Investigations on Machine Learning Algorithms for Energy Efficient Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) are used to span physical spaces to cyberspace, and there is a newlinewide spectrum of uses from environmental monitoring to industrial processes. The work outlined newlinein this thesis involves networks of microelectromechanical systems (MEMS) used as nodes on newlinemultiple areas and a base station operating on constrained energy supplies. In some cases, where newlinerecharging is not possible (hostile or remote locations), creative alternatives are needed to conserve newlineenergy and keep networks operating sustainably. newlineIn order to resolve these challenges, this work develops a framework depending on using Q newlinelearning and machine learning based clustering and cluster head selection in WSN. This research newlineintroduces a novel Optimized LEACH protocol architecture that is based on the Low-Energy newlineAdaptive Clustering Hierarchy (LEACH) protocol and is geared to single structure and multi newlinestructure WSNs. An enhanced protocol uses adaptive clustering, cluster head swapping techniques, newlineand dynamic transmission power control, where energy consumption is reduced and throughput newlineincreased. Moreover, the machine learning based predictive analysis is applied for cluster newlineformation, and improves leader election and resource allocation strategy. A Q learning framework newlineis integrated to address dynamic cluster head selection and routing issues, and produces near newlineoptimal solutions in real time environments. newlineSuch results show significant improvements in system level metrics such as network lifetime, newlinethroughput and latency. It has Q learning to learn and respond adaptively to network conditions, newlinesuch as energy levels, densities of nodes and traffic of data. Clustering processes receive artificial newlineintelligence treatment and may lead to more energy efficiency across the network