Creating representative urban motorway traffic scenarios: initial observations

Traffic patterns are useful for analyzing and identifying representative traffic scenarios from traffic data. Traffic scenarios are important when machine learning is used for traffic control to ensure good controller performance in all cases. This article tackles the problem of identifying relevant scenarios from clustered data for urban mobility analysis. The unsupervised learning approaches k-means, principal component analysis, and self-organizing maps were applied on real traffic data from Slovenian motorways to analyze and group traffic scenarios. Obtained observations present a solid foundation for future research on a wide-scale data-set, including data from more measuring points for creating relevant traffic scenarios for learning of traffic controllers.