Choice-based cluster consensus in multi-agent systems

In this paper, the cluster consensus problem for multiple agents whose state evolutions are described by identical linear models is investigated. Control decisions for each agent are dependent on the observed data and a parameter chosen at an initial time from a set of possible options. These options are referred to as agent choices. The observation pattern of the agents is summarized by an interacting graph. The key result of this paper shows that under suitable structural conditions of the interacting graph, there exists a choice-based distributed control algorithm that enables agents who make the same choices to converge to a common consensus state, while agents who make different choices will converge to distinct states. The structural condition is shown to be dependent on how the choice selections are distributed. For interacting graphs that satisfy the so called unicyclic property, an explicit characterization of the sets of choice selections that guarantee cluster consensus is presented.