Loading...

Multi-criteria Multi-class Traffic Assignment Using Mixed Generalized Extreme Value Models

Shahhoseini, Zahra | 2013

516 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 44673 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Poorzahedy, Hossain
  7. Abstract:
  8. Relaxing the underlying assumptions of transportation networks analysis and approaching to more realistic behavioral assumptions in route choice modeling, allows more precise predictions for equilibrium traffic pattern. Furthermore, enhancing the degree of sensitivity in modeling traffic equilibrium is crucial from planning perspective in the sense that it can make transportation analyst capable to evaluate and predict the impacts of a broad spectrum of supply and demand management policies. This research has concentrated on proposing some generalizations in route choice behavior modeling and relaxing some restrictive assumptions of the previous approaches in this area. Formally speaking, in most current models of traffic assignment, some or at least one of the following assumptions has been imposed to the modeling procedure: it has been assumed that: (a) Travellers have a perfect knowledge about network conditions. (b) Commuters have an unlimited ability in finding and calculating the “best” route. (c) All drivers follow a similar and consistent route choice behavior, and accordingly all are classified in a unified category or class in the sense of perceiving and sensitivity to route attributes. (d) Travel time is the exclusive representable explanatory variable which affects route choice decision. (e) Random utilities of different routes in a network are all uncorrelated, and (f) the coefficient of travel time variable in routes’ utility functions is same for the whole population of decision makers. In this research, assumptions (a) and (b) by using random utility, assumption (c) by multiclass modeling (classification for income and equipment by ATIS), assumption (d) through referring to multi-criteria or multi-variable modeling, assumption (e) through application of more generalized logit models than the multinomial logit, and assumption (f) through following mixed (or random coefficient) modeling approach have been relaxed. Adding more variables than the travel time can enhance the explanatory power of the models. Taking socio-demographic aspects of commuters into consideration in modeling route choice behavior has never been reported in traffic assignment literature. Making use of monetary cost variable makes the model capable to assess network pricing policies. Classification of network users into equipped and unequipped with ATIS can cope with the need for evaluating benefits of these systems; and, application of random coefficient choice models captures random taste variation and heterogeneity in perception and sensitivity to travel time, distributed over commuters’ population. In order to enjoy theoretical benefits of advanced generalized extreme value models, such as paired combinatorial logit or cross-nested logti (for capturing path-overlapping) and mixed choice models (for representation of heterogeneity), we have made use of an unexploited capacity of random utility theory for developing new models and have utilized “mixed generalized extreme value models”, which have not been used before in econometrics contexts. In addition, due to the intrinsic difficulties of calibrating route choice models based on revealed preferences, described in this research, we have referred to a set of stated choices data, collected in the city Tehran, for estimation of our models. Formulation of the problem and the solution algorithm for multi-criteria multiclass traffic assignment has been presented and application of the proposed models on illustrative networks has been investigated. The ability of the models for evaluation of supply and demand policies has also been demonstrated, and as well, a comparison has been conducted on the relative importance of different generalizations, proposed for route choice modeling, on the solution and result of traffic assignment
  9. Keywords:
  10. Multi-Criteira Traffic Assignment ; Multi-Class Assignment ; Mixed Generalized Extrem Value Models ; Stated Preference

 Digital Object List

 Bookmark

No TOC