Investigating the Scalability in Population Synthesis: Comparative Approach
Ismaïl Saadi, Universite de LiegeShow Abstract
Hamed Eftekhar, University of Liege
Jacques Teller, University of Liege
Mario Cools, University of Liege
In this paper, we investigate the influence of scalability on the accuracy of different synthetic populations using both fitting and generation-based approaches. Most activity-based models need a base-year synthetic population where the agents are described by various attributes. However, when an important number of attributes need to be synthesized, the accuracy of the synthetic population may decrease due the mixed effects of scalability and dimensionality. Based on the workforce survey carried out in Belgium in 2013, we analyze the two population synthesis methods for different level of scalability, i.e. two to five attributes and different sample sizes, i.e. 10%, 25% and 50%. The results reveal that the simulation-based approach is more stable than Iterative Proportional Fitting (IPF) when the number of attributes increases. However, IPF is less sensitive to changes in sample size when compared to the simulation-based approach. From a global perspective, the accuracy of the synthetic populations provided by the simulation-based approach outperforms the ones from IPF for any sample size and any number of attributes. In addition, we demonstrate the importance of choosing the correct metric to validate the population. In this regard, we show that the trends in terms of RMSE/MAE are different from those of SRMSE.
Stable Matching and Price of Stability in Real-Time Ridesharing Systems
PENGYU YAN, University of Electronic and Science Technology of ChinaShow Abstract
Cynthia Chen, University of Washington
Zhiqin Luo, University of Electronic and Science Technology of China
Ride-sharing, a sustainable transportation mode, allows drivers to flexibly share idle seats in their vehicles with others. For many users, it is also considered an easier behavior change than switching from single driving to, for example, public transport and non-motorized modes. However, in the real world, rider-sharing is not a popular used alternative. Two primary barriers exist: one is that it is difficult to match between drivers and riders due to their constraints and preferences; and the other is that it remains unclear about the effects of the cost allocation methods between drivers and riders on the society welfare and the benefits of users.
In this paper, we formulate the problem as a stable marriage problem, which means that no user can get a better match than the current match, thus a stable matching solution. We show that cost allocation methods directly affect the benefits of users and ultimately determine the performance of the stable match solutions on society welfare. Then, we evaluate four common cost allocation methods in the literature and proves the lower bounds of the loss of the society welfare for each. In the numerical experiment, we evaluate these methods with multiple metrics, including the loss of the society welfare, ratio of successful matches and average benefit to the users. Our theoretical and experimental results are expected to provide practical guidelines in real-world ride-sharing applications.
Comparison of Artificial Neural Networks and Statistical Copula-Based Joint Models
Nima Golshani, Georgia Institute of Technology (Georgia Tech)Show Abstract
Ramin Shabanpour, University of Illinois, Chicago
Seyed Mahmoudifard, University of Illinois, Chicago
Sybil Derrible, University of Illinois, Chicago
Abolfazl (Kouros) Mohammadian, University of Illinois, Chicago
In the last decade, both machine-learning and statistical techniques have been expanding, offering new modeling approaches for travel demand analysis. While traditionally, the field of transportation demand analysis has been dominated by statistical models, few studies recently showed that machine learning techniques outperform the statistical models in some instances, notably in their forecasting potential, mainly because of their ability to recognize nonlinear patterns in datasets and to deal with noisy data. This study compares the prediction capabilities of a copula-based model as a statistical approach with an artificial neural network model as a popular machine learning technique for modeling two critical trip-related decisions of travel mode and departure time. The copula approach is employed because these two decisions are highly interrelated, and there are many common influencing factors that affect their outcomes. Furthermore, we are able to compare these models in the contexts of both discrete and continuous decision variables. To do so, we form a joint distribution for error terms of departure time as a continuous variable and travel mode as a discrete variable based on their marginal distributions. The departure time is estimated by a log-linear regression model and the travel mode is estimated by a multinomial logit model. The results illustrate that in addition to much easier and faster implementation process, the neural network model offers a slightly better performance for both target variables. However, these models are not able to easily assess the role of each explanatory variable in estimating the target variables and their black-box nature limits their capability for policy analysis, like examining the effects of variable elasticity.
A Joint Discrete-Continuous Model of Travel Mode and Departure Time Choices
Ramin Shabanpour, University of Illinois, ChicagoShow Abstract
Nima Golshani, Georgia Institute of Technology (Georgia Tech)
Sybil Derrible, University of Illinois, Chicago
Abolfazl (Kouros) Mohammadian, University of Illinois, Chicago
Mohammad Miralinaghi, Purdue University
This paper presents a cluster-based joint modeling approach to investigate heterogeneous travelers’ behavior toward trip mode and departure time choices by considering them as a joint decision. To do so, first, a two-step clustering algorithm is applied to classify travelers into six distinct clusters to account for the heterogeneity in their decision-making behavior. Then, a joint discrete-continuous model is proposed for each cluster, in which the travel mode and departure time are estimated by a multinomial logit and a log-linear regression model, respectively. These two models are jointly estimated using copula approach. To investigate the performance of the proposed approach, its results are compared with an aggregate joint model on all non-clustered observations to assess the potential benefits of population clustering. The goodness-of-fit measures and prediction accuracy results demonstrate that the proposed cluster-based joint model significantly outperforms the aggregate joint model. Furthermore, the variations in the estimated parameters of different clusters indicate significant behavioral differences across clusters. Hence, the proposed cluster-based joint model, while offering higher accuracy, possesses a significant potential for transportation policy-making because it has the capability to target different types of travelers based on their decision-making behavior.
Travel Demand Model Evaluation: Graph-Theoretic Approach
Meead Saberi, University of New South WalesShow Abstract
Taha Rashidi, University of New South Wales
Milad Ghasri, University of New South Wales
Kenneth Ewe, Monash University
Modeling demand for the transport system requires development of complicated mathematical structures reflecting the response of users to the capacity and service provided in the transport network including the infrastructure and modes of transport. Such complicated competitive environment in which travelers try to achieve better service at lower cost can be evaluated from numerous angles when modeled. The classical travel demands are typically evaluated based on the fit to the observed origin-destination table (mostly unavailable), or the likelihood value of models used in the process of generating the origin-destination table. These conventional evaluation techniques do not assess the goodness-of-fit of the model in a robust way. This paper presents a graph-theoretic methodology to evaluate travel demand models. An innovative travel demand model, developed using several advanced data mining techniques, is considered for the evaluation exercise. Statistical and mathematical properties of the modeled networks are compared against the observed networks over time. Networks are constructed based on the trip origin-destination matrices. The proposed evaluation approach focuses on the network structure attributes reflecting network properties from many angles as means for evaluating the goodness-of-fit of the models that are not usually captured by traditional evaluation and validation methods. Results demonstrate the complexity involved in development, evaluation, and validation of travel demand models stressing on the need for using methods like the proposed approach in this study.
Econometric Investigation of the Influence of Transit Passes on Transit Users’ Behavior in Toronto, Canada
Khandker Nurul Habib, University of TorontoShow Abstract
Md Sami Hasnine, Massachusetts Institute of Technology (MIT)
The paper presents the results of an empirical investigation of the influence of transit pass ownership on daily transit service usage behaviour under a flat fare system. The transit system of the City of Toronto (named as TTC) is investigated to evaluate the factors that influence the choice of owning a transit pass and at the same time whether or not a pass ownership influences different transit ridership behaviour. Data from a household travel survey are used to estimate a joint econometric model of the choice of owning a transit pass as a function of benefits drawn from daily transit usage (trip rate and distance travel) along with different socio-economic and land use variables. Results clearly show that a transit pass has a profound and segmenting influence on ridership behaviour in terms of the daily frequency of transit trips and the total distance travelled by transit. Non-pass owners seem to draw higher benefit from daily transit usage than the pass owners. However, a transit pass gives perceived benefit/utility over and above its daily use of the transit service. Results clearly indicate that a transit pass is more than just unlimited access to transit services for riders and hence should be treated as a mobility tool. Empirical results also show the influence of different variables that can be exploited to cater to various policies that may encourage higher pass ownership rate and thereby reduce demand for private automobiles.
Retrospective Evaluation of Traffic Forecasting Accuracy: Lessons Learned from Virginia
Salwa Anam, Florida International UniversityShow Abstract
John Miller, Virginia Department of Transportation
Jasmine Amanin, Virginia Department of Transportation
Understanding the accuracy of techniques for forecasting traffic volumes for a future year, such as extrapolation of previous years’ traffic volumes, use of regional travel demand models, and use of local trip generation rates, can aid analysts in considering the range of transportation investments for a given location. To determine this accuracy, forecasts from 39 Virginia studies (published from 1967-2010) were compared to observed volumes for the forecast year. Excluding statewide forecasts, the number of roadway segments in each study ranged from 1 to 240 links. For each segment, the difference between the forecast and observed volume divided by the observed volume gives a percent error, such that a segment with a perfect forecast has an error of 0%. For the 39 studies, the median absolute percent error ranged from 1% to 134% with an average value of 40%.
Slightly more than one-fourth of the variation in accuracy (29%) was explained by three factors: the forecast method, the forecast duration (number of years between the base and forecast years), and the number of economic recessions in the same interval ( p≤ 0.04). Interaction effects matter: the first two factors have significant and expected impacts on accuracy only if economic changes are explicitly considered. Finally, link-by-link error in a study has sufficient variation ( p = 0.02) such that if this variation is not controlled, causal factors are impossible to detect. Although no forecast is perfect, this study provides an indication of expected forecast error for future studies.
Comparison of Reliability Valuation Methods for the Ranking of Transport Projects
Helene Le Maitre, France Ministry For Infrastructure Transport and HousingShow Abstract
Charlotte Coupe, France Ministry For Infrastructure Transport and Housing
Recent transport policies focus on managing existing facilities rather than expanding the network and a large share of recent transport projects are expected to reduce travel time variability without necessarily reduce the mean travel times. Therefore the inclusion of travel time reliability in the appraisal of transport projects is necessary. The literature on the valuation of travel time reliability is rather large and quickly increasing and various methods and reliability indicators might be used, based on different models. In France, reliability valuation is not mandatory and various methods are used, which is challenging in terms of comparison of reliability benefits especially for multimodal projects.
Rather than focusing on the absolute valuation of reliability benefits, the aim of this paper is to provide an empirical comparison of reliability valuation methods by assessing their comparative ranking of projects and by analysing the correlation between the valuation of reliability benefits for various methods for frequently observed travel time distributions.
We find that although the shape of travel time distributions differs a lot, most valuation methods are consistent in terms of rankings. Also, in most cases, two or three indicators seem to be sufficient to value reliability since results are close to those obtained by valuing the whole distribution. This will enable us to focus the efforts in terms of reliability forecasts on a limited number of indicators rather than having to use simulation tools to forecast the whole distribution of travel times.
A Finite Mixture Modeling Approach to Examine New York City Bicycle Sharing System (CitiBike) Users’ Destination Preferences
Ahmadreza Faghih Imani, Imperial College LondonShow Abstract
Naveen Eluru, University of Central Florida
Given the recent growth of bicycle-sharing systems (BSS) around the world, it is of interest to BSS operators/analysts to identify contributing factors that influence individuals’ decision processes in adoption and usage of bicycle-sharing systems. The current study contributes to research on BSS by examining user behavior at a trip level. Specifically, we study the decision process involved in identifying destination locations after picking up the bicycle at a BSS station. In the traditional destination/location choice approaches, the model frameworks implicitly assume that the influence of exogenous factors on the destination preferences are constant across the entire population. We propose a Finite Mixture Multinomial Logit (FMMNL) model that accommodates such heterogeneity by probabilistically assigning trips to different segments and estimate segment-specific destination choice models for each segment. Unlike the traditional destination choice based MNL model, in an FMMNL model, we can consider the effect of fixed attributes across destinations such as users or origins attributes in the decision process. Using data from New York City bicycle-sharing system (CitiBike) for 2014, we develop separate models for members and non-members. We validate our models using hold-out samples and compare our proposed FMMNL model results with the traditional MNL model results. The proposed FMMNL model provides better results in terms of goodness of fit measures, explanatory power and prediction performance.
Refueling Station Location Problem with Traffic Deviation Considering Route Choice and Demand Uncertainty
Mohammad Miralinaghi, Purdue UniversityShow Abstract
Yingyan Lou, Arizona State University
Burcu B. Keskin, University of Alabama
Yu-Ting Hsu, National Taiwan University
Ramin Shabanpour, University of Illinois, Chicago
Construction of refueling stations in a traffic network is a major step toward the promotion of hydrogen fuel vehicles in the metropolitan areas. This study provides two modeling approaches with considering the refueling demand uncertainty and the effect of travelers’ deviation to refuel in the network. First, considering the refueling demand uncertainty of the hydrogen fuel vehicles market, we propose a discrete and robust optimization model, in which refueling demand is formulated as an uncertainty set during planning horizon. A cutting plane algorithm is then adopted to solve this robust centralized planning model. Numerical results demonstrate that the robust model can lead to more reliable design in comparison to the nominal plan. Second, a link-based bi-level program is proposed in which the lower level problem describes the user equilibrium traffic condition characterized by the locations of refueling stations. Numerical examples indicate that the refueling station location pattern can change completely with the consideration of route choice of users.
On the Heterogeneity and Substitution Patterns in Mobility Tool Ownership Choices of Post-secondary Students: The Case of Toronto
Khandker Nurul Habib, University of TorontoShow Abstract
Adam Weiss, University of Calgary Schulich School of Engineering
Md Sami Hasnine, Massachusetts Institute of Technology (MIT)
The paper presents an investigation of the choices of mobility tool ownership of post-secondary students in Toronto. Data came from a 2015 survey of post-secondary students across four universities in Toronto. The choices of owning a basic mobility tool (driver’s license, car, transit pass and bicycle) or combinations of basic tools (composite tools) are investigated through estimation of cross-nested generalized extreme value (GEV) models. The paper proposes a parsimonious GEV model that drastically reduces the total number of parameters that are needed to be estimated, while accommodating the full range of substitution patterns among the choice alternatives. The model clearly shows the systematic interaction of basic mobility tool ownership utility is more prevalent than the random correlation that a GEV model can capture. Students’ personal and household related attributes influence the choice of owning combinations of mobility tools and influence multimodality. It is also found that older and male students are more multimodal than younger and female students. High car ownership levels play a pivotal role against the choice of owning transit passes.
Two-Stage Bicycle Origin Destination Demand Matrix Estimation
Seungkyu Ryu, Ajou UniversityShow Abstract
Anthony Chen, Hong Kong Polytechnic Universtiy
Jacqueline Su, University of California, Los Angeles
Keechoo Choi, Metropolitan Transport Commission of Korean Government
Urban communities are witnessing a transformation in mode share due to the rising popularity of cycling in recent years. As more people choose to travel by bicycle, transportation planners are recognizing the need to rethink the way they evaluate and plan transport facilities to meet local mobility needs. A modal shift towards bicycles motivates a change in transportation planning to accommodate more bicycles. However, the current methods to estimate bicycle volumes on a transportation network are limited. The purpose of this research is to address those limitations through the development of a two-stage bicycle origin-demand (O-D) matrix estimation process that would provide a different perspective on bicycle modeling. The bicycle O-D demand matrix estimation process undergoes two stages; the first uses survey data in a doubly constrained gravity model to construct an initial O-D matrix, and the second refines that initial matrix using a Path Flow Estimator (PFE) to produce the finalized bicycle O-D demand matrix. After a detailed description of the methodology, the paper concludes with a case study of the City of Winnipeg, Canada to demonstrate the applicability of the process.
Considering Dynamic Knowledge Updating in Bounded Rationality-Based Route Choice Modeling
Xiaowei Jiang, Southeast UniversityShow Abstract
Muqing Du, Hohai University
Wei Deng, Southeast University
To study the effect of en-route information on driver’s route choice behavior, a route choice modeling approach based on bounded rationality is proposed, which takes the driver’s dynamic knowledge updating process into consideration. A Bayesian network is developed to describe the en-route travel time updating process. Within the framework of the cumulative prospect theory, a choice model is conducted to analyze the driver’s route choice behavior at each decision node. A numerical experiment is carried out to illustrate the development of the dynamic route choice modeling approach. The results show that the dynamic updating process of travel time is affected not only by the geometry properties of route, but also by the traffic condition on route. The cumulative prospect theory without considering dynamic knowledge updating is also employed as a reference in the experiment. From the comparison, the results of the route choice model considering dynamic knowledge updating are closer to the results of stated preference survey, which indicates that the proposed approach could provide a more accurate description to driver’s route choice behavior under the conditions of uncertainty.
A Multi-criteria Bus Demand Assignment based on Minimizing Total System Costs
Ali Gholami, University of Nevada, RenoShow Abstract
Maysam Ziaee, Mashhad Traffic and Transportation Organization
Zong Tian, University of Nevada, Reno
The number of bus transit passengers that are attracted to a route is related to different levels of transit design including line configuration and frequency. One of the important factors of lines are terminals. Location of terminals are also related to line routing that is another level of transit system design. The output terminal locations, line routes, and frequencies affect different layers of related parties, in another word users, operator and society. Therefore bus transit demand assignment (BTDA) can be considered as a multilevel and multilayer problem. In this paper the link pheromone updating concept of Ant Colony Optimization (ACO) was used also for nodes to weight appropriate nodes during ACO process. While in traditional ACO links have pheromone, in the proposed method nodes also receive pheromone based on total system cost. The total system cost includes users, operator and society costs. This node pheromones gradually increase at good node candidates and decrease at poor locations. Since at each iteration of ACO in addition to terminal selection, routing and frequency calculation also is performed, the effect of these two levels of transit system design is considered at the same time on the assignment process. Thus with a small modification in the ACO process the model performs the BTDA as a multilevel-multilayer problem without increasing the complexity of the design.
Keywords: bus transit demand assignment, ant colony optimization, bus network, metaheuristic.
Activity Rescheduling within a Multi Agent Transport Simulation Framework (MATSim)
Milos Balac, ETHZ - Swiss Federal Institute of TechnologyShow Abstract
Kay Axhausen, Eidgenossische Technische Hochschule Zurich
People's desire or the need to perform certain activities during the day drives their activity-scheduling decisions. However, these decisions are dependent on the state of the transportation system, its supply and demand. The need for the tools able to deal with the kind of adaptations to the daily plans that come with these decisions, is ever growing. The introduction of new modes and services and the fast approaching era of autonomous vehicles, among other things, has increased the need for suitable tools to look at the induced/suppressed demand effects on the activity schedules.
The work in this paper presents a methodology for the adaptation of the activity schedules inside of the multi-agent transport simulation (MATSim), based on the changes of supply in the system. The first results show that the proposed methods are able to adapt people's schedules when they are faced with shorter or longer travel times and this with only 10\% in the computation time. However, further development is needed in order to more realistically represent human behavior, which is discussed at the end of this paper.
A Random Utility Based Estimation Framework for the Household Activity Pattern Problem
Zhiheng Xu, University at Buffalo, The State University of New YorkShow Abstract
Jee Eun Kang, University at Buffalo, The State University of New York
Roger Chen, University of Hawai'i at Manoa
This paper develops a random utility based estimation framework for the Household Activity Pattern Problem (HAPP). Based on the realization that outputs of complex activity-travel decisions form a continuous pattern in space-time dimension, the estimation framework is treated as a pattern selection problem. In particular, we define a variant of HAPP that has capabilities of forecasting activity selection and durations in addition to activity sequencing. The framework is comprised of three steps, (i) choice set generation, (ii) choice set individualization and (iii) multinomial logit estimation. The estimation results show that utilities for work, shopping and disuilities for travel time, time outside home, and average tour delay are found to be significant in activity-travel decision making.
A Vehicle Ownership Model for Conventional Four-Step Travel Models
Guang Tian, University of New OrleansShow Abstract
Reid Ewing, University of Utah
Jon Larsen, Wasatch Front Regional Council
Vehicle ownership modeling plays an important role in travel demand analysis as a pre-step in the conventional “four-step” travel demand modeling and forecasting process. However, the practice process fails to account for the full effects of the built environment on vehicle ownership. This study explores a short-term, low-cost model enhancement promoted by leading academics and consultants in the travel modeling field. It is the most efficient way to account for land use effects on household trip rates and mode choices, important effects ignored in practice models. In the new model, built environmental variables affect vehicle ownership, which in turn affect various aspects of household travel. Using household travel survey data from 23 diverse regions of United States, we estimate a multilevel Poisson model of vehicle ownership. The model has the expected signs on its coefficients and respectable explanatory power. Vehicle ownership decreases activity density, land use diversity, percentage of 4-way intersections, transit stop density, and with employment accessibility by transit after controlling for sociodemographic variables. The model captures the phenomenon of “car shedding” as development patterns become more compact. We also compare the multinomial logit model to the Poisson model, showing how the Poisson model performs better than the multinomial logit model in terms of model fit and prediction accuracy.
Robust Evaluation for Transportation Network Capacity under Demand Uncertainty
Muqing Du, Hohai UniversityShow Abstract
Xiaowei Jiang, Southeast University
Lin Cheng, Southeast University
Changjiang Zheng, Hohai University
When evaluating the capacity of a transportation system, the prescribed origin and destination (O-D) matrix for existing travel demand has been noticed to have a significant effect on the results of network capacity models. However, the exact data of the existing O-D demand are usually hard to be obtained in practice. Considering the fluctuation of the real travel demand in transportation networks, the existing travel demand is represented as uncertain parameters which are defined within a bounded set. Thus, a robust reserve network capacity (RRNC) model using min-max optimization is formulated based on the demand uncertainty. An effective heuristic approach utilizing cutting plane method and sensitivity analysis is proposed for the solution of the RRNC problem. Computational experiments and simulations are implemented to demonstrate the validity and performance of the proposed robust model.
Biased Standard Errors in Transport Model Calibrations Due to Heteroscedasticity Arising from Linear Data Projection
Wai Wong, University of Michigan, Ann ArborShow Abstract
S.C. Wong, University of Hong Kong
Linear data projection is a prevalently adopted data inference method for traffic data estimation. A recent study has proven that model calibration based on linearly projected data using the scaling factor mean that does not account for its variability may lead to systematically biased parameters. Adjustment factors for reducing such biases have been proposed for a generalized multivariate polynomial model. Nevertheless, the effects of linear data projection on the standard errors of the adjusted parameters have not been explored. Without appropriate statistics testing the significance of the parameters, the application validity of the calibrated model remains unknown and dubious. This study reveals that heteroscedasticity is inherently introduced when a data projection scheme is leveraged, thus parameter standard errors estimated by linearly projected data are definitely biased. A generic analytical distribution-free (ADF) method and an equivalent scaling factor (ESF) method are proposed to accurately estimate the parameter standard errors for a generalized multivariate polynomial model. The ESF method provides an alternative solution path for unbiased parameter estimations by transforming a transport model into a linear function of the scaling factor with an assumed scaling factor distribution. Simulations demonstrate that the ESF method outperforms the ADF method at high model nonlinearity. Six Macroscopic Bureau of Public Roads functions are calibrated using real-world global positioning system data obtained from Hong Kong to illustrate the applicability of the ESF method for the parameter standard error estimations.
A Fleet Sizing Algorithm for Autonomous Car Sharing
Renos Karamanis, Imperial College LondonShow Abstract
Ali Niknejad, Imperial College London
Panagiotis Angeloudis, Imperial College London
In the past ten years, the popularity of car sharing schemes has been growing rapidly. Depending on service policies, a customer might be required to return the vehicle to the same location or a different one. The latter type of service is termed as one-way car sharing since the client is not required to make a return trip to leave the vehicle. It is expected that the car sharing companies will start offering autonomous vehicles in their fleet in the future, in line with the growth of investment in autonomous cars in the automotive industry. Fleet management of autonomous vehicles is, therefore, an area which could be explored to devise and offer solutions to companies regarding autonomous vehicle applications. This paper explores the implementation of an approach with a mixed-integer programming (MIP) algorithm which is inspired by one-way car sharing schemes, to offer assistance on strategic decisions such as fleet size, depot location, and number, as well as depot capacity for shared autonomous vehicle (SAV) systems within cities. The proposed model uses stochastic demand based on expected origin and destination (OD) matrices for a modified version of the Sioux Falls network, incorporates relocations to serve demand and is subject to charging and maintenance constraints. The results show that the proposed model could potentially be used in real and larger networks, with expected demand of trips to be used as a relocation strategy.
Incorporating Land Use in Synthetic Population Generation Methods and Transfer of Behavioral Data
Konstadinos Goulias, University of California, Santa BarbaraShow Abstract
Elizabeth McBride, University of California Santa Barbara
Jae Hyun Lee, Korea Research Institute for Human Settlements
Adam Davis, University of California, Davis
In this paper a new method of population synthesis that includes land use information is described. The method is based on first identifying suitable land use summaries to build a spatial taxonomy at any spatial scale. This same taxonomy is used to classify household travel survey records (persons and households) and in parallel geographic subdivisions for the entire State of California. This land use information is the added dimension in the population synthesis methods for travel demand analysis. Synthetic population generation proceeds by expanding (recreating) the records of the households responding to the survey and the entire array of travel behavior data reproduced for the synthetic population. Selection of the variables to use in the synthetic population is based on first testing their significance in simplified specification in models of travel behavior that include land use as an explanatory variable and account for the shape of behavioral data (e.g., observations with no travel). The paper shows differences between synthetic populations with and without the use of land use data demonstrating the behavioral realism added by this approach.
Testing Spatial Transferability of Activity-Based Travel Forecasting Models
John Bowman, Bowman Research and ConsultingShow Abstract
Mark Bradley, RSG Inc
This paper reports results from the second phase of a two-phase FHWA-sponsored project to empirically test and demonstrate the transferability of activity-based (AB) model systems between regions. Using data obtained through the 2008-2009 National Household Travel Survey “add-on” program, the principal investigators estimated activity-based models simultaneously for thirteen metropolitan regions in seven U.S. states. Statistical tests were used to test transferability, including tests of regional differences in the model coefficients, likelihood ratio tests of model equivalence, and Transferability Indexes that measure the degree of model differences. In addition, differences in prediction sensitivity between locally estimated and transferred models was tested.
The project overall found evidence in favor of transferability, and that parameters associated with land use, logsum accessibilities and travel time and cost cause the biggest problems with transferability. Finally, and the primary focus of this paper, it found that transferring within a state or between regions with similar urban density improves transferability. This paper presents the data, models and testing methods used in the project, followed by details of all tests and results related to the improved transferability associated with model transfers from regions within the same state or with similar urban density. The conclusion is that agencies considering transfer of an AB model from another region would do well to find one within the same state or with similar urban density that has a model well-supported by a large household travel survey data set.
A Sensitivity-Based Linear Approximation Method to Estimate Time-Dependent Origin-Destination Demand in Congested Networks
Sajjad Shafiei, DATA61|CSIROShow Abstract
Meead Saberi, University of New South Wales
Ali Zockaie, Michigan State University
Majid Sarvi, University of Melbourne
This paper presents a bi-level optimization problem to estimate offline time-dependent origin-destination (TDOD) demand based on link flows and historical OD matrices. Conventionally, OD flows are linearly mapped to link flows using the assignment matrix proportions obtained from the dynamic traffic assignment (DTA), which is typically formulated as the lower-level problem. However, the linear relationship may be violated when congestion builds up in the network, resulting in a nonlinear relationship between OD flows and link flows. The nonlinearity leads to convergence of the solution to the points that may be far from the global optimum. To overcome this limitation, we propose a practical solution to capture the nonlinearity feature in real-size congested networks. The iterative solution algorithm is applied to a small study network and a real urban network. Results demonstrate the applicability and efficiency of the proposed method in congested networks.
Latest Urban Rail Demand Forecast Model System in the Tokyo Metropolitan Area
Hironori Kato, University of TokyoShow Abstract
Daisuke Fukuda, Tokyo Institute of Technology
Yoshihisa Yamashita, Creative Research and Planning
Seiji Iwakura, Shibaura Institute of Technology
Tetsuo Yai, Tokyo Institute of Technology
This paper reports on an urban rail travel demand forecast model system, which technically supported the formulation of the Tokyo Urban Rail Development Master-plan 2016. The model system was included in the forthcoming 15-year urban rail investment strategy for Tokyo. The model system was utilized to quantitatively assess urban rail projects including 24 new rail development projects, which had been proposed in response to expected changes in socio-demographic patterns, land-use market, and the government’s latest transportation policy goals. The system covers the entire urban rail network within the Tokyo Metropolitan Area of approximately 50-km radius with a population of over 34 million. The system must handle over 80 million trips per day. Three demand models are used to predict daily rail passenger link flows: the urban rail demand model, the airport rail access demand model, and the high-speed-rail rail access demand model. These practical models have unique characteristics such as incorporating differences in behavior between aged and non-aged travelers, reflecting expected influences of urban redevelopment on trip generation and distribution, highlighting urban rail access to airports or high-speed-rail stations, examining impacts of in-vehicle crowding on rail route choice, and deploying urban rail-station access/egress mode choice models for rail route choice. It is concluded that the model system would be well calibrated with observed data for reproducing travel patterns, identifying potential problems, assessing proposed projects, presenting results with high accuracy, and assisting decision-making of urban rail planners.
Misclassification in Travel Surveys and Implications to Choice Modeling: Application to Household Auto Ownership Decisions
Rajesh Paleti, Pennsylvania State UniversityShow Abstract
Lacramioara Balan, Old Dominion University
Travel surveys that elicit responses to questions regarding daily activity and travel choices form the basis for most of the transportation planning and policy analysis. The response variables collected in these surveys are prone to errors leading to mismeasurement or misclassification. Standard modeling methods that ignore these errors while modeling travel choices can lead to biased parameter estimates. In this study, methods available in the econometrics literature were used to quantify and assess the impact of misclassification errors in auto ownership choice data. The results uncovered significant misclassification rates ranging from 1% to 40% for different auto ownership alternatives. Also, the results from latent class models provide evidence for variation in misclassification probabilities across different population segments. Models that ignore misclassification were not only found to have lower statistical fit but also significantly different elasticity effects, particularly for choice alternatives with high misclassification probabilities. The methods developed in this study can be extended to analyze misclassification in several response variables (e.g., mode choice, activity purpose) that constitute the core of advanced travel demand models including tour and activity-based models.
On the Variance of Recurrent Traffic Flow for Statistical Traffic Assignment
Wei Ma, Carnegie Mellon UniversityShow Abstract
Sean Qian, Carnegie Mellon University
This paper generalizes and extends classical traffic assignment models as well as existing statistical assignment models to characterize the statistical features of varying Origin-Destination (O-D) demands, link/path flow and link/path costs. The generalized statistical traffic assignment (GESTA) model has clear multi-level variance structure. Flow variance is analytically decomposed into three sources, O-D demands, route choices and measurement errors. Consequently, optimal decisions on roadway design, maintenance, operations and planning can be made using estimated link/path flow distributions and their variance. The statistical equilibrium in GESTA is mathematically defined. Its multi-level statistical structure is consistent and well fits large-scale data learning techniques. The embedded route choice model is consistent with the settings of O-D demands considering link costs that vary from day to day. We propose a Method of Successive Averages (MSA) based solution algorithm to solve for GESTA. Its convergence and computational complexity are analyzed. Two example networks including a large-scale network are solved to provide insights for decision making and demonstrate computational efficiency.
Who Will Buy Alternative Fueled or Automatic Vehicles: A Modular, Behavioral Modelling Approach
Ioannis Tsouros, University of the AegeanShow Abstract
Amalia Polydoropoulou, University of the Aegean
Future car purchase affects a wide variety of areas ranging from CO 2 emissions to urban quality of life. For this reason, models and methods predicting car purchase are valuable to policy makers. This paper examines the choice of future cars purchase taking into consideration different levels of car attributes, as well as personality traits. The results should enable decision makers (including policy makers and manufacturers/distributors) to focus on certain market segmentation for the promotion of alternative fuel and automated vehicles. The paper proposes a hybrid choice model, with latent variables capturing the pro-environmental, exuberant and tech-friendly attitudes of individuals. The questionnaire presented to the respondents had the form of a menu. Participants could choose from five different types of vehicle characteristics (engine size, type of car, fuel type, car edition and level of automation) to construct their ideal vehicle. Results indicate that there is a negative correlation between having symbolic, exuberant attitudes towards automobiles, viewing the cars as symbols, and the willingness to purchase a hybrid or electric vehicle.
Explicit or Implicit Accommodation of Residential Self-Selection in Modeling Vehicle Ownership: The Right Approach
Sabreena Anowar, University of Missouri, ColumbiaShow Abstract
Naveen Eluru, University of Central Florida
In this research, we contrast different modeling frameworks that offer alternative ways of capturing observed/unobserved heterogeneity. The model systems compared are: ordered logit, exogenous segmentation residential location cluster based ordered logit model, mixed ordered logit, latent segmentation based ordered logit model and a copula based joint model. While the comparison across single dependent variable models is straight forward, the comparison with the copula based model requires post-processing to generate marginal for the choice of interest. A host of comparison metrics are computed to evaluate the performance of these alternative models. The performance of the alternative frameworks is examined in the context of model estimation and validation (at the aggregate and disaggregate level). The comparison exercise is conducted in the vehicle ownership context using O-D survey data of Greater Montreal Area (GMA), Canada. In all cases, the superior performance of the ordered part of the joint copula based model indicates that employing information from an additional dependent variable (such as residential location choice in our case) allow us to better understand and predict the main dimension of interest (vehicle ownership).
High-Granularity Dynamic Traffic Flow Prediction Model Based on Artificial Neural Network
Zhihong Yao, Southwest Jiaotong UniversityShow Abstract
Peng Han, Southwest Jiaotong University
Bin Zhao, Southwest Jiaotong University
Yangsheng Jiang, Southwest Jiaotong University
Bo Liu, Southwest Jiaotong University
Mengqiu Du, Southwest Jiaotong University
The traditional platoon dispersion model is based on the hypothesis of probability distribution, and the time granularity of the existing traffic flow prediction model is too big to be applied to the adaptive signal timing optimization. Based on the view of the platoon dispersion model, the relationship between vehicle arrival at downstream intersection and vehicle departure from the upstream intersection was analyzed. Then, the high-granularity dynamic traffic flow prediction model based on artificial neural network was proposed. The departure flow at the upstream was taking as the input and the arrival flow at downstream intersection was taking as the output in this model. Finally, the parameters of the proposed model were calibrated by the field survey data, and this model was implemented to predict the arrival flow of the downstream intersection. The result shows that the proposed model can better reflect the fluctuant characteristics of traffic flow and the prediction error is reduced by 11.1%, compared with Robertson’s model. Thus, the proposed model can be applied for signal timing optimization.
Using Work Location and Industry Classification Information in the Weighting of Household Surveys using Open Source Frameworks
Anurag Komanduri, Cambridge Systematics, Inc.Show Abstract
Karthik Konduri, Amazon.com
Activity-based models (ABM) that simulate travel are becoming commonplace owing to their value in supporting policy-driven scenario analyses. Due to the complex nature of these ABMs, only a limited number of data sources provide the detail necessary for the estimation and calibration of these models. Among them, household travel surveys are becoming increasingly central to the development and calibration of ABMs.
ABMs mimic rational decision-making and use a hierarchical decision-making framework that prioritizes mandatory travel and activities first which results in a constrained time interval for other non-mandatory activities. Therefore, it is critical that expanded household surveys represent mandatory travel and activity characteristics accurately. Traditionally, household surveys have been expanded to match household-level demographics such as household size, and number of vehicles. More recent weighting frameworks have included person-level demographics such as worker status, and age. However, there is limited research (if any) looking into the role of employment-level attributes (e.g. journey to work flow data) within the weighting procedure.
The authors build on the body of work in the areas of survey expansion, and synthetic population generation to incorporate two employment-level variables: industry-level employment totals, and home-to-work flow patterns in the expansion process. Further, the employment-level variables are matched at different spatial resolutions, namely, (a) industry-level employment totals at the region-wide level, and (b) home-to-work patterns at a sub-regional level to improve travel duration distributions. The resulting weights are then contrasted against results from traditional expansion methods by summarizing a variety of variables suitable for model validation.
Key Words: activity-based models, synthetic population, household survey expansion, iterative proportional fitting, model validation statistics
A Spatial Linear Programming Method for Estimating Zonal Space Use Coefficients and its Application for Integrated Land Use Transport Modeling
Bilin Yu, Wuhan UniversityShow Abstract
Ming Zhong, Wuhan University
John Hunt, University of Calgary
Huini Wang, Wuhan University
In order to develop an ILUTM, estimation of space use coefficients by space type by zone and theoretical space-use-rent curves are required. Existing studies mostly use census data and borrowed space use coefficients (SUCs) for synthesizing base-year floorspace and developing space-use-rent curves and their accuracy is largely unknown. In this study, a spatial linear programming method is proposed to estimate SUCs by category by zone, by assuming the zonal population, employment and floorspace total is available (which can easily collected/extracted from the census and remote sensing data). Study results show that the estimation method proposed can provide better SUCs and space-use-rent curves than the synthesis methods used before. Rather than applying a set of fixed SUCs across study area, the proposed method can clearly capture the spatial dependency between the rent and space consumption pattern and better space-use-rent curves is developed. It is found that, under the most disaggregate case in which 468 SUCs (9 SUCs for each of 52 zone clusters) are simultaneously optimized, for zonal floorspace totals, most absolute percent errors (APEs) for estimated zonal floorspace totals are below 17.1%. The sum of square errors (SSEs) between the synthesized space use ratios and the theoretical space-use rent curves are also significantly reduced to about 1/40 or even lower of the traditional method.
Empirical Demonstration of Traffic Flow Estimates from Repeated Passes of a Mobile Sensing Platform
Mark McCord, Ohio State UniversityShow Abstract
Rabi Mishalani, Ohio State University
Benjamin Coifman, Ohio State University
An empirical study is presented in which traffic flows are estimated from data collected from a sensor-equipped probe vehicle that repeatedly traversed a set of roadway segments. The probe vehicle serves as a "platform" upon which sensors are mounted to monitor ambient vehicles. A variant of the moving observer method is developed to estimate traffic flows for each pass of the platform on a segment and applied to vehicle counts automatically obtained from the mounted sensors. Comparisons with traffic flow estimates obtained from road tube data demonstrates that the individual pass flow estimates behave like estimates derived from short-term traditional estimates. The temporal patterns in the flow rates correspond well to known traffic flow patterns. The empirical results indicate the promise of using a vehicle, such as a transit bus, that repeatedly traverses roadway segments on a regular basis as a sensing platform for valid traffic flow estimation across a spatially extensive urban network.
Discrete-time Autoregressive Continuous Logit: Formulation and Application to Time of day choice modeling
Carlos Carrion, University of Maryland, College ParkShow Abstract
Sepehr Ghader, University of Maryland, College Park
Lei Zhang, University of Maryland, College Park
We formulate the Discrete-time Autoregressive Continuous Logit model as a novel continuous choice model capable of representing correlations across alternatives in the continuous spectrum, explicitly. These correlations are modeled for a priori specified set of discrete periods. This model is formulated by combining a continuous logit and a linear stochastic difference equation. In addition, we study analytically the properties of this model and also numerically with a Monte Carlo experiment. Furthermore, we discuss briefly the possible application of this model for time of day choice modeling.
A Novel Model Updating Method: Updating Function Model with Gross Domestic Product Per Capita
Nobuhiro Sanko, Kobe UniversityShow Abstract
When data are available from two points in time: older data with a larger number of observations and more recent data with a smaller number of observations, then model updating is utilised to use the merits of the both datasets. However, the author’s previous study questioned the merits of conventional model updating techniques: transfer scaling, joint context estimation, Bayesian updating, and combined transfer estimation. Although these model updating methods utilise datasets from two points in time, models using only the more recent data often produced statistically significantly better forecasts than the models updated. The present study proposes a novel updating method (called ‘updating function model’), where parameters are assumed to follow functions of gross domestic product per capita. The method was originally proposed by the author, but the present study aims to demonstrate that it is a novel updating method. While conventional model updating applies to a case where the number of observations from the more recent time point is smaller than that from the older time point, the present study also considered a case where the number of observations from the more recent time point is larger than that from the older time point. In both of the two cases, the present study demonstrated that the updating function model often produced statistically significantly better forecasts than models using only the more recent data. The study concluded that the updating function model is a useful model updating technique and extended the applicability of the model updating to the case where the number of observations from the more recent time point is larger than that from the older time point.
Modelling User Adaptation to a Campus Bicycle Share System
Cen Zhang, Kyoto UniversityShow Abstract
Jan-Dirk Schmoecker, Kyoto University
Understanding the changes in travel behavior over time in a transport system is essential to evaluate the performance and forecast the travel demand. However, long term travel behavior is difficult to observe and explain and even more difficult to forecast. This paper proposes an approach based on stochastic state equations to describe the gradual change of behavior over time by using panel data. Transition functions determine the likely change in behavior from one time period to another. To overcome the dynamic population of samples problem and explain special state transition phenomenon, we introduce “life cycle”, “potential demand” and “willingness to use” into our models. Then we discuss time-homogeneity issues and possibilities to identify states and calibrate the transition function. The model is applied to panel data from Kyoto University’s bicycle share system. The findings help us understanding the adaption, “usage recover” and drop out behavior. Also the errors between actual and estimated values are analyzed to evaluate the model. A comparison of various models are made to show the reliability of our models. Overall, the results offer promising insights to a wide range of applications.
Mobility as a Language: Predicting Individual Mobility in Public Transportation using N-Gram Models
Zhan Zhao, ViaShow Abstract
Haris Koutsopoulos, Northeastern University
Jinhua Zhao, Massachusetts Institute of Technology (MIT)
For public transportation agencies, the ability to provide personalized and dynamic passenger information is crucial for improving the efficiency of demand management and enhancing customer experience. This requires understanding and especially predicting individual travel behavior in the public transportation system, which is challenging because of the heterogeneity among passengers and the variability of their behaviors. This paper presents, to the best of our knowledge, the first attempt to predict individual spatiotemporal behavior of public transportation passengers using smartcard data. In this study, each trip is coded as a combination of trip start time, an entry station and an exit station. A passenger’s daily mobility is represented as a chain of travel decisions. We propose a new modeling framework, inspired by Bayesian n-gram models used in natural language processing, to estimate the probability distribution of the next decision in the sequence. Empirical analysis using Oyster card data from London shows promising results. It is found that the exact time of travel is most challenging to predict, but the difference between the predicted time and the true value is usually small. Model performance varies greatly across individuals for the prediction of entry and particularly exit stations. Overall, our proposed model shows significant improvement over the regular n-gram models, or Markov chain-based models in general. The improvement is even larger for weekend trips when travel behavior is flexible, irregular, and considerably less predictable.
An Early Look into Spectral Techniques for Travel Demand Modeling
Joseph Flood, Delaware Valley Regional Planning CommissionShow Abstract
Catherine Kostyn, Indianapolis Metropolitan Planning Organization
Suzanne Childress, Puget Sound Regional Council (PSRC)
Andrew Swenson, Indianapolis Metropolitan Planning Organization
Two major approaches to traffic assignment exist, static and dynamic traffic assignment. A major issue with static assignment is that the day is divided into multi-hour time periods within which network attributes are assumed to be constant. Dynamic traffic assignment allows for better temporal resolution, but there is an increased data storage requirement with the higher resolution. Fourier series are an efficient way to represent continuously-changing periodic data, such as transportation network attributes. Further, it is not a drastic departure from current modeling practice to obtain the coefficients of such series from model results. The theory behind such an approach will be discussed, as well as the results of an actual run of the Indianapolis regional travel demand model that was reworked slightly to produce Fourier coefficient outputs. Results suggest potential in future developments of such techniques.
Large-Scale Application of a Combined Destination and Mode Choice Model Estimated with Mixed Stated and Revealed Preference Data
Michael Heilig, Karlsruhe Institute of Technology (KIT)Show Abstract
Nicolai Mallig, Karlsruhe Institute of Technology (KIT)
Tim Hilgert, Karlsruhe Institute of Technology
Martin Kagerbauer, Karlsruhe Institute of Technology (KIT)
Peter Vortisch, Karlsruhe Institute of Technology
The diffusion of new modes of transportation such as carsharing and electric vehicles makes it necessary to consider them also in travel demand modeling. However, there are mainly two challenges for transportation modelers. First, the low share of usage still leads to lack of reliable revealed preference data for model estimation. Stated preference survey data are though a promising and well established approach to close this gap. Second, the state-of-the-art model approaches are sometimes stretched to their limits in large-scale applications.
In this paper, we present the integration of new transportation modes, namely electric bicylces, bikesharing and carsharing in the agent-based travel demand simulation framework mobiTopp with 2.5 million agents. Therefore, a combined destination and mode choice model has been developed. We describe the estimation process using mixed revealed and stated preference data and discuss important topics as well as the challenges of the modeling process, mainly caused by the large-scale application. We finally show that the results of the new combined model outperform those of the former sequential model.
Understanding Early Adopters of Carsharing and Expansion to the Whole Population
Michiko Namazu, Uber Technologies, Inc.Show Abstract
Don MacKenzie, University of Washington
Hisham Zerriffi, University of British Columbia
Hadi Dowlatabadi, University of British Columbia
This study addresses two questions: are early adopters of carsharing (CS) services representative of the general public and can outcomes associated with early adopters be projected for later adopters? Our study is based on a 2013 survey of residents in 110 apartment buildings in Metro Vancouver, Canada. Over two thousand responses were analyzed for possible differentiating factors for early adopters at the household level. We find that early adopters (about 25% of respondents) have more wage-earners per household, do not live with older family members and are more likely to be renters. Among non-CS membership holders (about 75% of respondents), roughly one-third stated they would never choose CS. The rest expressed interest in joining if CS accessibility was improved and usage/membership fees were lowered. These households are dissimilar to early adopters. Often they are living with elderly family members and are far more likely to own their own dwelling and automobile(s). The specific circumstances of early CS adopters mean that as CS memberships expand, the past patterns of vehicle utilization, car-shedding, vehicle kilometre travelled adjustment and greenhouse gas reductions are unlikely to be replicated.
Modelling Trip Generation Using Mobile Phone Data: A Latent Demographics Approach
Andrew Bwamable, University of LeedsShow Abstract
Charisma Farheen Choudhury, University of Leeds
Stephane Hess, University of Leeds
Trip generation plays a critical role in the four stage transport model because it is a key determinant of the accuracy of the subsequent stages. Traditional approaches to trip generation modelling rely on household travel surveys which are expensive and prone to reporting errors. Recently, there has been growing interest in the use of ubiquitous data sources (e.g. mobile phone data) for travel demand modelling. Mobile phone data has primarily been used for mobility pattern visualization and origin-destination matrix estimation. However, it has not been used in econometric models of travel demand (e.g. trip generation) due to missing demographic information. In this study, we address this limitation by making use of mobile phone data to calibrate a demographic group prediction model based on the phone usage and demographic information of a sub-sample of users. The calibrated model is then used to calculate the demographic group membership probabilities for each user in the data, and these are then used as an input in the estimation of a hybrid trip generation model. The hybrid trip generation model is calibrated using trip rates extracted from mobile phone data thereby mitigating the need for travel surveys. The hybrid model is compared with a traditional trip generation model that uses observed demographic variables. This comparative analysis shows that the hybrid model is a feasible alternative to the traditional model. This offers evidence that having the demographic information of a sub-sample of users makes mobile phone data a practical source of information for travel demand modelling.
Who Is Picking Up the Kid from Daycare? Understanding the Intra Household Dynamics in Drop Off and Pick Up Task Allocation for Households with Dependent Children
Adam Weiss, University of Calgary Schulich School of EngineeringShow Abstract
Khandker Nurul Habib, University of Toronto
This paper presents a joint model for the allocation of
drop-off and pick-up responsibilities and the choice of the daycare location for
two adult households with dependent children. This analysis aims to capture the
tradeoffs, which occur at the household level in terms of daycare location
selection and drop-off and pick-up allocation. The paper utilizes a stochastic
frontier modelling approach to predict feasible locations for each possible
allocation pair. The frontier model predicts the maximum distance an individual
would be willing to travel within a time budget constraint. This travel time
prediction is then applied to each individual for generating feasible location
sets given two endogenous anchor points. The paper then presents a joint
econometric choice model of task allocation and daycare location with
heterogeneous sampling correction factors for each possible allocation. Captured
within the model are the choice of who performs the drop-off and pick-up
activities, and the eventual location that is selected for the daycare. The
joint model clearly reveals that daycare location choices are more correlated
than the household level choice of task allocations (drop-off and pick-up). More
specifically, it seems that households with children would change day care
locations before changing task allocation choices. Furthermore, the household
structure provides key insights into both the choice to use daycare and the
allocation of drop-off and pick-up responsibilities.
Mitigating the Error Rate of an IPF-Based Population Synthesis Approach by Incorporating More Heterogeneity into the Initial Seed
Ismaïl Saadi, Universite de LiegeShow Abstract
Ahmed Mohamed El Saeid Mustafa, University of Liege
Jacques Teller, University of Liege
Mario Cools, University of Liege
In this paper, we propose a bi-level procedure for population synthesis to obtain good estimates of both the marginal and joint distributions. As a fitting-based algorithm, the Iterative Proportional Fitting (IPF) approach is capable of providing accurate synthetic population estimates from the marginal distributions perspective. The algorithm iteratively recalculates the different weights of the k-way contingency table until the deviation between the simulated and observed marginal distributions is minimized. Besides, the initial boundaries, referred to as the seed, are preserved in terms of proportions with respect to the full sample. This means that although the different values are updated along with the iterations, the correlation structure of the underlying population remains unaffected. In this paper, a hybrid model is proposed. Its originality resides in the incorporation of heterogeneity into the initial seed by using a Hidden Markov Model (HMM) (1). The enriched initial seed is then fitted to a set of stable marginal distributions. This new hybrid approach is very interesting as it for instance can be calibrated by an unlimited number of PUMS, and separate information can be merged. The estimation of the correlation structure (synthetic seed) is improved compared to the standard seed (stemming from a micro-sample) thanks to the HMM. In this regard, the results show that the coupling of IPF with the HMM method provides better estimates, i.e. decreased RMSE of 79.16% for 1% sample size, from both the marginal and the joint distributions points of view. This implies that enriching the micro-sample is absolutely necessary before fitting with any aggregate marginal distributions.
Discrete Choice Models with Dynamic Effects: Estimation and Application in Activity-Based Travel Demand Framework
Gaurav Vyas, WSPShow Abstract
Peter Vovsha, WSP
Most of the travel demand models in practice are estimated on a single-year cross-sectional data, which makes the model inherently static. This means that the underlying assumption in the model application for future years is that individual’s behavior will remain constant over time and the changes will be entirely driven by the explanatory variables such as socio-economic characteristics, land-use, and transportation Level of Service (LOS). However, this assumption is not necessarily true. Individual’s perception of different travel conditions and, thus, their travel behavior may also vary over time.
The paper illustrated how the model attributes which are dynamic over a period of time can be incorporated in a travel demand model in practice. The existing approach in practice is usually the aggregated calibration approach. In aggregate calibration approach, the core disaggregate model is estimated in a purely cross-sectional fashion (based on the latest available survey). Subsequently the model is calibrated for the base year and if needed re-calibrated to the future market share that has to be established externally. The model might benefit from adjustment of the corresponding constants for future years to reflect better the observed tendencies. However, this approach requires multiple scenarios to test. The approach described in this paper is a disaggregated approach where time trends were explicitly included in the estimated model as explanatory variables. The performances of these two approaches were compared with real data and concluded that two independent and different techniques proved to be in agreement at the aggregate level.
Impact of Site-Specific Data on the Accuracy of Volume Delay Functions
Ryley Stevens, Virginia Department of TransportationShow Abstract
Aidan Barkley, Virginia Department of Transportation
John Miller, Virginia Department of Transportation
Volume delay functions (VDFs) estimate travel speed based on volume, free flow speed, and capacity and help planners examine how route selection, mode choice, fuel economy, and other elements of transportation performance are affected by congestion. Although VDF calibration has garnered interest, determining why agencies should improve VDFs has received less attention. Using 2 years of observations at 8 interstate locations, the researchers quantified how site-specific data improve the accuracy of the Bureau of Public Roads (BPR), Akcelik, and conical VDFs and how this accuracy affects planning.
The results showed that using site-specific data to calibrate VDFs (compared to taking parameters and variables from the literature) improved mean absolute percent error by an average value of 20 percentage points and reduced the root mean squared error by 46%, from 16.7 to 9.0. However, the impact of such site-specific data on accuracy varied by VDF: it was greatest for the BPR VDF (improved error by 8.6 mph relative to taking values from the literature) but less for the conical VDF (reducing error by 3.5 mph). A key implication is that VDF selection depends on the availability of local data: there is not one VDF for all situations, but given a specific level of available local data, a significantly better VDF (p < 0.01) can be identified. Two case studies from the literature show that calibrating versus taking VDFs from the literature, can alter forecast fuel economy by 1.8%-14.0% and change forecast share for a mode of interest by 1.3 percentage points.
A Copula-Based Continuous Cross-Nested Logit Model for Tour Scheduling In Activity-Based Travel Demand Models
Sepehr Ghader, University of Maryland, College ParkShow Abstract
Carlos Carrion, University of Maryland, College Park
Liang Tang, University of Maryland, College Park
Arash Asadabadi, WSP
Lei Zhang, University of Maryland, College Park
This paper is focused on modelling the joint choice of arrival to the activity and departure from the activity. Each of the choices are modeled in continuous time using a modified version of Continuous Cross-Nested Logit (CCNL) model. The modified version has discrete nests which makes it computationally less cumbersome. Cross Nested Logit (CNL) model is able to capture various types of correlation between alternatives. CCNL generalizes CNL to capture correlations in continuous time, which better represents the real choice situation. In addition to correlation between alternatives, this paper uses Copula to capture correlation between two dependent choices of arrival to the activity and departure from the activity. Copula can derive the correlation structure from the data without knowing the actual bivariate distribution function. With its multidimensionality and power to capture different sorts of correlations, this modeling framework can be used as a time of day component for most activity-based models.
Lane Group Based Mesosopic Dynamic Network Loading Model for Congested Urban Network
Zongzhi Li, Illinois Institute of TechnologyShow Abstract
Xi Lu, Illinois Institute of Technology
This paper introduces a lane group based mesoscopic dynamic network loading (DNL) model along with an iterative computational process that captures the trajectory of each vehicle by considering interactions of vehicles in different lane groups of a multilane road segment or multilane approaches of an intersection. The model classifies each road segment or an intersection approach into a running section and a queueing section that is further split into a merging block and a discharging block, differentiates the lane group from the movement group, treats vehicles being discharged by vehicle movement category, contains rules to build vehicle queuing profiles, and accounts for lane geometry features of discharging zone and traffic responses to intersection control. This helps accurately estimate vehicle travel time on a road segment and delays at an intersection, making it possible to analyze the time-dependent progression of vehicles using a travel path from the origin to the destination. An urban street network in the densely populated Chicago central district is selected for model application. The modeled traffic volumes and vehicle speeds are found to be consistent with field measurements as indicated by the low root mean square errors (RMSE) and Geoffrey E. Havers (GEH) values. The proposed model can be used in conjunction with models for dynamic traffic assignment to develop effective congestion mitigation measures.
Dummy Coding vs Effects Coding for Categorical Variables in Choice Models: Clarifications and Extensions
Andrew Daly, University of LeedsShow Abstract
Thijs Dekker, University of Leeds
Stephane Hess, University of Leeds
This paper revisits the issue of the specification of categorical variables in choice models, in the context of ongoing discussions that one particular normalisation, namely effects coding, is superior to another, namely dummy coding. For an overview of the issue, the reader is referred to  or . We highlight the theoretical equivalence between the dummy and effects coding and show how parameter values from a model based on one normalisation can be transformed (after estimation) to those from a model with a different normalisation. We also highlight issues with the interpretation of effects coding, and put forward a more well-defined version of effects coding. Our results suggests that analysts using choice models in transport, where categorical variables regularly arise, can largely make their choice of normalisation on the grounds of convenience.
Development of External and Truck Components for a Regional Travel Model
Matthew Stratton, WSPShow Abstract
Christina Bernardo, WSP
Ashish Kulshrestha, WSP
Gaurav Vyas, WSP
Peter Vovsha, WSP
Rebekah Straub Anderson, Ohio Department of Transportation
Gregory Giaimo, No Organization
A regional travel model has to include various components in addition to the core part that generates travel demand of the regional population with the trip ends in the modeled region. These include external passenger trips with either origin or destination outside the modeled region (externals) as well as truck trips (both external and internal). This paper describes a possible approach to improve the existing practices of modeling external trips. This approach is based on two principles. The first one is that external trips should be derived from a larger “mega” model that plays a meta role with respect to the external-internal proportions. The second is that the core internal model has to be properly scaled to reflect on the external trip shares predicted by the inter-regional meta-model.
In the current project, full advantage was taken of the Ohio Statewide model to provide a basis for the external and truck components for the regional ABMs developed for three major cities (Columbus, Cleveland, and Cincinnati) and corresponding MPOs. The ABM structure was adjusted accordingly to eliminate possible double-counting between internal and external trips. The travel and traffic components that have to be extracted from the Statewide model to regional ABMs include external auto trips, commercial vehicles, and trucks.
The paper describes the corresponding technical details, results, and a discussion about possible further improvements of this important model component.
Pricing and Reliability Enhancement in the San Diego Activity-Based Travel Model
Nagendra Dhakar, RSGShow Abstract
Joel Freedman, RSG
Mark Bradley, RSG Inc
Wu Sun, San Diego Association of Governments
The estimation of demand for priced highway lanes is becoming increasingly important to agencies seeking to improve mobility and find alternative revenue sources for the provision of transportation infrastructure. However, many modeling tools fall short of what is required for robust estimates of demand with respect to toll and managed lanes in two key areas; the value-of-time is often aggregate and not consistently defined throughout the model system, and the reliability of transport infrastructure rarely taken into account. This paper describes an effort which implemented recommendations of the Strategic Highway Research Program C04 and L03\L04 tracks on pricing and reliability within a regional activity-based modeling system for the San Diego, California region. The implemented SHRP recommendations include distributed travel time sensitivities across the synthetic population and special travel markets, continuous cost sensitivity based on income, and multiple value-of-time bins in highway skimming and assignment. The work also included innovative research related to the analysis of travel time variability based upon a temporally disaggregate (1-minute interval) dataset of auto travel speeds for a majority of auto links in the San Diego network for the entire month of October 2012. The authors estimated regression equations that relate the standard deviation of travel time to link characteristics, incorporated reliability in auto travel skims, incorporated those skims in the travel demand model system, and calculated toll elasticities on toll roads in San Diego County. The enhanced model matched observed toll demand better than the original model. Resulting elasticity values were generally found in the ranges reported in the literature.
A Practical Method to Test the Validity of the Standard Gumbel Distribution in Logit-Based Multinomial Choice Models of Human Travel Behavior
Xin Ye, Tongji UniversityShow Abstract
Venu Garikapati, National Renewable Energy Laboratory (NREL)
Daehyun You, Maricopa Association of Governments
Ram Pendyala, Arizona State University
Most multinomial choice models, particularly in practice (e.g., multinomial logit model), assume an extreme-value Gumbel distribution for the random components of utility functions. The use of this distribution offers a closed-form likelihood expression when the utility maximization principle is applied to model choice behaviors. The maximum likelihood estimation method can be easily applied to estimate model coefficients. However, maximum likelihood estimators are consistent and efficient only if distributional assumptions on the random error terms are valid. It is therefore important to test the validity of underlying distributional assumptions that form the basis of parameter estimation and policy evaluation. In this paper, a practical but strict method is proposed to test the distributional assumption of the random component of utility functions in both the multinomial logit (MNL) model and multiple discrete-continuous extreme value (MDCEV) model. Based on a semi-nonparametric approach, a closed-form likelihood function that nests the MNL or MDCEV model being tested is derived. Then, the traditional likelihood ratio test can be applied to test violations of the standard Gumbel distribution assumption. Simulation experiments are conducted to show that the test yields acceptable Type-I and Type-II error probabilities at commonly available sample sizes. The test is then applied to three real-world discrete and discrete-continuous choice models. For all three models, the proposed test rejects the validity of the standard Gumbel distribution in most utility functions, calling for the development of approaches that overcome adverse effects of violations of distributional assumptions.
Can a Better Model Specification Avoid the Need to Move Away from Random Utility Maximisation?
Stephane Hess, University of LeedsShow Abstract
Matthew Beck, Institute of Transport and Logistics Studies
Romain Crastes dit Sourd, University of Leeds
An ever increasing number of applications in choice modelling in transport are looking at moving away from random utility maximisation (RUM) models, lured by the promise of more realistic behaviour in alternative structures. The most prominent example of such a structure in recent years has been the random regret minimisation model (RRM), though there are others. While these alternative structures are behaviourally interesting, researchers seem to at times forget the many reasons why RUM has been the workhorse in choice modelling for several decades, notably its firm grounding in economic theory. The present paper puts forward the idea that a more flexible treatment of heterogeneity in preferences across decision makers may reduce the benefits of moving away from RUM. We illustrate this point on the basis of three typical stated choice datasets from transport studies, offering strong support to our hypotheses.
A Multimodal Trip Generation Model to Assess Travel Impacts of Urban Developments in District of Columbia
Ryan Westrom, Ford Motor CompanyShow Abstract
Stephanie Dock, District Department of Transportation
Jamie Henson, Kittelson & Associates, Inc. (KAI)
Mackenzie Watten, Fehr & Peers
Anjuli Tapia, Fehr & Peers
Matthew Ridgway, Fehr & Peers
Jennifer Ziebarth, Fehr & Peers
Niranjani Prabhakar, Fehr & Peers
Nazneen Ferdous, Jacobs
Ramgiridhar Kilim, Jacobs
Raj Paradkar, Kimley-Horn and Associates, Inc.
The research effort described in this paper aims to develop a state of the practice methodology for estimating urban trip generation from mixed-use developments. The District Department of Transportation’s (DDOT) initiative focused on (1) developing and testing a data collection methodology, (2) collecting local data to substitute for Institute of Transportation Engineers’ (ITE’s) national data in trip rates estimation, and (3) developing a model/tool that incorporates factors identified affecting overall trip rate as well as trip rate by mode.
The final model accurately predicts total person trips and mode choice. The full set of models achieve better statistical performance (in terms of average model error and R-square) than ITE and other available existing research, while also including sensitivity to local environment and onsite context. Results indicate that this model is a more powerful and sensitive predictor of urban trip generation than the available existing research. The model makes two major steps forward: directly estimating total person trips, and is sensitive to the amount of parking provided on site – a major finding in the connection between parking provision and travel behavior at a local site level. The methodology will allow agencies to improve their assessment of expected trips from proposed buildings and therefore the level of impact a planned building may have on the transportation system.
Identification of Representative Time-Use Activity Patterns Using Fuzzy C-Means Clustering
Mohammad Hesam Hafezi, Dalhousie UniversityShow Abstract
Lei Liu, Dalhousie University
Hugh Millward, Saint Mary’s University
Analysis of the time-use activity patterns of
Does Compact Development Increase or Reduce Traffic Congestion?
Reid Ewing, University of UtahShow Abstract
Guang Tian, University of New Orleans
Torrey Lyons, University of North Carolina, Chapel Hill
Kathryn Terzano, University of Utah
From years of research, we know that compact development that is dense, diverse, well-designed, etc. produces fewer vehicle miles traveled (VMT) than sprawling development. But compact development also concentrates origins and destinations. No one has yet determined, using credible urban form metrics and credible congestion data, the net effect of these countervailing forces on area-wide congestion. Using compactness/sprawl metrics developed in an earlier project at the University of Utah, and congestion data from the Texas Transportation Institute’s (TTI’s) Urban Mobility Scorecard Annual Report database, this study seeks to determine which opposing point of view is correct. It does so by (1) measuring compactness, congestion, and control variables using the best national data available for U.S. urbanized areas and (2) relating these variables to one another using multivariate methods to determine whether compactness is positively or negatively related to congestion. Our models suggest that an increase in compactness reduces VMT, but also concentrates those VMT. The two effects roughly cancel each other out. This analysis does not support the idea that sprawl acts as a “traffic safety valve,” as some have claimed. However, it also does not support the reverse idea that compact development offers a solution to congestion, as others have claimed. Developing in a more compact manner may help at the margin, and providing more transit service may help at the margin, but the great reduction in congestion appears to be achievable through expansion of surface streets and higher highway user fees.
Modeling Workers' Daily Out-of-Home Maintenance Activity Participation and Duration
You-Lian Chu, ParsonsShow Abstract
This paper applies a multiple discrete-continuous choice model to determine worker’s decisions on interdependencies of maintenance activity participation and duration over different time periods in a worker’s day. To accommodate household effect, each time period is further defined according to whether the maintenance activities were conducted independently or jointly with other household members. Using the household survey data from the New York metropolitan area, the worker’s participation and duration decisions were jointly estimated for four time periods, including the morning commute, midday, evening commute, and after evening commute. To account for the censored nature of the duration data, a “censored system of equations” proposed by Shonkwiler and Yen was adopted and modified to increase asymptotic efficiency of the parameter estimates. Despite simplicity of the model and its ease of implementation, the estimation results offer valuable insights into the effects of socio-demographics, land use density, transportation service, and work duration characteristics on worker’s daily scheduling of maintenance activities and travel in a highly urbanized environment.
Temporal Origin-Destination Matrix Estimation of Passenger Car Trips in Medellin, Colombia
Carlos Gonzalez-Calderon, Universidad Nacional de ColombiaShow Abstract
John Jairo Posada-Henao, Universidad Nacional de Colombia
Susana Restrepo-Morantes, Universidad Nacional de Colombia
This paper develops a demand synthesis model based on entropy maximization to estimate origin-destination matrices for passenger cars trips using traffic counts for different years in an urban area. The model discussed in this paper improve current Origin-Destination matrix estimation techniques incorporating the total number of trips in the network as a constraint, and secondary data sources for different years, giving a point of comparison for the results. The performance of the formulation for different years (2005-2008) is tested in the Medellín (Colombia) network.
Development of a Future Year Large-Scale Microscopic Traffic Simulation Model
Craig Jordan, Old Dominion UniversityShow Abstract
Peter Foytik, Old Dominion University
Andrew Collins, Old Dominion University
R. Michael Robinson, Old Dominion University
This paper addresses the development of a future year large-scale microscopic traffic simulation model and identifies the challenges faced during the process. The construction of the model used a regional travel demand model and a calibrated base year microscopic simulation model and is described in this paper. The study area covered over 300 square miles and included the City of Virginia Beach as well as a section of the City of Chesapeake, Virginia. Roadway classifications range from local streets to freeway facilities and consist of 3,403 links, 1,808 nodes, and 874 intersections (438 signalized). The development of the future year model presented challenges not found in traditional microscopic models (data collection methods, conversion of OD matrices, calibration benchmarks, etc.). Identifying practical and efficient applications to overcome these challenges are areas of potential future research.
Spatiotemporal Traffic Forecasting: Review and Proposed Directions
Alireza Ermagun, Northwestern UniversityShow Abstract
David Levinson, University of Sydney
This paper systematically reviews studies that forecast short-term traffic conditions using spatial dependence between links. We synthesize 130 extracted research papers from two perspectives: (1) methodological framework, and (2) approach for capturing and incorporating spatial information. From the methodology side, spatial information boosts the accuracy of prediction, particularly in congested traffic regimes and for longer horizons. There is a broad and longstanding agreement that non-parametric methods outperform the naive statistical methods such as historical average, real time profile, and exponential smoothing. However, to make a conclusion regarding the performance of neural network methods against STARIMA family models, more research is needed in this field. From the spatial dependency detection side, we believe that a large gulf exists between the realistic spatial dependence of traffic links on a real network and the studied networks. This systematic review highlights that the field is approaching its maturity, while it is still as crude as it is perplexing. It is perplexing in the conceptual methodology, and it is crude in the capture of spatial information.
Combinatorial Tour Mode Choice
Peter Vovsha, WSPShow Abstract
James Hicks, WSP
Gaurav Vyas, WSP
Vladimir Livshits, Maricopa Association of Governments
Rebekah Straub Anderson, Ohio Department of Transportation
Gregory Giaimo, No Organization
In most Activity-Based Models (ABMs) in practice mode choice decisions are modeled in two steps. First the entire-tour mode combination is predicted based on the location of the primary destination of the tour (at this step the modeled tour is largely treated as a simple round trip). Secondly, a detailed trip mode is predicated conditionally upon the tour mode and given the specific origin and destination location for each trip.
The mode choice model applied for the ABMs recently developed for the Maricopa Association of Governments (MAG) and Ohio state DOT (ODOT) has a different structure where the tour-level and trip-level choices are integrated in a network combinatorial representation. The model considers all feasible trip mode combinations on the tour (in a similar way how a path dependent shortest path is built in a transportation network) and the tour mode combination emerges as the joint choice of trip modes. This model formulation imposes a lot of additional constraints compared to the two-step structure and in particular, with respect to the conditional linkages between different trip mode choices within the tour. This structure explicitly tracks the car status at origin and destination of each trip and constraints multi-model combinations such as park and ride to consider a logical location of the parking lot.
In terms of practical application, this approach suites some of the recent ABMs developed in practice where tour formation sub-model precedes tour mode choice in the model chain and eventually both models are equilibrated.
Travel Time Reliability with Boundedly Rational Travelers
Chao Sun, Jiangsu UniversityShow Abstract
Lin Cheng, Southeast University
Wenyun Tang, Southeast University
Jie Ma, Southeast University
Senlai Zhu, Southeast University
This paper presents a definition of boundedly rational confidence level (BRCL) which is the probability that a trip arrives within the shortest travel time budget plus the acceptable travel time difference (i.e., boundedly rational threshold). Using the method of data fitting, boundedly rational threshold is estimated. Then a reliability-based boundedly rational user equilibrium (R-BRUE) model that explicitly considers both travel time reliability and travelers’ bounded rationality in the route choice decision process is proposed. This new model hypothesizes that for each origin-destination (OD) pair no traveler can improve his or her BRCL by unilaterally changing routes. A route based quasi method of successive average algorithm is developed to solve the R-BRUE model. Numerical examples illustrate the essential ideas of the proposed model and the applicability of the proposed solution algorithm.
Complement or Competitior? Comparing car2go and Transit Travel Times, Prices, and Usage Patterns in Seattle
Xiasen Wang, University of WashingtonShow Abstract
Don MacKenzie, University of Washington
Zhiyong Cui, University of Washington
This paper compares the travel time and costs of trips taken by the free-float carsharing service car2go in Seattle, USA with the time and costs of taking the same trips by transit, to better understand how car2go may complement and/or compete with transit. We identify 237,000 car2go trips by observing available vehicles each minute from January 23 to May 10, 2016 through car2go’s application programming interface (API). We estimate driving times using the Google Maps API, developing a correction factor to account for temporal variation in traffic conditions, and develop a method to infer the average walking time to reach a car2go vehicle based on the locations of nearby vehicles. We estimate that the average car2go trip in Seattle costs $7.27 and involves 5.5 minutes of walking and 11 minutes of driving time, compared with 6 minutes walking, 13.6 minutes waiting, and 28 minutes in-vehicle time for the same trip by transit. Surprisingly, we find that although car2go complements transit by serving late-night trips and providing a faster but more costly alternative, it is not used disproportionately for trips where it offers especially large travel time savings compared with transit.
P2p Ridesharing with Ride-Back on Hov Lanes: Towards a Practical Alternative Mode for Daily Commuting
Roger Lloret-Batlle, University of California, IrvineShow Abstract
Neda Masoud, University of Michigan, Ann Arbor
Daisik Nam, University of California, Irvine
We design an economic benchmark for P2P ridesharing that takes advantage of time savings from HOV lanes. The ridesharing system is presented as an alternative mode for daily commuting, that is, we ensure a ride-back for the matched riders. The modelling is based on the Vickrey-Clarke-Groves (VCG) mechanism that is known to be efficient, incentive compatible and individually rational. Since it is known that VCG runs on a budget deficit, we classify agents into drivers and riders according to a novel multiparameter reserve price that also solves the revenue shortage. The allocation rule is formulated as a min-cost max-flow problem that is exact and fast solvable on polynomial time. The parametric study uses origin-destination demand data from Southern California Association of Governments (SCAG) and travel times are extracted from a professional webmap mapping service. The reserve prices are calibrated for empirical distributions of value of time and unit distance cost. Results show the method has revenue surplus over most of the reserve price parameter space, as well as offers high matching rates due to the inclusion of HOV travel time savings and reserve price structure.
Endogeneity Due to Missing Observations in a Learning Model For Travel Choice
Yue Tang, University of Massachusetts, AmherstShow Abstract
C. Angelo Guevara, Universidad de Chile
Song Gao, University of Massachusetts, Amherst
Learning-based models that can capture travelers'\
day-to-day learning process in repeated travel choice have gained popularity in
recent years. Meanwhile, the ever-increasing availability of ubiquitous sensors
such as smartphones provides abundant individual-level longitudinal data to help
validate and improve such models.
Latent Class Analysis of Residential and Work Location Choices
Rajesh Paleti, Pennsylvania State UniversityShow Abstract
Sabyasachee Mishra, University of Memphis
Khademul Haque, RSG
Afrid Sarker, California Department of Transportation (CALTRANS)
Mihalis Golias, University of Memphis
This paper developed a two-stage modeling framework for analyzing residential and work location choices with probabilistic choice sets. In the first stage, a household (or a worker) was assumed to select a neighborhood (such as central business district, urban, suburban etc.) to live (or work). In the second stage, the household (or worker) was assumed to choose a specific zone conditional on the selected neighborhood. The neighborhood choice model component takes the form of Manski model with latent choice sets. The model was used to analyze residential and work location decisions in Nashville, Tennessee. The model results indicate significant heterogeneity in the consideration probability of different neighborhood alternatives both in the residential and work location choices.
Choice Set Formation Behaviour in Selecting Travel Routes: Application of an Interactive Online Suvery Platform
Kiran Shakeel, University of New South WalesShow Abstract
Taha Rashidi, University of New South Wales
S. Travis Waller, University of New South Wales
Among various challenges associated with the analysis of route choice modelling, one of the major concerns is to formulate the choice set of alternatives that has the potential to provide somewhat precise prediction of demand for travel routes. Due to enormous and nearly infinite route alternatives, the subset of route alternatives in the choice set should be relevant and feasible conditional on the attributes considered most by the travellers in the decision making of route choice. This paper investigates the role and significance of different route choice set formations particularly with a focus on the perspective of travellers and that of the modeller. A revealed preference data is collected for Sydney residents regarding the information on their choice of route for their last commute trips. The survey tool is programmed in which Google maps APIs are utilised to effectively collect the route choice information, including the selected route and considered set of routes. Three discrete choice models are utilised to investigate the traveller’s inclination towards certain attributes of route considering both car and public transit routes. Further, the effect of possible bias that is generated due to the formation of route choice from the perspective of modeller is also analysed and presented with the results. The results show the intuitive signs of attributes with the travel time being the significant factor for route choice. The difference between the choice set considered by the traveller and by the modeller also suggests that those considered by the modeller possess enough variation providing the possibility of better capturing important factors affecting route choice behaviour.
Paradigms for Integrated Modeling of Activity-Travel Demand and Network Dynamics in an Era of Dynamic Mobility Management
Ram Pendyala, Arizona State UniversityShow Abstract
Daehyun You, Maricopa Association of Governments
Venu Garikapati, National Renewable Energy Laboratory (NREL)
Karthik Konduri, Amazon.com
Xuesong Zhou, Arizona State University
This paper describes various levels of an integrated transport modeling framework capable of reflecting the impacts of dynamic traffic management strategies and real-time traveler information provision on activity-travel demand and network dynamics. Many existing integrated model systems involve a rather loose coupling of activity-travel demand models (ABMs) and dynamic traffic assignment (DTA) models. These models do not have the capabilities necessary to simulate the impacts of emerging dynamic strategies that rely on real-time connectivity, communications, and proactive travel demand management to optimize network performance. In an effort to advance the development of integrated modeling platforms that can address “planning for operations”, this paper presents the design of increasingly detailed levels of the tightly integrated modeling framework called SimTRAVEL. In this modeling framework, the ABM and the DTA model communicate with one another along the continuous time axis with a view to reflect how activity-travel patterns evolve in response to network dynamics, and how network dynamics manifest themselves as travelers go about their daily activity-travel schedules. The highest level of the framework is capable of reflecting the full suite of behavioral and network dynamics that may result from a network disruption, information provision scheme, or dynamic mobility management strategy. The efficacy of the framework is demonstrated by implementing the platform using openAMOS as the ABM component and DTALite as the DTA model. Scenario analyses conducted using the Sioux Falls network show that the integrated modeling framework is able to provide behaviorally intuitive predictions of the impacts of real-time information provision strategies.