Cost performance based control reconfiguration in multi-agent systems

In this paper, a reconfigurable control protocol for a linear multi-agent system seeking consensus in presence of actuator faults and saturations is investigated. The control protocol consists of two parts: healthy system controller and faulty system controller. The main objective of the healthy controller that is designed off-line is to determine the protocol which guarantees team consensus and minimizes the team performance index. On the other hand, the faulty system controller is designed on-line and aims at stabilizing the consensus error signals and optimizing the faulty team performance index subject to constrained input signals. Both centralized and semi-distributed architectures for control reconfiguration are developed. In the centralized scheme, the team is considered as a single unit and the central module reconfigures the entire system controllers, while in the semi-distributed approach faulty agent uses its immediate neighbours information (i.e. measurements and control gains) to reconfigure its controller.