Prolonging the lifetime of wireless sensor networks using light-weight forecasting algorithms

The longevity of wireless sensor network (WSN) deployments is often crucial for real-time monitoring applications. Minimizing energy consumption by utilizing intelligent information processing is one of the main ways to prolong the lifetime of a network deployment. Data streams from the sensors need to be processed within the resource constraints of the sensing platforms to reduce the energy consumption associated with packet transmission. In this paper we carried out both simulation and real-world implementation of light-weight adaptive models to achieve a prolonged WSN lifetime. Specifically, we propose a Naive model that incurs virtually no cost with low memory footprint to realize this goal. Our results show that, despite its minimal complexity, the Naive model is robust when compared with other well-known algorithms used for prediction in WSNs. We show that our approach achieves up to 96% communication reduction, within 0.2 degrees error bound with no significant loss in accuracy and it is comparable in performance to the more complex algorithms like Exponential Smoothing (ETS).

I. Introduction

In recent years wireless sensor networks have received lots of attention due to their increasing capabilities, nevertheless, WSN deployments are still characterized by few bottlenecks, mostly related to energy resource management because they are mainly powered by batteries. Extending the lifetime of WSN deployment means using the available energy resources more efficiently. Most practical application scenarios will require a long-term monitoring capability, for example, a continuous monitoring of the evolution of temperature changes in a building could provide enough information for modelling the thermal behaviour of the building envelope and thus help the architects and civil engineers to decide when to carry out the necessary maintenance. Similarly, enough information on the background and statistical properties of a system are required to understand its behaviour, hence WSN plays a very important role in this process.