A Post-processing Technique for the Four-step Travel Demand Modeling Executed Through a Feedback Loop

摘要

The four-step travel demand model (FSTDM) is the most widely used technique in practice for estimating traffic flow pattern in the transportation network. Traffic assignment problem (TAP) is a key step of the four-step process that determines the flow pattern which forms the basis for scenario analysis of a transportation improvement project. In many cases, different scenarios may differ by little, but FSTDM may lead to network flows which differ significantly or suggest improvements which are inconsistent with the network situation. This inconsistency arises due to solution noise. There are two main sources of this noise; first, the interdependency between the trip distribution and trip assignment steps of FSTDM, and second, the lower level of convergence in the traffic assignment step. This paper presents a methodology to address the aforementioned two issues by a post-processing technique incorporated through a feedback mechanism in the FSTDM. The post-processing technique consists of SMPA-hybrid, perturbation assignment and Origin-Destination (O-D) prioritization schemes. SMPA- hybrid is an improved implementation of traffic assignment algorithm labeled slope-based multi-path algorithm (SMPA) developed by Kumar and Peeta (2010). There are three methodological contributions of this paper. First, the paper presents an enhanced travel demand modeling framework, second, it formulates a hybrid approach by combining the merits of sequential approach and simultaneous approach of solution algorithms for the TAP, and third, it provides a methodology for the O-D prioritization in TAP. The results of computational experiments suggest that the SMPA-hybrid has a superior rate of convergence compared to the SMPA. The results further reveal that a warm start using perturbation assignment and O-D prioritization has significant benefits over the base case of cold start and non-prioritized implementation of the SMPA-hybrid.