Energy-Efficient Reprogramming In Wsn Using Constructive Neural Networks

In this paper, we propose the use of neural network based technologies to carry out the dynamic reprogramming of wireless sensor networks as an alternative to traditional methodology. An analysis and comparison of the energy costs involved in reprogramming wireless sensor networks was done using rule-based programming (TP), standard feedforward neural networks (FF), and the C-Mantec (CM) algorithm, a novel method based on constructive neural networks. The simulation results, first performed on an array of sensor networks under the COOJA simulator (considering best, medium and worst case scenarios for three benchmark problems) and finally evaluated on a case of study with identical conditions, show that the use of neural network based methodologies (FF & CM) produces a significant saving in resources, measured by the number of packets transmitted, the energy consumed and the time needed to reprogram the sensors.