Multidimensional Sensor Data Prediction.

The ubiquity of modern sensor nodes capable of networking has incentived the development of new and intelligent embedded devices. These devices can be deployed anywhere and by connecting with each other they form a wireless sensor network (WSN). WSN add a new dimension to the world of information which enables the creation of new and enriched services widely applied in different industrial and commercial application areas. The subject of this paper is comparison of different techniques for reduction of overall energy consumption in WSN. Many techniques try to reduce the amount of data sent by the sensor nodes by predicting the measured values both at the source node and at the sink (base station). By doing that, transmission would only be required if the predicted value differs from the measured value by a predefined margin. The algorithms presented here treat the sensor data as part of a time series. They provide great reduction in power consumption and do not require any a-priori knowledge.