Machine learning based selfmanagement framework for unattended wireless sensor networks

Abstract

An Unattended Wireless Sensor Network (UWSN) consists of battery operated sensor nodes that operate in challenging operating conditions with minimal human intervention. Unattended Wireless Sensor Networks are used for extreme environmental monitoring, battlefield surveillance, and volcano monitoring, to name but a few. The dynamic nature of traffic arising from event-based applications of UWSN leads to congestion, high energy consumption, and presence of malicious nodes in the network cause poor performance. Thus, designing an energy efficient and reliable self-management framework which adapts dynamically to the change in network conditions is an important research. newlineIn this research work, we implemented the self-management framework in duty-cycled UWSN with randomly deployed static sensor nodes for both static sink and mobile sink. The duty-cycled UWSN consumes less energy because the nodes transit to different states such as active, idle, and passive based on change in network conditions. The self-management framework integrates the autonomic computing modules such as self-configuration, self-optimization, and self-protection. The self-management framework comprises of features such as boundary detection, topology control, data dissemination, and security. newlineThe boundary information is useful to determine the level of coverage of the network. The boundary nodes (or convex nodes) are used as a backup path during heavy load conditions in static sink scenario. newline