Building Distributed Soft Sensors

Abstract

Advances in digital electronics have enable the development of low-cost, low-power, multifunctional sensor nodes that are small in size and communicate in short distances. These tiny sensor nodes consist of sensing, data processing, and communication components, leverage the idea of Wireless Sensor Networks (WSN) based on collaborative effort of a large number of nodes. Despite the high number of publications dealing with WSN applications, there are still some potential to be explored in WSN development and maintenance. Recent contributions of Intelligence Modeling have made possible the construction of Smart Sensors, reducing maintenance costs and time prototyping. In this scenario, Neural Networks (NNs) have found many successful applications in nonlinear system identification and control, digital communication, pattern recognition, pattern classification, etc. Many similarities between NN and WSN can be found and explored to improve WSN application process by reducing the development costs. For example, the sensor node itself can be seen as an artificial neuron, since the WSN application shows characteristics such as distributed representation and processing, massive parallelism, learning generalization ability, adaptively, inherent contextual information processing, fault tolerance and low computation. This paper examines the hybridization with NN and WSN into a Smart Home application, called Smart Table. Preliminary prototypal results have shown that Multilayer Perceptron is good candidate for using into low-cost System-on-a-Chip (Soc) such as PIC microcontrollers.