The Day-to-day Traffic Dynamics Based On Travelers' Learning Behaviors

Travelers' day-to-day route choice behavior lead to the traffic dynamics.When travelers make route choice,they face complex and uncertain environment.Travelers' day-to-day route choice is a learning and adoptive process under the influence of traffic information.There are three kinds of traffic information which can effect travelers' route choice: ATIS information,travelers' own experience and other travelers' experience.So far,there have been much research about the influence of ATIS information on the day-to-day traffic dynamics.However,the effect of travelers' own experience and other travelers' experience have not been researched sufficiently.According to the shortcomings of the existing studies,this dissertation uses reinforcement learning and social learning to study the influence of travelers' own experience and other travelers' experience on dayto-day traffic dynamics.The methods of questionnaire survey,statistics,simulation and mathematical analysis are used in this dissertation.Travelers' day-to-day route choice are analyzed and modeled.The multi-agent simulate model and analytical model about the network flow are also be modeled.The rules and characteristics of the evolution of network flow are analyzed.The main contents and innovations of the dissertation are summarized as follows:(1)Reinforcement learning is the possible learning style when travelers only depend on their own experience to make route choice.However,the research which uses reinforcement learning to study travelers' route choice is not sufficient.Therefore,this dissertation uses reinforcement learning to model travelers' day-to-day route choice,travelers' memory,learning rate and experience cognition are taken into account.The model is used to analyse the properties of the evolution of the network flow with the influence of reinforcement learning.Results show that the network flow can converge to user equilibrium(UE)if travelers can remember all travel time they experienced,but which is not necessarily in the case under limited memory;Learning rate can strengthen the flow fluctuation,but memory leads to the contrary side.Moreover,high learning rate results in the cyclical oscillation of flow distribution.Finally,two scenarios are analyzed to illustrate the model's application.(2)So far,there have been research which comfirms that individuals' mode choice and destination choice can be influenced by other travelers' behaviors,but not in the aspect of route choice.In order to fill the research gap,the dissertation surveys the influence of other travelers on individuals' route choice by using questionnaire.Statistical results show that: it is very common that travelers communicate travel experience in daily life;many travelers have the experience that they change route choice because of the influence of other travelers.These verify that individual traveler's route choice is indeed effected by others' travel experience.Furthermore,the mechanism that others' experience influence individuals' route choice is analyzed.The research model is built and the verification of structural model shows that: sources' expert,sources' trustworthiness,information's relevance and information's accuracy all play significant roles in influencing an individual's perceived usefulness for other travelers' experience.Perceived trust on information indirectely affect an individual's adoption for others' experience.Extroverts are more likely to adopt other travelers' experience.Travelers with field-dependent congnitive style and those with fieldindependent congnitive style don't show significant differences in the aspect of adopting others' experience.(3)The day-to-day traffic dynamics considering other travelers' experience are modeled.The exsiting research of day-to-day traffic dynamics rarely considers the influence of other travelers' experience.Individuals can obtain others' experience though social interaction.After comfirming the influence of others' experience on individuals' route choice,the dissertation models travelers' day-to-day route choice under the effect of social interaction and social learning.The bottom-up method(from the microscopic level to the macroscopic level)is used: travelers' day-to-day route choice is modeled firstly and then the evolution of network flow are deduced based on the route choice model.The existence of equilibrium and the stability of equilibrium are analyzed and proved.Some interesting properties of the evolution of network flow and travelers' behaviors are also analyzed and proved.Two interesting phenomena are found in our study: A traveler's interaction number can affect his route choice strategy at the individual-level.However,it exerts no influence on the evolution of network flow at the aggregate-level;When the evolution of network flow reaches equilibrium,route choice strategy at individual-level is not necessarily invariable.Finally,two networks are used as numerical examples to show the consistencies of the microscopic model and macroscopic model as well as the properties of the two models and the interesting phenomena.