Laguerre neural network-based smart sensors for wireless sensor networks

A wireless sensor network comprises of several nodes (also called motes). A mote communicates with other nodes based on the information collected through the sensor module attached with multiple sensors, e.g., accelerometer, pressure, temperature and humidity sensors. It is important that the sensors provide accurate readout of the physical quantity that they sense, especially when the motes are operated in harsh environments. In this paper we propose intelligent sensors for the sensor module using a computationally efficient Laguerre neural networks (LaNN) to auto-compensate for the associated nonlinearity and environmental dependence, and provide linearized sensor readout even when the motes are operated in harsh environments. By taking an example of a capacitive pressure sensor, through computer simulations we have shown that the LaNN-based sensor model can provide highly linearized sensor output. The performance of the LaNN sensor model is compared with a multilayer perceptron-based sensor model, and it is observed that the former model is superior in terms of computational efficiency while providing similar linearity performance.