Controllability and observability of linear multi-agent systems over matrix-weighted signed networks

Abstract

In this paper, the controllability and observability of linear multi-agent systems over matrixweighted signed networks are analyzed. Firstly, the definition of equitable partition of matrixweighted signed multi-agent system is given, and the upper bound of controllable subspace and a necessary condition of controllability are obtained by combining the restriction conditions of the coefficient matrix and matrix weight for the case of fixed and switching topologies, respectively. The influence of different selection methods of coefficient matrices on the results is discussed. Secondly, for the case of heterogeneous systems, the upper bound of controllable subspace and the necessary condition of controllability are obtained when the dynamics of individuals in the same cell are the same. Thirdly, sufficient conditions for controllable and uncontrollable union graphs are obtained by taking advantage of the concept of switched systems and equitable partitions, respectively. Finally, a necessary condition of observability is obtained in terms of the dual system and the constraints of the coefficient matrix, and the relationship between the observability and the controllability of the matrix-weighted signed multi-agent systems is discussed.

Controllability of discrete‐time signed multiagent networks

“…Controlling the swarm behaviour of MASs is one of the fundamental problems of complex networks. Considering the limitations of computing or communication resources required for agents to exchange their state information in the network, the real-time exchange of information between agents is very expensive and the feedback control loop is usually implemented on microprocessors, which leads to computational and communication problems when MASs run consensus protocols [34]. Pinning control is widely used to control the synchronous behaviour of complex networks, whose basic idea is to realize the purpose of adjusting all agents' states to the pre-set states by using the interrelated characteristics of agents' dynamic behaviour in the networks [35,36].…”