At the 2018 annual flagship conference of the IEEE Intelligent Transportation Systems Society (ITSS), Yaroslav Hernandez will be presenting a paper on “Unsupervised Incident Detection Model in Urban and Freeway Networks” (Tuesday, 14:10-14:30, Paper TuCT2.4).
Reducing the effects of incidents through early detection is a crucial requirement for incident management. This paper presents an automated incident detection model based on an unsupervised approach that uses only traffic observations as model input. First, a new self-tuning statistic is introduced as a feature generation function to capture the spatio-temporal relationship of traffic data in both urban and freeway networks; these features are then used as input in the segment-based mixture model that learns complex data distributions and their parameters. The Mahalanobis Distance is used to determine whether the traffic observation corresponds to an incident or recurrent traffic state. The model performance is demonstrated for two networks: freeway with real data, and urban with simulated data. Results show that the developed method achieves high accuracy rates and early incident detection compared to widely used approaches, such as the California algorithm series and their extensions.