Embedded Artificial Neural Network for Data Prediction in Energy Efficient Wireless Sensors Networks

Description

In this paper, we propose a novel strategy for minimizing energy consumption in a Wireless Sensor Network using embedded machine learning techniques. Our approach uses a Convolutional Neural Network model embedded in a cluster head node to predict nodes' data to minimize communication instead of transmission of real one. Thanks to this method, the sensor nodes can remain in idle mode and save 90% of their energy in comparison to normal transmission. In order to validate our algorithm, we have developed a wireless network consisting of electronic boards that can communicate in LoRaWAN, which are compatible with The Things Network server. The results of this test bench show that with this method, it becomes possible to reduce the number of transmissions in the wireless sensor network and to put the sensor nodes in idle mode much longer. This technique allows us to reduce Energy Consumption, and therefore, increases the overall lifetime of the wireless sensor network.

Abstract

International audience

Additional details