Multi-objective Optimization In Wireless Sensor Network Deployment

Wireless sensor network(WSN) deployment is a prerequisite for the WSN to achieve its corresponding function. Excellent deployment scheme leads to robust network topology, and long network lifetime. In general, many factors should be considered in the WSN deployment, for example numbers of node, surveillance area coverage, total energy consumption, network lifetime and connectivity.WSN is modeled as a graph, where energy consumption can be determined by estimating the paths between the pairs of nodes. The minimum spanning tree of the graph structure is used to guarantee the connectivity and total energy consumption of whole network and energy balance quality are used to determine the lifetime. The deployment problem becomes a multi-objective optimization problem whose goals are keeping network connectivity, maximizing network coverage, minimizing total energy consumption and balancing the nodes' energy consumption.This problem is solved by a generic algorithm based on the non-dominated sorting(IMNSGA) to compute the non-dominated hierarchy of each individual. Congestion factor is utilized to compensate the population diversity. In order to improve its performance, this e multi-objective optimization algorithm replaces the base point with ideal point to build the multi-dimensional decision space. In the solving process, dimensionality reduction approach is adjusted and spatial crowding factor is utilized to examine individual of populations. In order to avoid the local aggregation, a monitoring function that can detect whether the population has a local accumulation is presented and abruption of individual is used when the local aggregation occurs to increase the opportunity of optimal solution.The deployment scenario of the WSN is obtained through the simulation experiment and IMNSGA is implemented to generate deployment schemes in these simulations. We compare the IMNSGA with the multi-objective optimization algorithm NSGA-II and MOEA/D and the existing deployment solution in the WSN. Simulation results show that the IMNSGA performs better than NSGA-II, MOEA/D and the traditional WSN deployment. The resulted Pareto optimal solution reflecting coverage area, total energy consumption and balancing of nodes energy dominate the solutions obtained through NSGA-II and MOEA/D. Compared with the existing deployment approach, the solution under the threshold of the coverage leads to a longer lifetime.