The aim of this work is to realize a device with semiconductor gas sensors for the measurement of the two more important air pollution components: ozone and ''automotive traffic'' pollution. This device is made with three different thick film layers deposited on an alumina heating substrate. Electrical conductivities are measured at different temperatures. Two kinds of calibration models are made with a Neural Network Algorithm and are tested: a conventional one made with artificial gas standards and a more original where real pollution measurements are used to calibrate the multisensor. Interesting results are obtained and portability, or universality, of the models is discussed.