Travel demand matrix estimation methods integrating the full richness of observed traffic flow data from congested networks

Abstract

Road travel demand matrix estimation fuses prior or synthetic travel demand matrices with observed flow data. Due to technological advances, ever more observed link flows, speeds and densities are available, whereas rising congestion levels trigger the urgency to use robust and sound estimation procedures on them. This paper addresses difficulties when estimating travel demand using link flows observed on congested networks. Active bottlenecks on these networks influence flow values both upstream (queues will form) and downstream (flow is metered). This implies that, on such a network, observed link flow values may represent either 1) the unconstrained travel demand for that link, 2) a proportion of the capacity of a set of upstream links, 3) the capacity of the normative downstream link; or 4) a combination of these quantities. Which quantity each observed link flow represents depends on the specific traffic conditions in the network. If the assignment model used to assess the relationship between travel demand and link flow does not strictly adhere to link capacity constraints, flow metering effects of bottlenecks (2) are not accounted for and all traffic is considered unaffected (1), thereby forcing incorrect assumptions upon the estimation. Current practice is to derive unconstrained link demand values from flows affected by congestion (2, 3 or 4) and then, instead of the actual observed flows, use these link demand values during matrix estimation. As such, these methods exhibit poor tractability and robustness and do not integrate any information from the assignment model about the composition of routes on the observed links. This paper describes and compares three novel demand matrix estimation methods for large scale strategic congested transport models that use assignment models that strictly adhere to link capacity constraints and explicitly consider the conditions under which link flows are observed. It compares these methods to the current practice and gives practical insights from applications, demonstrating that these methods are more tractable and robust and allow for usage of observed congestion patterns and travel times from (big) data sources. Furthermore, these methods reveal inconsistencies between model link capacities and observed congestion patterns and between count values, allowing the modeler to correct the model network and other matrix estimation input.