Development of Laguerre Neural-Network-Based Intelligent Sensors for Wireless Sensor Networks

The node of a wireless sensor network (WSN), which contains a sensor module with one or more physical sensors, may be exposed to widely varying environmental conditions, e.g., temperature, pressure, humidity, etc. Most of the sensor response characteristics are nonlinear, and in addition to that, other environmental parameters influence the sensor output nonlinearly. Therefore, to obtain accurate information from the sensors, it is important to linearize the sensor response and compensate for the undesirable environmental influences. In this paper, we present an intelligent technique using a novel computationally efficient Laguerre neural network (LaNN) to compensate for the inherent sensor nonlinearity and the environmental influences. Using the example of a capacitive pressure sensor, we have shown through extensive computer simulations that the proposed LaNN-based sensor can provide highly linearized output, such that the maximum full-scale error remains within +/-1.0% over a wide temperature range from -50 degrees C to 200 degrees C for three different types of nonlinear dependences. We have carried out its performance comparison with a multilayer-perceptron-based sensor model. We have also proposed a reduced-complexity run-time implementation scheme for the LaNN-based sensor model, which can save about 50% of the hardware and reduce the execution time by four times, thus making it suitable for the energy-constrained WSN applications.