Consensus Control For Distributed Systems Under Communication Constraints

By aid of network communication, information and automation technologies, many successful distributed systems have been gradually constructed in the past years. For guaranteeing the system stability and further providing prominent system features such as controllability, flexibility, expandability, etc., design of feasible control strate-gies becomes an urgent requirement. Generally, the final objective of distributed control systems is to perform effectively control via data exchanging through communication networks so as to solve problems which are beyond the individual capability. However, due to limit capacity of communication facility and its induced network constraints such as time-varying delays, data packet dropouts, communication topology changes, exoge-nous disturbances, etc., the control performance may be deteriorated even the stability of overall system may be endangered. Therefore, how to design some feasible control strategies for tolerating these communication imperfections while guaranteeing system stability is a challenging problem. Up to now, the intrinsical complementary mechanism between the system, network and control are not be thoroughly revealed, therefore, it is necessary to further probe some effective distributed system modeling and control s-trategies. For investigating some feasible control strategies as well as completing control theory research, the methods of active queue management and event-trigger are thor-oughly investigated at first; afterwards, considering communication imperfections, some distributed control strategies are proposed by which the consensus of distributed system can be guaranteed meanwhile the communication imperfections can be tolerated. The main contribution of this dissertation is given as below.(1) The network congestion phenomena usually appears in the network routing n-ode when the required data transmission amount is exceed its data transmission capacity, however, the existing congestion control strategics mostly require large computational burden and show weakness in system robustness. For alleviating data routing congestion in Internet wherein system state cannot be directly measured, a robust observer based ac-tive queue management will be investigated. By Ihe proposed method, the queue length of routing node can be stabilized and maintained around the pre-given equilibrium point, thus the date routing congestion can be alleviated.(2) In real networked systems, many kinds of sensors are dispersed in differen-t locations and massive data is originated by these sensors. Since the amounts of data greatly increase the operating pressures in data transmission, processing and control, the event-triggering scheme is firstly investigated for reducing data transmission amoun-t. Afterwards, considering some communication induced imperfections including data transmission time-varying delay, switching topology and data packet dropout, a com-munication tolerant cooperative control scheme is proposed for guaranteeing the desired consensus of multi agent systems (MASs). Furthermore, for achieving synchronization of heterogeneous MASs (He-MASs) while reducing the data exchanging burden, a novel event-triggered distributed cooperative control strategy is investigated. By the proposed method, the consensus region of He-MASs can be regulated meanwhile the Zeno phe-nomena is excluded.(3) The exogenous disturbances ubiquitous usually exist in networked control sys-tems which are capable of deviating system information and deteriorating system perfor-mances. For the perturbed distributed networked systems in presence of multiple exoge-nous disturbances, a point-to-area H? performance is firstly proposed for suppressing the negative effect of multi-source disturbances. Afterwards, considering data transmis-sion time-varying delay and switching topology change, a distributed cooperative control law is proposed for achieving consensus of distributed systems while tolerating commu-nication imperfections.(4) As far as large scale systems (LSSs) are concerned, the traditional multi agent systems (MASs) based cooperative control laws will easily loss the applicability since the nodes of LSSs are already organically coupled and the LSSs are usually unneces-sary or even impossible for all of its nodes to be controlled. For determining the key node(s) of LSSs, a novel node selection principle is proposed. For further reducing the controller burden as well as the communication pressure, a pinning control law and an event-triggered control command updating scheme will be simultaneously investigated. By the proposed method, the group consensus of LSSs can be guaranteed only few nodes are controlled and seldom control commands are updated, meanwhile the data transmis-sion pressure as well as the controller burden can be simultaneously reduced.