Road Traffic Noise Prediction: An Artificial Intelligence Approach

Present road traffic noise prediction models, such as TNOISE, use semi-empirical adjustments to account for factors that influence the noise level impacting a receiver. Most adjustments are based on actual sound level measurements, for example of noise attenuation by different ground types, and hence present models perform satisfactorily for the simple situations in which the measurements were made. However, accurate noise prediction in more complex situations is beyond the ability of such models, because determination of a comprehensive set of adjustments is defeated by the numerous possible variations in terrain characteristics, building geometries, and so forth. This paper describes how this problem can be overcome using a neural network approach to road traffic noise prediction. We demonstrate how a simple neural network easily mimics one of the present road traffic noise models, and how neural networks trained on grid-based data can learn to predict road traffic noise in complex situations.