| Summary: | 碩士 === 淡江大學 === 交通管理學系 === 86 === The focus of this thesis is on the applying fuzzy set theory in commuters’day-to-day dynamic departure time and route choice behavior. Emphasizing the commuters’ dynamic decision behavior mechanism including the learning mechanism and individual prediction process.
The drivers’ travel choice behavior is by its own nature a complex issue which makes observation difficult. A controlled experiment was therefore implemented in this study where a group of selected auto-driving commuters interacting with a simulated commuting context were repeatedly observed. The participants’departure time and route choices are then recorded along with their individual statement regarding their own perception to the decision related attributes.
Under a day-to-day dynamic framework, driver perceptions of uncertain outcome of attributes affecting their route choice is considered due to the vagueness rather randomness. A rule-based reasoning process is therefore applied to model the observed behavior rather than the commonly used utility maximization. The dynamic learning mechanism of drivers’fuzzy recognition and fuzzy indifiereace bands were established through fuzzy time series model and fuzzy linear regression model. Finally, to realize commuters’predicting behavior, a fuzzy predicting model were established through fuzzy linear regression model as well.
Main findings may be concluded in the following:
1. Individuals did present day-to-day dynamic recognition of travel decision related attributes in the forms of changing shape of the corresponding fuzzy membership function.
2. The fuzzy indifference band for both departure time and route choice is around three minutes.
3. Using the calibrated fuzzy model for comparative validation, we fund the percent right of prediction is quite high, most cases showed over 70% or higher.
4. Further analysis has shown that information availability as well as travel decision dimension itself would affect the individual’s prediction for travel time.
5. This study was successfully combined a modeling structure originated from probabilistic model but to implement fuzzy elements and reasoning process.
This study can provide useful information as a priori for studying dynamic travel behavior for both modeling structure and behavior analysis when fuzziness rather than randornness was involved in the future. Results of this study is also useful as supplement for issues regarding dynamic traffic assignment and traffic control.
|
|---|
