An Investigation On Characteristics And Travel Mechanism Of Changing Commute Mode And Commute Time Of Relocated Residents Living In The Urban Area

Triggered by the requirement of the urbanization and urban developments,many cities in China are undergoing a large scale of new district construction and traditional area reconstruction,making cities expand their urban area,remodifying their urban functions in spaces,and releasing the over-burdened population pressure in the city center.Meanwhile,multiple factors such as housing system reforming and housing marketing make effects resulting in different flows of residential relocation occurring in the urban area.Moreover,the single land use pattern and the inconsistent development of industry and residence in suburbs cause a serious job-housing imbalance in the newly moved area,where the commute distance of these relocated residents is therefore largely lengthened and the proportion of car use in commuting is significantly increased.All these threaten the city with a series of issues such as the frequent traffic tidal commuting flows,the worsening traffic congestion in peak hour,the severe air pollution caused by vehicle emission,and environmental pollution.In view of this,more and more researchers in different fields such as urban planning,transportation planning and geography area start to pay their attention to the changes and mechanism of relocated commuters’ travel pattern,making trials to design urban planning or transportation planning policies that are able to alleviate the urban traffic problems triggered by residential relocation.Based on projects funded by National Natural Science Foundation of China titled “Multidimensional travel decision mechanism of living relocation commuters’ activity chain”,“The influential mechanism and feedback optimization of spatial changes in the outlying area of old city on traffic travel changes”,and “The mechanism of the built environment on the rail transit mode chain and the modal shift of car usage”,this research focuses on the changing commute mode and commute time of relocated commuters before and after residential relocation.It proceeds with steps as follows: establish the quasi-longitudinal database for relocated commuters based on multiple data sources from elaborately designed travel survey conducted in Nanjing and corresponding online data,which captures relocated individual’s personal and household attribute,relocation related attribute and commute travel attributes before and after relocation and measures the changing extent on how they perceive the built environment attribute surrounding their before and after communities;develop models respectively aiming to investigate influential factors and mechanism of these residents’ commute motorizing trend and travel time pattern after relocation;simulate the macro characteristics of the commute travel of motorized relocated residents by utilizing the multi-agent system;and based on the simulation models conduct the scenario test to measure the impacts of built environment optimization on commute mode choice and commute time.The main content of this dissertation is composed of four sections as follows:(1)Learning that the residential relocation which occurred in urbanizations of traditional area reconstruction and new district construction has posed a series of commute issues to these relocated residents,this study made trials to design a quasi-longitudinal survey collecting commute travel data of relocated residents from the individual’s perspective.Differing from previous cross-sectional data,the quasi-longitudinal data mainly covers relocated commuters’ characteristics relating to their personal and household,relocation and commute attributes,and particularly focuses on their perceptional changes on the built environment characteristics surrounding their residences before and after relocation.Two main data sources are utilized in the study: the survey data and the matched online data accessed through the API.We conducted a case study in Nanjing,China.Combining the reasonable sampling survey and online data,we established a quasi-longitudinal database on the commute behavior relocated residents before and after relocation in Nanjing.Using the basic statistical methods such as ANOVA test,we made a sample description on respondents’ commute characteristics and investigated the general correlation between commuters’ attributes and commute mode,commute time,and commute distance.To investigate the relationship between built environments and commute behavior,we developed a probit model indicating the general rule between the built environment and relocated individual’s commute mode choice,which mainly serves for the following commute mode choice analysis.(2)Focusing on the motorizing trend of relocated commuters after relocation,we developed two mode shift models for residents who shifted their commute mode to motorized mode after relocation.They were previous non-motorized users and public transit riders.To provide evidence for the guidance and control of people in the usage of the motorized modes,we investigated the commute mode shift of relocated residents from different perspectives.As the commute mode shift of the main motorizing group(the previous nonmotorized commuters)involves many factors which are potentially interrelated,we applied the Bayesian network which is more flexible and capable for the modeling.Beforehand a data preprocess on discretization and variable selection was made by the rough set and mutual information methods.Within the framework of attitude-based travel theory,we developed a Bayesian network for the mode shift of these previous nonmotorized users by Three-Phase Dependency Analysis(TPDA)algorithm and Maximum Likelihood Estimation(MLE)and measured the extent to which each factor influences the mode shift with the reasoning method.Meanwhile,we employed the Logistic regression model based on random utility theory to identify key influential factors of the previous public transit users’ mode choice decision.Both models have comprehensively incorporated the personal and household attribute,relocation attribute,and built environment attribute,and explored the detrimental elements and mechanism of motorizing mode shift of relocated commuters.(3)Viewing the characteristics of commute time change and current commute time pattern of relocated commuters,this study utilized statistical methods such as cross-sectional tables and ANOVA tests to analyze the general rules of commute time pattern before and after relocation,taking account of personal and household attribute,relocation attribute,built environment and commute related factors.To investigate the commute time characteristics of relocated residents after relocation,we established a multivariate linear regression model to determine its key factors.Additional trials were made to recover the current commute time pattern by discretizing the commute time through the equal frequency method and the surveyed optimal trip time.With the optimal time discretization,we developed three decision tree classifications using the J48 algorithm by three variable selection approaches,which are multiple linear regression,the Logistic regression,and information gain methods.The tree structure with the best performance obtained in the study was to interpret commute time pattern to induce corresponding policies that could favor the reduction of commute time.(4)Based on survey data on relocated residents’ commute travel in Nanjing,this research employed the Q-learning algorithm to develop a multi-agent simulation system for these motorized relocated commuters.Considering the correlation between commute travel decisions of these relocated commuters,this system constructed the agent’s reward function with the incorporation of models that were made for modal shifts of two main motorized groups and the commute trip duration in previous chapters to represent the commute mode choice,commute departure time,and commute trip duration of these target commuters.Using the reinforcement learning we developed a multi-agent system to simulate the relocated individual’s commute trip.Based on the simulation model,we tested the impacts of 9 scenarios on commute mode choices and commute trip duration for two types of relocated commuters who have shifted their travel modes to motorized modes.The scenario test was formed by policies that refine built environment in respect to public transit services and the slow-mode travel.Measuring results could be worked as effective assistances supporting urban planners in the policy design and optimizing the arrangement of built environment surrounding residents’ community.All effects are made to encourage the more usage of public transit and less car usage in the commute of relocated residents.This dissertation expects to draw more attention of urban planners and transportation planners to the commute travel behaviors of relocated residents in urban areas.The initial procedure in new peripheral area construction and traditional area reconstruction is more flexible and shapable in built environment arrangement around communities.Therefore,the timely adjustments on multi-dimensional built environments in community scale would attract more urban commuters into the public transit commute,and to the large extent decrease the negative impacts on urban transportation caused by the urban migration.Lastly,this study expects to provide theoretical supports and empirical references for the design of urban planning policies and traffic control countermeasures.