Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 43999 (09)
- University: Sharif University of Technology
- Department: Civil Engineering
- Advisor(s): Zakaei Ashtiani, Hedayat
- Abstract:
- Accurate and realistic forecasts of traffic pattern in the networks can affect many decisions of urban transportation planning. In recent years, traffic assignment (TA) using stochastic models have received more attention by researchers of this field. Stochastic approach of route choice modeling using random utility theory has specifically developed to relax restrictive assumptions of traditional deterministic model and to address randomness nature of travelerʼs behavior. Together with development of discrete choice theory, more advanced models have been utilized to better represent route choice behavior. Although there have been considerable researches in the context of stochastic traffic assignment (STA), but most of them have concentrated on limited issues such as model specification (or model selection) and proposing efficient algorithms to probabilistic loading of networks, and hence some aspects of this approach have been neglected. In this research, in addition to a comprehensive review on theoretical principles of choice models used in TA, the results of applying these models on pedagogical networks have also been investigated. Some basic issues of this thesis are as follows:
All stochastic models of TA contain calibration parameters which reflect route choice behavior and hence include important information from modeling point of view. But the problem of estimation of these parameters has received little attention in the literature of TA and most of researchers have used predetermined typical values for these parameters and run their models with these typical parameters. In this research, a heuristic and practical method is proposed to calibrate these models. Although this method utilizes a somewhat strong assumption, but at least can be viewed as an alternative for using raw typical values which in some occasions can be highly misleading. Some estimates have also been proposed based on an experimental data set collected in the city of Tehran. Our experience clarified that model calibration can affect accuracy of prediction of traffic pattern much more than model selection. Using simple but calibrated models can be more justifiable than using sophisticated models with typical parameters.
Path-based algorithms of TA have been implemented more in recent years. Before, there was an insistence on using link-based methods, just for computational considerations. It has been shown that explicit treatment of path flow variables, which are actual alternatives of travelers, allows using more advanced models with sounder assumptions. In this research, we have used path-based methods of STA. We have applied a simulated path generation algorithm which produces fixed sets of alternatives prior to running assignment procedure. Our investigations showed that at least for the scale of our illustrative network, solving TA problem using path flows is not computationally expensive as it may seem, since generating too many paths is not actually required.
Another superiority of stochastic approach of TA is its more flexibility than the deterministic one. To illustrate this potential of flexibility, we have proposed and applied a multi-class model of STA which has the capability to evaluate benefits of traveler information systems in the transportation networks - Keywords:
- Logit Model ; Probit Model ; Stochastic Traffic Assignment ; Route Choice Modeling ; Discrete Choice Model
Digital Object List
-
محتوي پايان نامه
- view
Bookmark