Completed D-STOP Projects

The Composite Marginal Likelihood (CML) Inference Approach with Applications to Discrete and Mixed Dependent Variable Models

Project ID: 101
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 9/30/13
Completion Date: 9/30/14

Project Summary     Read the report (PDF).

The composite marginal likelihood (CML) inference approach is a relatively simple approach that can be used when the full likelihood function is practically infeasible to evaluate due to underlying complex dependencies. The history of the approach may be traced back to the pseudo-likelihood approach of Besag (1974) for modeling spatial data, and has found traction in a variety of fields since, including genetics, spatial statistics, longitudinal analyses, and multivariate modeling. However, the CML method has found little coverage in the econometrics field, especially in discrete choice modeling. This project will fill this gap by identifying the value and potential applications of the method in discrete dependent variable modeling as well as mixed discrete and continuous dependent variable model systems.

Objective

A New Estimation Approach to Integrate Latent Psychological Constructs in Choice Modeling

Project ID: 102
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 9/30/13
Completion Date: 9/30/14

Project Summary     Read the report (PDF).

We propose a new multinomial probit-based model formulation for integrated choice and latent variable (ICLV) models, which has several important advantages relative to the traditional logit kernel-based ICLV formulation. Combining this multinomial probit (MNP)-based ICLV model formulation with Bhat’s maximum approximate composite marginal likelihood (MACML) inference approach resolves the specification and estimation challenges that are typically encountered with the traditional ICLV formulation estimated using simulation approaches. Our proposed approach can provide very substantial computational time advantages, because the dimensionality of integration in the log-likelihood function is independent of the number of latent variables. Further, our proposed approach easily accommodates ordinal indicators for the latent variables, as well as combinations of ordinal and continuous response indicators. The approach can be extended in a relatively straightforward fashion to also include nominal indicator variables. A simulation exercise in the virtual context of travel mode choice will be designed to evaluate the ability of the MACML approach to recover model parameters.

Objectives

Transit Demand and Routing after Autonomous Vehicle Availability

Project ID: 104
University: University of Texas at Austin
Principal Investigator: Stephen Boyles
Date Started: 1/1/14
Completion Date: 12/31/15

Project Summary     Read the report (PDF).

Autonomous vehicles (AVs) create the potential for improvements in traffic operations as well as new behaviors for travelers such as car sharing among trips through driverless repositioning. Most studies on AVs have focused on technology or traffic operations, and the impact of AVs on planning is currently unknown. Development of a planning model integrating AV improvements to traffic operations and the impact of new traveler behavior options will soon be of practical interest as AVs are currently test-driven on public roads. The altered traveler preferences may affect mode choice, leading to changes in transit demand and transit provider cost. An analysis of the model on metropolitan planning data will provide predictions on the impact of general AV ownership on network conditions.

Objectives

Semi-Autonomous Parking for Enhanced Safety and Efficiency

Project ID: 105
University: University of Texas at Austin
Principal Investigator: Sriram Vishwanath
Start Date: 4/10/14
Expected Completion Date: UPDATED: 06/30/2017

Project Summary     Read the report (PDF).

This project focuses on the use of tools from a combination of computer vision and localization based navigation schemes to aid the process of efficient and safe parking of vehicles in high density parking spaces. The principles of collision avoidance, simultaneous localization and mapping together with vision based actuation in robotics will be used to enable this functionality.

Objectives

Transportation Data Discovery Environment

Combining Millimeter-Wave Radar and Communication Paradigms for Automotive Applications: A Signal Processing Approach

Project ID: 106
University: University of Texas at Austin
Principal Investigators: Robert Heath and Chandra Bhat
Date Started: 9/1/14
Completion Date: 5/31/16

Project Summary      Read the report (PDF).

This project proposes to develop a conceptual mathematical model for combined paradigm of millimeter-wave communication and radar using a signal processing perspective. In particular, it will explore and investigate different possible signal frameworks for joint communication and radar paradigms, both with simultaneous or non-simultaneous applications. For these mathematical frameworks, novel algorithms will be developed targeting automotive applications. Our algorithms will leverage the performance of the joint paradigm by sharing information between the radar and communications signal frameworks. These algorithms will be further optimized to meet varied performance objectives in both rural and urban areas. This would require identifying the scenarios of interest in transportation environments. A design-trade off analysis will be carried out to meet the conflicting requirements of both the signal frameworks.
This project will also address the challenges unique to the combined mathematical framework such as beamforming, signal design and mutual interference. Furthermore, the performance of the combined paradigm would be compared with the individual signal frameworks of radar and communication. This will involve a detailed survey, mathematical model development and simulation of both these systems separately. The outcomes of the project are expected to dramatically improve safety for vehicles, bicycles, and pedestrians in all weather conditions and on all roadways.

Objectives

Coherence Time and Beam Alignment for mmWave Vehicular Communications

Project ID: 107
University: University of Texas at Austin
Principal Investigator: Robert Heath
Date Started: 9/1/14
Completion Date: 10/31/15

Project Summary     Read the report (PDF).

The goal of this project is to develop a learning based approach to significantly reduce the overhead leveraging side information including user positioning information and network geometry. After initial design, we use offline learning to construct an initial mapping from user feedback to the beam selection. During operation, new data are collected and used to refine the initial mapping. The online learning part can be viewed as a multi-armed bandit (MAB) problem and solved by leveraging the existing results on the subject.

Objectives

Improved Traffic Operations through Real-Time Data Collection and Control

Project ID: 108
University: University of Texas at Austin
Principal Investigators: Stephen Boyles and Sanjay Shakkottai
Date Started: 6/1/14
Completion Date: 5/31/16

Project Summary     Read the report (PDF).

New data collection technologies enable real-time traffic control more precise and efficient than what was earlier possible. This project develops novel control strategies based on this data, with an emphasis on two types of traffic control: (1) signalized intersection control, where cycle lengths and phasing may be adjusted based on observed demands and coordination with nearby intersections, and (2) pricing strategies, where tolls are adjusted in real time based on observed demand, in order to influence travelers to avoid congested areas. Both of these share a common methodological basis of adjusting traffic controls to prioritize particular vehicles to minimize congestion, accounting for human behavior and learning. The project will involve combining wireless routing algorithms with traffic engineering knowledge to create innovative control policies.

Objectives

Models for High Dimensional Mixed Regression

Project ID: 109
University: University of Texas at Austin
Principal Investigators: Constantine Caramanis and Chandra Bhat
Date Started: 9/30/13
Completion Date: 9/30/2016

Project Summary     Read the report (PDF).

We propose to consider the mixed regression problem in high dimensions, under adversarial and stochastic noise. We will consider convex optimization-based formulations with the aim of showing that it provably recovers the true solution. This agenda will seek to provide upper bounds on the recovery errors for both arbitrary noise and stochastic noise settings. We also will seek matching minimax lower bounds (up to log factors), showing that under certain assumptions, our algorithm is information-theoretically optimal. Our preliminary results represent the first (and currently only known) tractable algorithm guaranteeing successful recovery with tight bounds on recovery errors and sample complexity.
Mixture models treat observed data as a superposition of simple statistical processes. Thus they are particularly relevant in the transportation setting, when city-wide phenomena are often mixtures of simple processes (cut-through traffic, intra-city movement, etc.).

Objectives

Streaming PCA with Many Missing Entries

Project ID: 110
University: University of Texas at Austin
Principal Investigator: Constantine Caramanis
Date Started: 9/30/13
Completion Date: 12/31/15

Project Summary     Read the report (PDF).

We propose to consider the streaming memory-constrained principal component analysis (PCA) problem with missing entries, where the available storage is linear in the dimensionality of the problem, and each vector has so many missing entries that matrix completion is not possible. For this problem, we propose a method based on a block power update approach introduced in our previous work. We show on synthetic as well as benchmark data sets that our approach outperforms existing approaches for streaming PCA by a significant margin for several interesting problem settings. We also consider the popular spiked covariance model with randomly missing entries, and obtain the first known global convergence guarantees for this problem. We show that our method converges to the true “spike” using a number of samples that is linear in the dimension of the data. Moreover, our memory requirement is also linear in the ambient dimension. Thus, both memory and sample complexity have optimal scaling with dimension.
Streaming PCA is extremely relevant in the setting where the resolution of our sensors outpaces our ability to store massive amounts of data. This is precisely the setting we face as we increase the amount of intelligence and high precision/resolution sensors on the fleet of private and commercial vehicles.

Objective

Greedy Subspace Clustering

Project ID: 111
University: University of Texas at Austin
Principal Investigator: Constantine Caramanis
Date Started: 12/1/14
Completion Date: 9/30/16

Project Summary     Read the report (PDF).

We propose to consider the problem of subspace clustering: given points that lie on or near the union of many low-dimensional linear subspaces, recover the subspaces. To this end, one first identifies sets of points close to the same subspace and uses the sets to estimate the subspaces. As the geometric structure of the clusters (linear subspaces) forbids proper performance of general distance based approaches such as K-means, many model-specific methods have been proposed. In this paper, we provide new simple and efficient algorithms for this problem. Our statistical analysis shows that the algorithms are guaranteed exact (perfect) clustering performance under certain conditions on the number of points and the affinity between subspaces. These conditions are weaker than those considered in the standard statistical literature. Experimental results on synthetic data generated from the standard unions of subspaces model demonstrate our theory. We also show that our algorithm performs competitively against state-of-the-art algorithms on real-world applications such as motion segmentation and face clustering, but with much simpler implementation and lower computational cost.

Objective

An Empirical Investigation into the Time-Use and Activity Patterns of Dual-Earner Couples With and Without Young Children

Project ID: 112
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 3/1/14
Completion Date: 9/30/15

Project Summary      Read the report (PDF).

This project examines the time-use patterns of adults in dual-earner households with and without children as a function of several individual and household socio-demographics and employment characteristics. A disaggregate activity purpose classification including both in-home and out-of-home activity pursuits will be used because of the travel demand relevance of out-of-home pursuits, as well as to examine both mobility-related and general time-use related social exclusion and time poverty issues. The study uses the Nested Multiple Discrete Continuous Extreme Value (NMDCEV) model, which recognizes that time-decisions entail the choice of participating in one or more activity purposes along with the amount of time to invest in each chosen activity purpose, and allows generic correlation structures to account for common unobserved factors that might impact the choice of multiple alternatives. The 2010 American Time Use Survey (ATUS) data is used for the empirical analysis.

Objectives

A New Generalized Heterogeneous Data Model (GHDM) to Jointly Model Mixed Types of Dependent Variables

Project ID: 113
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 3/1/14
Completion Date: 9/30/15

Project Summary     Read the report (PDF).

This project examines the time-use patterns of adults in dual-earner households with and without children as a function of several individual and household socio-demographics and employment characteristics. A disaggregate activity purpose classification including both in-home and out-of-home activity pursuits will be used because of the travel demand relevance of out-of-home pursuits, as well as to examine both mobility-related and general time-use related social exclusion and time poverty issues. The study uses the Nested Multiple Discrete Continuous Extreme Value (MDCNEV) model, which recognizes that time-decisions entail the choice of participating in one or more activity purposes along with the amount of time to invest in each chosen activity purpose, and allows generic correlation structures to account for common unobserved factors that might impact the choice of multiple alternatives. The 2010 American Time Use Survey (ATUS) data is used for the empirical analysis.

Objective

A New Spatial (Social) Interaction Discrete Choice Model Accommodating for Unobserved Effects due to Endogenous Network Formation

Project ID: 114
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 3/1/14
Completion Date: 9/30/15

Project Summary     Read the report (PDF).

This project formulates a model that extends the traditional panel discrete choice model to include social/spatial dependencies in the form of dyadic interactions between each pair of decision-makers. In addition, the formulation accommodates spatial correlation effects as well as allows a global spatial structure to be placed on the individual-specific unobserved response sensitivity to exogenous variables. We interpret these latter two effects, sometimes referred to as spatial drift effects, as originating from endogenous group formation. To our knowledge, we are the first to suggest this endogenous group formation interpretation for spatial drift effects in the social/spatial interactions literature. The formulation is motivated in a travel mode choice context, but is applicable in a wide variety of other empirical contexts.

Objective

High-precision GPS Vehicle Tracking to Improve Safety

Project ID: 115
University: University of Texas at Austin
Principal Investigators: Jennifer Duthie and Todd Humphreys
Date Started: 1/15/15
Completion Date: 8/31/16

Project Summary   Read the report (PDF).

Commercial Global Positioning System (GPS) devices are being used in transportation for applications including vehicle navigation, traffic monitoring, and tracking commercial and public transit vehicles. The current state-of-practice technology in GPS devices typically has 10-meter accuracy and can properly answer the needs in the above applications. When it comes to safety, where driver behavior is important, new technologies for high-precision (i.e., centimeter-level) mobility detection are required. It is with high-precision GPS traces that safety can be evaluated by identifying when drivers are drowsy or distracted, and anticipating problems before they occur. Through this project, the team will build low-cost high-precision GPS devices, obtain GPS traces by placing the devices on buses, and analyzing the traces to identify driver behavior indicators that could anticipate a safety concern before it occurs.

Objective

Infrastructure-Informed Travel Sheds

Project ID: 116
University: University of Texas at Austin
Principal Investigator: Jennifer Duthie
Date Started: 9/1/14
Completion Date: 12/31/15

Project Summary     Read the full report (PDF).

An infrastructure-informed index is needed for pedestrians and bicyclists to relate the natural and built environment with its impact on perceived travel distance and time. The objective is to develop an easy-to-use metric for use at all levels, allowing transportation planners to make better-informed decisions when planning or redeveloping a city or area. Building off of previous research efforts, attributes were determined and weighted to capture the characteristics of a link, then summed to create the infrastructure informed index for pedestrians and bicyclists, respectively. Pedestrian perception data collected by the PI previously was used to determine the attributes. The indices were then visualized using ArcGIS mapping tools, creating a service area around specific origin or destination points to see the effective area a pedestrian or bicyclist can travel taking into account the effects of the infrastructure along the route.

Objectives

The Formulation and Estimation of a Spatial Skew-Normal Generalized Ordered-Response Model

Project ID: 117
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 1/1/15
Completion Date: 6/30/16

Project Summary      Read the report (PDF).

Ordered-response (OR) choice models are now widely used in many different disciplines, including sociology, biology, political science, marketing, and transportation. OR models may be used when analyzing ordinal discrete outcome data that may be considered as manifestations of an underlying scale that is endowed with a natural ordering. In this proposal, we will use the GOR structure as the starting point, and extend the formulation in two different directions. The first direction relates to the distribution of the kernel error distribution, and the second relates to spatial dependence. We will apply the proposed model to examine the determinants of bicycling frequency using data from the Puget Sound Regional Council in the state of Washington.

Objective

A Latent Class Multiple Constraint Multiple Discrete-Continuous Extreme Value Model of Time Use and Goods Consumption

Project ID: 118
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 1/1/15
Completion Date: 6/30/16

Project Summary    Read the report (PDF).

The aim of this proposal is to develop a microeconomic time-use framework that (a) accommodates technological relationships between time allocated to activities and goods consumption, and (b) proposed a discrete distribution for the response coefficients. This latent class model will be able to identify different segments of the population, each one of them with different effects of the exogenous variables on time allocation, activity participation, and goods consumption. This endogenous segmentation will be compared in a comprehensive fashion with the typical segmented estimation of microeconomic time use models (of the type discussed in the first paragraph of this abstract) from a theoretical, conceptual, and empirical data fit standpoint. The empirical analysis will be pursued using a 2012 Dutch data set on weekly time use and good expenditure.

Objective

A Comprehensive Dwelling Unit Choice Model Accommodating Psychological Constructs Within A Search Strategy for Consideration Set Formation

Project ID: 119
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 1/1/15
Completion Date: 12/31/15

Project Summary     Read the report (PDF).

This project adopts a dwelling unit level of analysis and considers a probabilistic choice set generation approach for residential choice modeling. In doing so, we accommodate the fact that housing choices involve both characteristics of the dwelling unit and its location, while also mimicking the search process that underlies housing decisions. In particular, we proposed to model a complete range of dwelling unit choices that include tenure type (rent or own), housing type (single family detached, single family attached, or apartment complex), number of bedrooms, number of bathrooms, number of storeys (one or multiple), square footage of the house, lot size, housing costs, density of residential neighborhood, and commute distance. Bhat’s (2014) generalized heterogeneous data model (GHDM) system will be used to accommodate the different types of dependent outcomes associated with housing choices, while capturing jointness caused by unobserved factors. The proposed analytic framework will be applied to study housing choices using data derived from the 2009 American Housing Survey (AHS), sponsored by the Department of Housing and Urban Development (HUD) and conducted by the U.S. Census Bureau.

Objective

On Accommodating Spatial Interactions in a Generalized Heterogeneous Data Model (GHDM) of Mixed Types of Dependent Variables

Project ID: 120
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Date Started: 1/1/15
Completion Date: 12/31/15

Project Summary     Read the report (PDF).

Multi-dimensional dependent outcome models are of interest in several fields, including land-use and transportation, biology, finance, and econometrics, just to name a few. The primary motivation for modeling dependent outcomes jointly is that there may be common underlying unobserved factors (attitudes, values, and lifestyle factors) of decision-makers that impact multiple dependent outcomes simultaneously. Even as there has been increasing emphasis on mixed data outcome modeling, there also has been a growing interest in accommodating spatial (and social) dependency effects among decision-makers in mixed data modeling. This is because spatial/social interactions can be exploited by decision-makers to achieve desired system end-states. In the current project, we use the important insight that the analyst can generate spatial dependence across multiple and mixed outcomes by specifying spatial dependence in the “soft” psychological construct (latent) variables underlying the many outcomes.

Objective

Travel Modeling in an Era of Connected and Automated Transportation Systems: An Investigation in the Dallas-Fort Worth Area

Project ID: 122
University: University of Texas at Austin
Principal Investigator: Chandra Bhat
Completion Date: 12/31/16

Project Summary      Read the report (PDF).

There is substantial anticipation and excitement in the area of connected/automated vehicles (CAVs) and transportation systems in terms of their potential to improve transportation safety and enhance mobility and accessibility. CAVs can have a substantial impact on travel patterns and roadway performance, and on mobile source-emissions, fundamentally altering strategic planning decision. Within this context, and given that metropolitan planning organizations (MPOs) incorporate a multi-decade (20 or more years) planning horizon in their long-term planning process, it is important that NCTCOG considers the implications of CAVs early on.The research team includes Natalia Ruiz Juri, James Kuhr, Mason Gemar, and Jen Duthie.

Objective

 Analyzing Millimeter Wave Vehicular Communication Systems in Urban Areas

Project ID: 123
University: University of Texas at Austin
Principal Investigator: Robert Heath
Start Date: 5/1/2016
Expected Completion Date: 5/31/2017

Project Summary      Read the report (PDF).

The project aims at developing a tractable model by stochastic geometry to analyze the performance of these vehicular networks in urban environments and the effects of the on line-of-sight (LOS) distance, non-line-of-sight (NLOS) interferers. Given the intensity of the streets and transmitters as well as the statistics of vehicular mobility, it is possible to evaluate the vehicular network in terms of coverage probability and capacity. These results will reveal insights about how to give the optimal strategy of deploying the base stations under different vehicular environment.

Objectives

Exploiting DSRC Information to Reduce Millimeter Wave Beam Alignment Overhead in Vehicular Environments

Project ID: 124
University: University of Texas at Austin
Principal Investigator: Robert Heath
Start Date: 5/1/2016
Expected Completion Date: 5/31/2017

Project Summary      Read the report (PDF).

This project aims at developing an efficient beam alignment algorithm for millimeter wave (mmWave) vehicular communications taking advantage of the side information available in vehicular context. Due to its propagation characteristics, the antenna gains from directional beams are necessary to maintain link quality, but these gains can be achieved only when beams are properly aligned. Conventional methods, such as the one proposed in the IEEE 802.11ad standard, do not use any side information and must endure large overhead that is unacceptable in mobile environments. In this project, we will focus on the DSRC, which exchanges basic safety messages that convey potential useful information for beam alignment such as position, speed, and acceleration of each vehicle. The outcome of this project is an algorithm with two components. The first one will use position information to efficiently perform initial beam alignment, while the second one will make use of kinematic information (e.g., velocity) to maintain the alignment. This algorithm will be essential in providing robust mmWave links that support high data rates to enable automated driving.

Objectives

Online Learning for Freight—An Examination of Queueing Regret

Project ID: 127
University: University of Texas at Austin
Principal Investigator: Sanjay Shakkottai
Completion Date: 12/31/16

Project Summary     Read the report (PDF).

As recently noted by The Economist, a large part of the freight trucking market is unconsolidated. There is a major opportunity to develop online platforms that match individual consumers (people who need freight delivery) with the resource suppliers (individual trucks and truckers). Such platforms can be game-changers, and dramatically improve the efficiency of freight haulage. At the core of this platform is a matching algorithm. This needs to learn from online data—learn both current supply (available trucks and their preferences) and demand (packages to be transported)—and match supply and demand in an online manner. Our research will focus on developing matching algorithms for this setting.

Objectives

Large-Scale Linear Programs in Planning and Prediction

Project ID: 128
University: University of Texas at Austin
Principal Investigator: Constantine Caramanis
Start Date: 4/1/2016
Expected Completion Date: 6/30/2017

Project Summary     Read the report (PDF).

Large-scale linear programs are at the core of many traffic-related optimization problems in both planning and prediction. Moreover, many of these involve significant uncertainty, and hence are modeled using either chance constraints, or robust optimization. Chance constraints and robust optimization are by now classical approaches for dealing with uncertainty. The ultimate goal in each of these areas, is to find an explicit convex reformulation that provides some approximation to the original (uncertain) optimization problem. The work in these areas has helped us obtain a nearly comprehensive understanding of when convex reformulations (and approximations) are possible, and what the quality of the approximation is. Yet little has been said about truly tractable solutions—solutions where running time for the uncertain problem is comparable (perhaps even less than!) the time to solve the problem without any uncertainty. As networks grow in size, and our ability to capture more data rapidly increases, it is of paramount importance to rethink our theory of robust and uncertain optimization for transportation applications, to one that is computationally oriented.

Objective

Imputing Missing Data via Sparse Reconstruction Techniques

Project ID: 129
University: University of Texas at Austin
Principal Investigator: Constantine Caramanis
Start Date: 4/1/2016
Expected Completion Date: 6/30/2017

Project Summary     Read the report (PDF).

The State of Texas does not currently have an automated approach for estimating volumes for links without counts. This research project proposes the development of an automated system to efficiently estimate the traffic volumes on uncounted links, in the event of rare disturbances of the typical traffic flow. The idea we plan to leverage is that the road network provides a mixing effect, whereby localized disturbances (accidents, flooding, road damage, etc.) have an impact whose effect can be measured at many places across the city. This forms the important analog to the well-known uncertainty principle, whereby a signal cannot be sparse in both the time and frequency domains—a result that is critically utilized in the signal reconstruction algorithms for fMRI.

Objectives

Performance Trade-Off for a Joint Automotive Radar and Communication System

Project ID: 133
University: University of Texas at Austin
Principal Investigator: Robert Heath
Start Date: 5/1/2016
Expected Completion Date: 5/31/2017

Project Summary     Read the report (PDF).

The IEEE 802.11ad waveform can be used for automotive radar by exploiting the Golay complementary sequences in the preamble of a frame. The performance of radar, however, is limited by the preamble structure. In this paper, we propose an adaptive preamble design that permits a trade-off between radar parameters’ estimation accuracy and communication rate. To quantify this trade-off, we propose a minimum mean square error (MMSE) metric based on rate distortion theory. The simulation results demonstrate that by adapting the preamble structure, we can achieve decimeter-level range mean square error (MSE) per symbol duration and gigabit per second (Gbps) data rates simultaneously for a distance up to 280 m.