Advancements in persistent and non-persistent delay forecasting: a macroscopic intersection modelling approach

Abstract

The ability to forecast travel times in road networks is important for transport planning and traffic management. On urban roads, travel times are predominantly influenced by intersection delays, which consist of persistent and non-persistent delays. Traffic assignment models are commonly used to predict traffic flows and travel times, with network loading (NL) being a core component that propagates traffic from origins to destinations. However, existing macroscopic NL models cannot effectively address both persistent and non-persistent delays. Persistent delays are inadequately modelled by existing node models due to their incapability of (i) accounting for the influence of approach lane configurations, and (ii) consideration of endogenous exit link capacities due to differences in exit speeds for different movements. Moreover, non-persistent delays are generally completely ignored in macroscopic NL. This thesis addresses these deficiencies through a comprehensive macroscopic intersection modelling approach. Firstly, a systematic literature review on the development of node models is conducted. Subsequently, this thesis proposes three advancements, namely: (i) an innovative macroscopic node model that allows for separate queues for each turning movement, while considering approach lane configurations, (ii) an approach to incorporate endogenous exit link capacities in node models, and (iii) a method to endogenously incorporate non-persistent delays into the NL propagation scheme. Each advancement is mathematically formulated, and efficient solution algorithms are developed. The capability of these approaches is tested through numerical examples, demonstrating their ability to improve travel time forecasts in urban intersections. The advancements proposed in this thesis not only enhance the current modelling techniques but also provide a foundation for future research in macroscopic transport modelling and real-time traffic management applications.