Compressed Cluster Sensing in Multiagent IoT Control

Traditional multiagent system control relies primarily on inter-agent local communications. In large-scale IoT systems it may appear hard to synthesize local control actions in a simple manner. Cluster control possibilities are thus thoroughly discussed, with possible control goals stated. The relation between multiagent state sparsity and cluster patterns is illustrated, for further utilization of sparsity for cluster control. Consequently, the problem of cluster identification is stated, and a possible solution is proposed in the form of the compressed cluster sensing algorithm. The algorithm utilizes compressed sensing methodology for a compact representation of agent states, which is then used to synthesize compact control actions in a low-dimensional space, without requiring to specify cluster locations.