The Use of Artificial Neural Networks in Predicting Air Pollution in Cities - Hardware Implementation Issues

The paper presents a concept of an air pollution monitoring system based on a wireless sensor network, in which particular sensors are able to predict future pollution levels within a specific time horizon. In the solutions reported in state-of-the art literature, forecasting the changes in pollution levels is usually based on the use of artificial neural networks (ANNs), globally trained for a given area, e.g. a city. In contrast, we propose to equip the sensors with hardware-implemented ANN s that could make the prediction based on data that the network obtains from the place where particular sensor is installed. Such networks require appropriate hardware optimization to maintain low energy consumption of the overall sensor. For this reason, in this paper we present a method of optimizing one of more complex components that is involved in determining the value of neuron weights in the network training process. In this work, we focus on the backpropagation (BP) learning algorithm, which results from the widespread use of this type of network in predicting air pollution levels.