Abstract
Multi-hop network paths and path redundancy enhance the reliability of communications in the Graph Routing of Industrial Wireless Sensor Networks (IWSNs). However, the centralised management of IWSNs creates unbalanced energy consumption between battery-powered wireless sensor nodes. This creates a hotspot challenge in Graph Routing. The inability to balance energy consumption with single-objective paths in mesh topologies was revealed in our previous work that used the Covariance-Matrix Adaptation Evolution Strategy (CMA-ES) to improve Graph Routing [1]. In this paper, we address this problem by combining the single-objective paths of the Graph Routing algorithm with a pre-set unequal clustering topology. First, this study detects isolated nodes within pre-set unequal clustering and then creates new clusters for these nodes. Second, the objective function of CMA-ES is used to select the best cluster head by considering node centrality between the nodes in the same cluster and the distance between other cluster heads or the gateway. Once the cluster heads are selected, single-objective paths of Graph Routing that are minimum-distance (PODis), maximum residual energy (POEng), and minimum end-to-end transmission time (POE2E) can be effectively evaluated. Simulation experiments reveal that using the single-objective paths of Graph Routing by the topology of pre-set unequal clustering results in more balanced energy consumption.