Investigating Freeway Speed-Flow Relationships for Traffic Assignment Applications

摘要

Developments in high resolution traffic sensors over the past decades are providing a wealth of empirical speed-flow data. Travel demand models use speed-flow relationships to assign traffic flows to network links. However, speed-flow relationships have not been revalidated against new detailed traffic sensor data. Therefore, it is necessary to revisit speed-flow relationships based on actual measured conditions on network links rather than assuming constant speed-flow relationships over entire highway network systems. Speed-flow relationships have been particularly difficult to calibrate and estimate when traffic volumes approach capacity, i.e. when the v/c ratio approaches one. This thesis empirically evaluates the speed-flow relationships for v/c u3c 1 using field data. For congested conditions (v/c u3e 1) a theoretical approach is taken. A new methodology to determine the distribution of the activation of bottlenecks, bottleneck duration, and bottleneck deactivation is proposed. This thesis is a new contribution to understand the stochastic nature of freeway capacity as well as bottleneck duration, activation, and deactivation. Unlike previous research efforts, this thesis studies speed-flow relationships at the lane level and later presents a method to estimate speed-flow relationships at the link level.