Run-time Reconfiguration in Wireless Sensor Networks

Abstract

A wireless sensor network (WSN) consists of multiple small and simple computers (nodes), whose performance is tightly linked to its unpredictable deployment environment. It is nearly impossible to design a WSN that performs well in every scenario; instead they are developed for a specific context, with performance rapidly decreasing when environment properties move away from the optimum. Enabling a WSN to adapt to a changing environment could be a solution for this problem. The goal of this thesis project is to develop software for sensor nodes that is able to reconfigure a wireless sensor network at run-time, allowing the sensor network to perform within its requirements under changing conditions. An approach is proposed in which developers provide knowledge about the WSN’s requirements to sensor nodes during design-time, enabling the nodes to reason about their configuration at run-time. A middleware solution is designed that combines the run-time environment data with the knowledge about the requirements and indicates which reconfiguration is required. A proof-of-concept implementation is developed which reasons about reconfiguration within milliseconds. The performed evaluation includes the implementation of a reconfigurable modal analysis application and shows the processing and memory overhead introduced by the middleware remains low. The approach proves to be a solid basis for future developments on reconfigurable sensor networks.