across a mutual communication network. Decentralized structures are attractive and widely used solutions for controlling large and complex systems, such as energy hosts control, robotics, water networks, wireless sensor networks, traffic control, among others. However, the inclusion of the network introduces dropouts, which greatly influence the stability of the system. In this paper, an innovative DNMPC solution is presented to compensate for dropouts when using a wireless network. Moreover, the decentralized control performance is improved through the implementation of a cooperative strategy where controllers exchange signal predictions. The effect of interactions and different rates of information exchange between the subsystems have been investigated. Experiments using a real-time network simulator demonstrated the effectiveness of the proposed approach to deal with missing sensor information, systems uncertainties and strong interactions between the subsystems while maintaining a good performance. The approach offers an effective and innovative solution to improve the reliability of decentralized networked control systems.