Abstract

Driverless (or fully-automated) vehicles (AVs) are expected to fundamentally change how individuals and households travel and how vehicles interact with roadway infrastructure. Privately-owned AVs (PAVs), when operated within households, offer travel options that distinguish them from conventional vehicles (CVs), such as remote parking, returning home to park, and serving other household members. These options—available through deadheading—can lead to an increase in vehicle miles traveled (VMT). The goals of this dissertation are to (i) explore the expected travel patterns of PAVs, (ii) analyze their impacts on transportation system performance, and (iii) propose design and policy changes to mitigate the negative impacts of PAVs and leverage their benefits.In this context, this dissertation presents three models and corresponding case studies. First, I propose a parking assignment model to analyze the impact of PAV parking behavior on travel patterns and parking facility demand and performance. The case study finds that significant VMT increases occur due to PAVs traveling to remote parking locations after dropping off travelers at activity locations, and that balancing fees and capacities of parking spaces can reduce the extra VMT. Second, I introduce a new policy and infrastructure system aimed at reducing VMT that is similar to a park-and-ride (PNR) system. Instead of traditional fixed-route transit, my proposed system includes transfer stations where travelers can switch from their PAVs to on-demand, door-to-door shared-use AVs (SAVs) that enhance traveler convenience and service reliability. By optimizing transfer station locations, the case study demonstrates significant reductions in both VMT and vehicle hours traveled (VHT) within the region. Third, I extend the routing and scheduling of PAVs to the decision-making process within households. I introduce the Household Activity Pattern Problem with AV-enabled Intermodal Trips (HAPP-AV-IT) that incorporates SAV, public transit, and transit-based intermodal travel options. The case study results reveal that travelers are likely to choose long deadheading options, such as returning home, to optimize household vehicle operations. The model also demonstrates that intermodal trips can reduce both the household’s travel distance and overall travel costs. Although the precise performance of AVs on road networks remains uncertain, the findings of this dissertation suggest that additional VMT from PAV deadheading could negatively affect transportation systems. As we move closer to the era of widespread AV adoption, it becomes increasingly important for planners and researchers to develop policies and infrastructure systems that reduce PAV deadheading miles. The methodological advancements and practical insights presented in this dissertation provide a strong foundation for addressing these challenges and preparing for the transformative impact of AVs.

Abstract

Time-varying networks and techniques developed to study them have been used to analyze dynamic systems in social, computational, biological, and other contexts. Significant progress has been made in this area in recent years, resulting from a combination of statistical advances and improved computational resources, giving rise to a range of new research questions. This thesis addresses problems related to three lines of inquiry involving dynamic networks: data collection designs; the conditions needed for structural stability of an evolving network; and the computational scalability of statistical models for network dynamics. The first contribution involves a commonly neglected problem concerning data collection protocols for dynamic network data: the impact of in-design missingness. A systematic formalization is offered for the widely used class of retrospective life history designs, and it is shown that design parameters have nontrivial effects on both the quantity of missingness and the impact of such missingness on network modeling and reconstruction. Using a simulation study, we also show how the consequences of design parameters for inference vary as a function of look-back time relative to the time of measurement. The second contribution of this thesis is related to a fundamental question of network dynamics: when or where are changes in a network most likely to occur? A novel approach is taken to this question, by exploring its complement — what factors stabilize a network (or subgraphs thereof) and make it resistant to change? For networks whose behavior can be parameterized in exponential family form, a formal characterization of the graph-stabilizing region of the parameter space is shown to correspond to a convex polytope in the parameter space. A related construction can be used to find subgraphs that are or are not stable with respect to a given parameter vector, and to identify edge variables that are most vulnerable to perturbation. Finally, the third contribution of this thesis is to scalable parameter estimation for a class of temporal exponential family random graph models (TERGM) from sampled data. An algorithm is proposed that allows accurate approximation of maximum likelihood estimates for certain classes of TERGMs from egocentrically sampled retrospective life history data, without requiring simulation of the underlying network (a major bottleneck when the network size is large). Estimation time for this algorithm scales with the data size, and not with the size of the network, allowing it to be employed on very large populations. 

Abstract

Trip chaining is a common phenomenon generally known as linking multiple activities and trips in one travel process. A good understanding about trip chaining complexity is important for travel demand model development and for transportation policy design. However, most of the existing studies on trip chaining limit the complexity classification scheme on number of trips chained and neglect other dimensions that also elevate the degree of complexity. The purpose of this study is to develop a new approach, Tour Complexity Index (TCI), that integrates the multi-dimensional nature of trip chaining into the complexity assessment.

The study contains three analysis components. The first component introduces the TCI approach as a trip chaining complexity measure that not only considers number of trips chained but also includes the spatial relationship across destinations, the route arrangement, and the urban environment of the destinations. By comparing descriptive statistics and generalized linear model results from TCI approach with those from traditional approach, we find that the TCI approach offers more information regarding trip chaining and mode choice. The application of TCI is further demonstrated in the following components. The second component investigates the intrapersonal daily and weekly travel variability with travel characterized by TCI and mode choice. The result reinforces an argument in current literature that the common single-day travel survey may produce biased estimation due to the day-to-day variance in travel behavior. Result also finds that proximity to a new transit service from place of residence is connected with a decline in variability. The third component explores a framework for travel pattern recognition where pattern is characterized by TCI as well. The discrepancy analysis which is a generalized analysis of variance (ANOVA) method is applied to associate individual characteristics with travel pattern. In addition, both components use Sequential Alignment Method (SAM) for travel pattern representation. The TCI approach and proposed analysis frameworks are validated using the longitudinal GPS trajectory data collected between 2011 and 2013 at west Los Angeles area for Expo Study.

Abstract

As sensors that measure daily human activity become increasingly affordable and ubiquitous, there is a corresponding need for algorithms that unearth useful information from the resulting sensor observations. Many of these sensors record a time series of counts reflecting two behaviors: 1) the underlying hourly, daily, and weekly rhythms of natural human activity, and 2) bursty periods of unusual behavior. This dissertation explores a probabilistic framework for human-generated count data that (a) models the underlying recurrent patterns and (b) simultaneously separates and characterizes unusual activity via a Poisson-Markov model. The problems of event detection and characterization using real world, noisy sensor data with significant portions of data missing and corrupted measurements due to sensor failure are investigated. The framework is extended in order to perform higher level inferences, such as linking event models in a multi-sensor building occupancy model, and incorporating the occupancy measurement from loop detectors (in addition to the count measurement) to apply the model to problems in transportation research.

Abstract

Estimating and forecasting travel demand have been a popular study topic among transportation researchers; however the research needs to pursue new direction with the advent of data from the potential availability of newer types of data previously not envisaged. In this dissertation, the author reviews previous studies on this topic and develops approaches for two aspects of travel demand analysis in the transportation network: A newer OD estimation method and a household activity-based demand modeling framework. First, a trip-based dynamic OD estimation model is developed. Several previous studies on OD trip table estimation focused on a static problem and many recent dynamic OD estimation methods also have not sufficiently proved their practical applicability. In order to overcome the shortcomings, this dissertation introduces supplementary information (i.e., vehicle trajectory data) to a dynamic OD estimation model. However, the trip-based approach has certain well-known shortcomings. OD estimation results can not give satisfactory solutions for forecasting purposes, and the estimated OD table only contains materialized trips, which implies that no latent travel demand is included in the table. Therefore, the estimated OD table does not have sufficient information for identifying the real travel demand pattern and it is not so useful for transportation planning works. Contrarily, a standard four-step model has a better capability for explaining a travel demand pattern. However, when we load the OD trip table calculated by the four-step model, we might see some discrepancies between simulated traffic patterns and the ground truth. The discrepancies can come from various factors such as insufficient network capacities and unexplained influencing factors. When the discrepancy is caused by insufficient network capacities, then it can be solved by an iterative adjusting procedure. Using the ground truth such as link traffic counts, it might be updated correctly. However, if the discrepancies come from incapability of the four-step model, then we should look for a new approach. The capability of the four-step model already has been criticized continuously by numerous activity researchers because a trip-based approach does not correctly consider the real motivation of travel. To overcome these drawbacks, the second item of fucud in the dissertation is in developing a dynamic agent-based household activity and travel demand simulation model framework named DYNAHAP. The framework calculates a demand pattern in terms of activity chains generated by synthetic families. A traffic simulator then executes the activity chains, and finally an aggregated dynamic traffic pattern is generated. In order to calibrate DYNAHAP, huge activity data should be gathered. Such tasks had been regarded very difficult or even nearly impossible before, but with the development of data collecting technologies, currently we have several ways for collecting the activity chains of individuals. Like vehicle trajectory data, sample activity chains collected from personal communication devices such as PDA (Personal digital assistant) could be used for DYNAHAP calibration. Some numerical test results also will be given for proving the performance of the developed models. In last chapter, some important issues for future study are also discussed.

Abstract

Current travel forecasting models have had limited sensitivity to policy decisions. One of the primary challenges with travel forecasting models (both experimental and those implemented) is limitations in the data. The primary data source, the daily travel diary, is limited in both accuracy and sample size. The daily travel diary has known problems with underreporting, time inaccuracies, respondent fatigue, and other human errors. Global positioning systems (GPS) have been recently used to supplement the daily travel diary. As GPS becomes more accurate, reliable, and cost effective, could it entirely replace the daily travel diary? A number of efforts have used GPS data for route choice studies and to supplement daily travel diaries by providing more accurate time data, and determining under-reporting rates. GPS is also used in computer assisted daily travel diaries, reminding respondents of activities they may have forgotten to report. GPS devices record times and locations of each activity and the trips between those activities. To use GPS data to replace the daily travel diary one need only predict the activity types. The goal of this research is to develop and test a model to predict activity types based solely on: (1) GPS data from devices placed on the individual’s vehicle or person, (2) Land use data, such as location type, expressed as GIS data, and (3) Demographic data for the individual and the household. This thesis summarizes models developed using discriminant analysis and classification/regression trees. The models predicted in which of 26 different activity types the individual participated. Accuracy for out of home activities for the best model was 63%. When combed with the activity of being at home (which can be accurately predicted if we know the individuals home location) an accuracy of 79% was achieved (72% if you consider that GPS data may miss as much as 10% of trips). Since travel diaries have been known to underreport trips by as much as 25%, GPS data with the model developed can be very competitive. It is even more appealing considering the time inaccuracies and human error associated with travel diaries.