A Fuzzy Framework for Self-Aware Wireless Sensor Networks

Abstract

We have developed a fuzzy logic framework to simulate self-awareness in wireless sensor networks (WSNs) that envisages data transmission losses due to abnormalities in the operating environment. Fuzzy logic is selected to model the awareness, because it requires less data to encode, and occupies less memory, making it suitable for sensor nodes with limited resources. We view WSN as a multi-layered infrastructure, where a central-server in the top-layer manages the gateways in the intermediate-layer, and the gateways are aggregating the raw sensed data from the bottom layer’s sensors. The self-awareness is simulated using Mamdani Fuzzy inference system (FIS) that utilizes a custom-defined membership functions of the selected parameters, such as temperature, humidity, wind speed, and battery residual-level to guide WSN through its data transmissions. The input and output parameters are defined with three linguistic variables (low, medium and high) and a predetermined set of rules is applied to the selected FIS to determine the impact level. The impact is mapped into packet error rate and the gateway then decides whether to carry out the data transmissions while being aware of the selected uncertainties and their impact. The experimental results on a multivariate dataset integrating real-time meteorological data with randomly generated battery voltage revealed that the proposed FIS framework demonstrated 38% energy savings by non-executing unworthy communications against normal transmissions.