This study employs Structural Equation Modeling techniques to investigate the relationship between demand for taxi, Uber and Lyft services, socioeconomic and built environment characteristics as well as access to transit and bike-sharing modes, for taxi zones within New York City. The results show that income and car ownership levels influence demand for these for-hire modes. Additionally, higher activity density and higher extent of mixed land-use are associated with increased demand for for-hire modes, while pedestrian-friendly street networks are linked with lower demand levels. Also, temporal destination accessibility as well as accessibility to transit and bike-sharing significantly influence demand for taxi, Uber and Lyft. The findings provide a better understanding of the link between for-hire modes and built environment as well as accessibility to other modes, which can be used to improve demand forecasting of taxi, Uber, and Lyft services in large cities.

Shared parking has received considerable attention over the past decade due to its potentiality of alleviating urban parking headache through improving utilization of the existing parking resources, especially in residential zones when those residents drive their vehicles out. However, little attention has been paid to the risk that the returned residents have no parking spaces, which quite affecting residents’ engagement. Based on the precondition that any resident has a parking space, the objective of this paper is to explore the potential for the shared parking spaces. Firstly, a novel reservation and allocation mechanism of shared parking was proposed. Secondly, an agent simulation model was developed to track the NP-hard problem, regarding the shared procedure as a queuing system. Finally, the optimal number of shared parking spaces was obtained through numerical tests. The result has proved that the proposed shared strategy has brought about vast improvements of both utilization and turnover rate of parking spaces, compared with the non-shared circumstances. This paper provides a novel method to the solutions of the proportion of shared parking spaces in the residential area.

The collection of traffic data is the first step of urban traffic flow forecasting. It is very important to grasp the change rules of urban traffic flow, grasp the evolution characteristics of urban road traffic system, and formulate scientific traffic control measures. The sampling data of traffic detectors is usually a part of the missing high-dimensional form. The traffic data in this paper contains rich multi-mode features to interpolate the missing parts. Taking the local road network of Hefei as the research object, three tensor restoration methods were used to repair the traffic flow data under different loss rates respectively, and the accuracy of data repair under the different missing rates of two cases of random loss and structural loss were discussed.

With the increase of urban rail transit system complexity, the disaster process usually show the disaster chain form, the structure presented by multiple disasters and its communication process between the two has the complex dynamic network. This paper stud-ies the critical evaluation model of the disaster transmission process in metro disaster chain by using the universal disaster chain network in urban rail transit system.First, it constructs the evaluation index system of important degree of communication process. Secondly, based on the characteristics of the subway disaster chain, an important evaluation model of the dis-aster propagation process is given. In the end, an analysis of a major fire accident at king's cross station in London, England, has proved the validity of the model.

Bikesharing programs in their current form have been in place for several years in many cities across the United States. Encouraging people to use bikesharing for their daily routine travels has numerous social, economic, environmental, and health benefits. Therefore, it is important to understand factors influencing bikesharing usage in different urban areas in order to improve the system and encourage more use. This paper investigates how built environment at both local and regional scales influences bikesharing usage in five large metropolitan U.S. areas in the U.S. The study areas include Boston, Chicago, Philadelphia, Minneapolis, and Washington, D.C. and the data consists of around 9 million bike trips in over 1,500 stations over a one-year period. Multi-level mixed effect regression model is built to predict the number of trips originated from each station with respect to the station’s built environment pattern, as well as the overall urban form in the entire city. The results are consistent with previous research on the effect of land use at the local level on bikesharing demand. At the regional level, results suggest that the overall walkability and job accessibility via bikesharing networks are significant factors influencing bikesharing activities and demand. Models developed in this study could be applied to other communities that are seeking to improve and/or expand their bikesharing systems, as well as cities planning to launch new bikesharing programs.

This paper explores an emerging strategy of transportation demand management, where dynamic and personalized incentives are transferred to travelers in order to influence their travel choices toward a more socially desirable state. We first show that user-based incentives are futile under the complete information and full rationality assumptions, as other travelers instantaneously fill up the room left by incentivized travelers, thus canceling any marginal benefit improvements. Therefore, day-to-day traffic dynamics models are adopted, where commuters update their perceived travel time through trying and learning, based on which choices are made. A mathematical programming model is proposed for the system operator to design individually customized incentives on a daily basis. The effectiveness of travel incentives in improving the system benefits in the short and long runs is explored through numerical analyses. We identify a paradox that increasing incentive limits do not necessarily improve the system efficiency in the long run, due to probabilistic responses of travelers, i.e., we can influence while not control travelers’ choices.

The recent development in location-aware technologies and georeferenced social media provides us massive amount of data regarding individuals’ travel and activity participations in urban space across time. This paper develops a framework to systematically visualize the GPS-based activity diaries and associated emotions to better understand emotion changes across time and different activities/travel modes. The three components in the framework are: (1) data representation from Lagrangian or Eulerian perspectives, (2) classes of elementary and synoptic tasks in exploratory analysis of human emotions, and (3) selected visualization methods, with their representation of data and their achievable tasks. The paper presents a taxonomy for data representation and task specification and uses example visualizations to demonstrate how elements in the taxonomy can determine the selection and usage of certain visualization methods. This framework can serve as a reference to systematically select and organize visualizations for a project or identify needs for new methods given the characteristics of new data available in the future.

Non-motorized travel modes—mainly referred to as biking and walking—have drawn research attention for years. However, people’s knowledge of the non-motorized travel demand is still interrupted and limited. Existing studies either looked into the microscopic behaviors of bicyclists and pedestrians with high temporal resolution, or focused on the macroscopic behaviors with lower temporal resolution (e.g., annual). However, the combination of macroscopic spatial scale and relatively high temporal resolution have yet to be examined for the non-motorized modes. Hence, the paper proposes a two-module framework to estimate the number of biking and walking trips monthly at the metropolitan level. Various public domain data sources are utilized, including the American Community Survey (ACS), non-motorized count data, and regional household travel survey. They help to first estimate the number of annual non-motorized trips in the study area, which is later disaggregated by the monthly trend factors derived from the Poisson multilevel model (PMM). The application in the Washington–Arlington–Alexandria, DC–VA–MD–WV metropolitan statistical area (D.C. MSA) demonstrates the feasibility and reliability of the proposed method.

This paper proposed a framework that can be employed in trip-based models to account for time-of-day component. The framework utilized hybrid choice method to model departure-time choice and unobserved preferred departure-time simultaneously using observed explanatory variables, such as socio-economic and network-related variables. Scheduling preferences are important factors in time-of-day modeling, but preferred schedules data is usually unavailable. In the proposed hybrid choice framework, preferred schedules were the latent variables of hybrid choice model. Using latent variables in hybrid choice model not only improves efficiency, but also adds behavioral realism to the model. Furthermore, hybrid choice models can capture unobserved heterogeneity through the distribution of preferred departure time. The proposed methodology was applied to Maryland statewide trip-based model to obtain temporal distribution of the demand. The temporal distributions were compared for two scenarios (year 2007 and year 2030) to analyze demand shifts and peak spreading. The comparison between the predicted demand distributions showed how changes in demand distribution occur. It also showed the importance of time-of-day modeling for capturing temporal changes and having proper prediction for future demand.

This paper utilizes the well-known prospect theory to study how travel mode choice is affected by travel-time reliability. Prospect theory is developed to model decision making under risk. Travel-time reliability is related to uncertainties in travel-time, and its effect on travel behavior is a good candidate for applying prospect theory. The prospect theory is combined with a discrete choice model to build a mode choice framework. The choice model parameters, in addition to prospect theory parameters, are estimated using a combination of revealed preference household travel survey data and empirically observed reliability data for a real-world application. This application showcases how real-world observational data can be used in a prospect theory-based mode choice model. The proposed model’s estimated parameters are discussed and goodness-of-fit is compared with the utility-based mean-variance model. The paper focuses on mode choice, but its extension to other choice dimensions is discussed.

In this paper, we introduce a novel person re-identification method under the problem of path selection. Unlike other supervised person re-identification algorithms, our method can identify persons in the video without artificial markers. Our method divides each image sequences into several slices and selects the most distinguished ones, which can improve the performance. More crucially, this model is unsupervised and hence readily scalable to real-world large scale ReID settings and more suitable to previous path selection, i.e. no need of exhaustively collecting a large number of cross-view pairwise labels for each camera pair as required by most existing ReID models, which need supervised training. Experimental results show that our method can outperform other methods and achieve excellent results.

Promoting the government’s procurement of urban public transport services is helpful for upgrading the supply quality of public transport services and improving the efficiency of government’s financing. Based on the field research in more than ten cities in China and a thorough analysis on their policies, this paper summarizes three modes of public transport services procurement in China, i.e. selection of the best through bidding & quotation based on competitive cost, direct granting of operation rights & affirmation of costs by the government, direct granting of operation rights & extensive subsidy compensation. By analyzing Foshan and Suzhou’s exercises as typical cases, this research points out existing problems in the current unsound system and working mechanism, the insufficient development of the service procurement market and the inadequate financial guarantee. This article further puts forward countermeasures in specifying the governmental duties, improving the pattern of market operation and strengthening fund guarantee.

As the foundational, precursory and service industry of the national economy, speed up construction of the transportation industry credit system, which is an important way to perfect the socialist market economy system, strengthen and innovate the social governance. Based on the current situation of credit information management system of road transportation Industry and its existing problems, according to the planning objectives of credit informatization in Beijing, this paper tried to construct a credit information management and service system of the road transportation Industry in Beijing, on the basis of the existing credit evaluation system in several transportation departments. Furthermore, detailed introduction of functions and characteristics of the overall framework, data architecture, application architecture, business architecture, and security system.

Artificial intelligence methods are widely used in travel mode detection based on passively collected GPS data. This paper presents a jointly trained single-layer model and deep neural network for travel mode detection. Being “wide” and “deep” at the same time, this model combines the advantages of both types of models to be able to make sufficient generalizations using multi-layer deep learning and capture the exceptions using the wide single-layer model. The model is empirically tested on a GPS dataset collected in the Washington D.C. and Baltimore metropolitan regions. We also innovatively line-up the multimodal transportation network to the GPS trajectories in order to infer the closeness to the nearby rail (both underground and aboveground) and bus lines. To the best knowledge of the authors, this paper is the first to use land use data to infer underground metro modes. The empirical test showcases the superior goodness-of-fit and high precision and recall rates of the proposed wide and deep learning model compared with other benchmark machine learning models.

Each person has his/her own driving style, which can be used to identify the driver. Mining driving characteristics or habits from numerous travel data to identify drivers is the Driver Identification topic concerned in this paper. Real data sets are collected, including GPS, Acceleration and Angular Velocity. Some Deep Learning algorithms, such as AlexNet and Recurrent Neural Network, are novelly applied to solve the problem. Moreover, their performance are compared with Multilayer Perceptron and Random Forest in Machine Learning. In addition, drawing on the idea of speech recognition, Dynamic Time Warping is applied to find common track sections as verification set for Deep Learning identifications, which is a brand new crossover application. The overall designed identification process is tested by experiments, in which 5,500,000 driving records for each of 25 drivers are collected, and accuracy, precision, recall, verification rate and speed are consided as perfomances. The results show that Deep Learning are feasible for the problem and the two algorithms have different advantages. Especially, the testing accuracy of AlexNet is 46.15%, and the DTW verification accuracy can reach 67.77% . The methods proposed in the paper can be widely applied to similar problems in traffic, such as fleet drivers identification

Responsive ramp metering operates in adjusting on-ramp flow to enter the mainline freeway, dependent upon the detected congestion level of the mainline freeway. This technique theoretically sounds to alleviate recurring and non-recurring congestions, but greatly relies on prompt on-ramp and mainline freeway traffic detections of speed and flow variations. However, the traditional fixed detectors are a barrier to achieving the designated goal because of their incapability of quickly capturing the spatiotemporal patterns of congestion. With the emerging connected vehicle (CV) and/or autonomous vehicle (AV) technology (or CAV), it potentially provides a solution to the addressed problem trough make the CAV vehicles “floating sensors” that seamlessly cover the concerned highway over continuous time horizon. As a cost-effective, risk-free approach to quantitatively capture the CAV-affected driver behaviors, the developed method is to explore the synthesis on the operation of ramp meters at a real-world freeway facility with CAV support. The simulation results suggest positive benefits of improving freeway operation with ramp metering facility in CAV environment.

The Connected Vehicle (CV) safety systems improve the safety performance of a highway facility via assisting drivers in (re)acting properly under risky traffic conditions. This paper discloses and quantifies the cause-and-effect mechanism between such a behavior impact of CV and the traffic safety. A synthetic approach has been adopted to integrate the CV-affected behavior parameters into state-of-the-art traffic flow models. Resulting models are used to reproduce vehicle trajectories under the CV environment. Such trajectory data serve as the basis for computing the surrogate safety measurement, and thus describe the systematic safety performance of a highway facility. The effectiveness of the CV is examined in a case study for a freeway site in the greater Cincinnati area, Ohio. The study results find that the CV-affected perception-reaction time, desired headway and desired speed are responsible for the reduction of the traffic conflict frequency and decrease of the conflict intensity. The quantitative contribution of these behavior parameters has also been determined based on statistical analysis. The findings pave the technical foundation for successful deployment of the CV safety technologies into existing highway transportation systems.

This study investigated the long-term properties of recycled lime-flyash treated aggregates (RTA) as pavement base and subbase layers. First, the both the physical and chemical properties of the waste lime-flyash treated aggregate (WTA) materials were tested. Then, RTA were designed according to the specifications and further examined for its long-term durability properties including shrinkage, freeze-thaw, fatigue and permeability. The RTA investigated were actually cement-stabilized RTA, a stabilized blend of 70% WTA and 30% virgin aggregates for base layer, and a stabilized 100% WTA for subbase layer, respectively. Test results were as follows: 1) WTA particle surfaces were becoming rougher as the size of the particles decreased, which was closely related to the degree of shrinkages of the RTA. 2) Both the value of moisture shrinkage and the weight of water loss of RTA increased as the curing time of RTA increased, indicating the contribution of the weight loss to the shrinkage of the RTA. 3) The total thermal shrinkage increased as temperature decreased although its rate of increase decreased for every temperature increment. 4) Overall, the fatigue life of RTA was short, indicting its structure was closer to that of suspension dense type.

As technologies enabling connected and autonomous vehicles (CAV) are rapidly advancing, a 2 question remains not fully answered is that how peer effects influence users’ adoption of CAVs. 3 The answer is especially unclear for less-developed regions and residents not keen on new 4 technology adoption. This paper designed a stated preference (SP) survey and collected opinions 5 from five medium cities in China. Among all the determinants, peer effects showed significant 6 influence. 87% of respondents confirmed the effects of peer effects on their adoption of CAVs at 7 different levels. Ordered Probit model was then used to further analyze the impacts of peer effects. 8 Results indicated that those who are experienced in driving or know CAV tend to be less dependent 9 on peer effects in choosing CAV. This study offers valuable insights for understanding CAV 10 adoption, especially for less developed regions.

Recently, the emergence of deep learning has facilitated many research fields including 7 transportation, especially traffic pattern recognition and traffic forecasting. While many efforts 8 have been made in the exploration of new models for higher accuracy and larger scale, few 9 existing studies focus on learning higher-resolution traffic patterns. The most representative 10 example is the lack of research in multi-lane pattern mining and forecasting. To this end, this 11 paper proposes a deep learning framework that can learn multi-lane traffic patterns and forecast 12 lane-level short-term traffic conditions with high accuracy. Multi-lane traffic dynamics are 13 modeled as a multi-channel spatial-temporal image in which each channel corresponds to a 14 traffic lane. The constructed multi-channel image is then learned by a convolutional neural 15 network, which can capture key traffic patterns and forecast multi-lane traffic flow parameters. 16 One-year loop detector data for a freeway segment in Seattle are used for model validation. The 17 results and analyses demonstrate the promising performance of the proposed method.

Conventional travel behavior data collection methods such as the National Household Travel Survey (NHTS) have been the primary source of travel behavior information for transportation agencies. However, the relatively high cost of traditional travel surveys often prohibits frequent survey cycles. With decision makers increasingly requesting recent and up-to-date information on multimodal travel trends, establishing a sustainable and timely travel monitoring program based on available data sources from the public domain is in order. This paper develops a package of methods that are tailored to data of different quality for different modes in the public domain, and can collectively reveal month-to-month travel trends dynamically in a metropolitan area. The proposed methods will be demonstrated through case studies in three different metropolitan areas. A comparison with mode split trend based on household survey data collected in the same metropolitan area showed the effectiveness of the proposed method. Future studies will further address the data gap and reliability issue.