Consensus For Networked Multi-agent Systems Based On The Predictive Control Scheme

The problem of consensus has received extensive attention in the distributed coor-dination and cooperative control of networked multi-agent systems (NMASs), due to its broad applications in cooperative control, formation control, design of sensor networks, flocking of social insects and so on. Roughly speaking, the consensus means to design protocols such that the group of agents can agree on certain quantities of interest based on the local information. To achieve consensus and cooperative behaviors for NMASs, it is important that the agents in the group are capable of exchanging information through communication networks, which implies that the network-induced delay can occur in-evitably, due to the limited bandwidth of communication channels and the finite transmis-sion speed. Time delays can degrade the performance of networked multi-agent systems and even cause divergence. Hence, this dissertation systematically and deeply investigates the problem of consensus for NMASs with communication delays.The problem of consensus for networked linear discrete-time homogeneous multi-agent systems (NHMASs) with communication delays is investigated. Firstly, for NHMASs with a common and constant communication delay, based on the networked predictive control scheme (NPCS), a distributed consensus protocol is proposed to compensate for the communication delay actively. In the directed and fixed topology case, necessary and/or sufficient conditions of the consensus are given, which indicate that consensus is related to the agent dynamic structure and communication topology. Secondly, the above results are extended to the NHMASs with different and constant communication delays. Using the NPCS, the problem of consensus is transformed into the problem of asymptot-ical stability for linear discrete-time systems with multiple constant delays. Thirdly, the results of consensus for NHMASs with constant communication delays are extended to the case of diverse and time-varying communication delays. For NHMASs with bounded and time-varying communication delays, the problem of consensus is transformed into the problem of asymptotical stability for linear discrete-time switched systems with a time-varying delay.Because the structure of each agent cannot be same in NMAS, the problem of con-sensus for networked linear discrete-time hetergeneous multi-agent systems with a con-stant communication delay is investigated. Due to economic costs or constraints on mea- surement, it is often difficult or even unavailable to obtain the information of all the a-gents’states in practice. When the states of all agents are not available but their outputs are measurable, based on the NPCS, a distributed protocol is proposed to compensate for the communication delay actively. The delay-independent necessary and/or sufficient criteria are obtained. As a special case, the protocol design and consensus analysis are presented in the case of no network delays. Besides, under some circumstances, the exact output of each agent is also not completely available, but the output differences between an agent and its neighboring ones (relative outputs for short) can be measured, such as vehicle coordination and the networked lock synchronization. Secondly, when the ex-act output of each agent is not completely immeasurable, and yet relative outputs are available, based on the NPCS, a distributed protocol is proposed and delay-independent necessary and/or sufficient conditions are obtained. Finally, in order to increase freedom and flexibility of the design, based on the NPCS, a distributed protocol in the form of dy-namic compensator is put forward and delay-independent sufficient conditions are derived under mild assumptions.In cooperative control, the agents in a group must achieve agreement even if changes take place. A phenomenon that frequently occurs is that the agreements are different with the changes of environments, situations, cooperative tasks or even time. Consequently, a critical problem is to design appropriate protocols or algorithms such that the agents in a network reach more than one consistent states, i.e., the group consensus problem. Firstly, the problems of the group consensus for first-order NMASs with a fixed topology and with a switching topology are investigated, respectively. Under mild assumptions, necessary and/or sufficient conditions are derived. For the fixed topology case, the group consensus can be guaranteed using the Hurwitz stability of coefficient matrices. For the switching topology case, the group consensus is proved to be equivalent to the asymptot-ical stability of a class of switched linear systems under arbitrary switching signal. Sec-ondly, the results of group consensus for first-order NMASs are extended to high-order NMASs. For NMASs with the fixed topology and agents described by linear continuous-time time-invariant systems, necessary and sufficient conditions are derived under mild assumptions. Thirdly, the problems of group consensusability for continuous-time and discrete-time high-order NMASs with the fixed topology are investigated, respectively. Necessary conditions of group consensusability with respect to the given admissible con-trol set are derived under a mild condition. Different from the existing methods, the NPCS of a single control-loop control sys-tem through a real-time network is extended to the NMAS consisting of multiple sub-systems, where information exchanged among all subsystems is achieved via a network. Based on the NPCS, the communication delays are compensated actively, the predictions of data at current time are obtained, novel distributed protocols are proposed, and nec-essary and sufficient conditions of the consensus are given. Research results of this dis-sertation will contribute to promoting the application of the networked predictive control method in NMASs, and further understanding the consensus of NMASs.