A decision-making framework for stochastic deployment of wireless sensor networks

Due to reliance on stochastic deployment, delivery of large-scale Wireless Sensor Networks (WSN) presents a major problem in the application of WSN technology. When deployed in a stochastic manner, the WSN has the utmost challenge of guaranteeing acceptable operational efficiency upon deployment. Therefore, a systematic approach is required for planning stochastic deployments that meet acceptable efficiency levels. This dissertation presents a decision-making framework for planning and optimizing stochastic WSN deployments based on multiple efficiency metrics. The framework uses simulation, experimental design techniques, the Analytical Hierarchic Process (AHP), and Response Surface Methodology (RSM) to provide an innovative and complete approach that helps decision-makers determine deployment strategies from a set of alternatives.;The presented framework uses a two-pronged approach for prioritizing and optimizing deployment strategies. The first approach tackles deployment scenarios where data cannot be modeled using mathematical equations. In these cases, a two-step methodology is employed. First, the solution set is analyzed using the Vertical Variance Trimming (VVT) approach to remove deployment strategies that are statistically redundant. After execution of VVT, the Analytical Hierarchy Process (AHP) is used to prioritize deployment strategies based on decision-maker's deployment goals. The second approach provides analysis and optimization of stochastic deployments when data can be modeled using mathematical equations. In these cases, through Response Surface Methodology (RSM) and Desirability Functions, the datasets provided by simulation are analyzed to create non-linear predictive models using the polynomial approximation technique. From the prediction models, a unified efficiency metric representative of overall deployment efficiency is created using Desirability Functions. Finally, the response surface of the unified metric is optimized to provide results that represent optimum values of individual optimum responses, which are representative of the best deployment alternatives for a given scenario.;Through case studies, both approaches in the framework are proven successful. The first approach (i.e., VVT & AHP) resulted in 35% reduction of deployment strategies using the VVT technique. This greatly simplified the application of AHP by systematically eliminating deployment alternatives that result in higher cost and decreased network lifetime, but equal connectivity and coverage to existing alternatives. The second approach (i.e., RSM & Desirability Functions) is proven successful in modeling individual efficiency metrics, and in providing a way for analyzing deployments based on numerous efficiency metrics. In some cases, where deployment goals included optimization of all efficiency metrics, the overall deployment desirability (i.e., efficiency) resulted in 73%. In other cases, where deployment goals included optimization of partial efficiency metrics, individual optimized efficiency metrics resulted in high levels, however overall deployment desirability reached as low as 19%, which quantified the effects of partial efficiency metric optimization on overall deployment efficiency. Overall, the framework presented in this research proved to be feasible for efficiently managing and planning stochastic WSN deployments based on multiple efficiency metrics.