Multi-Class Hierarchical Random Networks for Consensus-Based Information Filter

This study proposes a general multi-class hierarchical random network with an arbitrary number of agents and an arbitrary connecting success probability to analyze the minimum communication costs with the minimum consensus iterations in distributed multi-agent networks. The proposed hierarchical random networks facilitate a flexible network topology that is well-suited for implementing consensus filters and verifying network connectivity. Further, the study validates the connectivity conditions, determines the upper bound of diameters, and calculates the minimum number of expected network connections for the proposed multi-class hierarchical random network. Total network connections, communication costs, and root mean square errors of consensus fused estimations in the proposed multi-class hierarchical networks are compared via a 2-D target tracking scenario. The simulation result demonstrates that the proposed multi-class hierarchical consensus filter can achieve maneuvering target tracking while avoiding the polynomial increment of communication costs.