Design of Robust Distributed Control Schemes with Applications to Multi-Agent Systems

点击翻译按钮获取中文摘要

The problem of controlling an interconnected large-scale system by means of structurally constrained controllers is considered in this proposal. In many physical systems such as sensor networks, power systems, coordinated motion of automated vehicles and traffic networks, the interacting subsystems are geographically dispersed. Usually, traditional centralized control for this type of system is not as efficient, due to the distributed nature of the network and also the complexity of the required on-line computations. Furthermore, continuous transmission of data from the subsystems to the controller and vice versa is typically expensive. Another obvious disadvantage of such structure is that the resultant closed-loop system has a single point of failure. In a distributed control structure, on the other hand, one can generate the control input of each subsystem by only concentrating on those outputs which have a relatively high impact on the corresponding subsystem (through interconnections). This reduces the complexity of the computations, and at the same time improves the efficiency of the control topology. It is known that inexact modeling, actuator limitations, sensor constraints, measurement error and time delay, if not taken into account in the controller design, can lead to performance degradation or even instability of the interconnected system. Due to the complexity of the control analysis and synthesis with the above-mentioned practical considerations, they are often neglected in the problem formulation. In this work, effective approaches will be used to tackle the problem in the most practical setting, and sufficient conditions will be derived to guarantee robust performance. In particular, the effects of these practical issues on connectivity preservation, consensus, containment, formation control and tracking problems in a multi-agent system will be investigated. Such considerations in control design and analysis will make the results of this research more applicable to real-world multi-agent networked control systems.