Consensus Based Multi-Agent Control Structures

In this paper a new approach to decentralized overlapping control of complex systems has been proposed in the form of multi-agent network were the agents are using dynamic consensus strategy in order to reach the agreement upon the control actions. Several structures and algorithms have been proposed which differ, on one hand, on the choice of the variables upon which the agreement among the agents is made (control or estimation variables), and, on the other hand, on the local controller structures derived from the decentralized control laws. Properties and performance of the proposed algorithms have been discussed and illustrated by several examples.