Evaluation of data-driven performance measures for comparing and ranking traffic bottlenecks

Abstract

Traffic analysis procedures are becoming more robust over time. However, there is still significant room for improvement both in practitioners’ adoption of robust methods and in robustness of the methods themselves. One example of oversimplified practice involves peak hour or 30th highest hour analysis, which fails to capture the impacts of varying demands and operating conditions. To address this limitation, detailed reliability modelling procedures were developed for the Highway Capacity Manual (HCM), but these procedures bring challenging input data and calibration requirements. A new generation of data-driven tools is capable of real-time congestion identification, but their performance measures are only beginning to improve and evolve. Finally, comparing and ranking congested locations (i.e. bottlenecks) on the basis of experience and judgment lacks credibility unless backed by quantitative results. This study discusses the development of new and innovative performance measures for congestion measurement. A case study of ranking eight real-world bottlenecks based on the proposed measures and existing HCM measures produced new insights that could improve both approaches. Ideally, these new insights and methods would be accepted by agencies and/or commercial products for a new level of robustness in congestion measurement.