Ongoing D-STOP Projects

Real-Time Signal Control and Traffic Stability

Project ID: 125
University: University of Texas at Austin
Principal Investigator: Stephen Boyles
Start Date: 5/1/2016
Expected Completion Date: 8/31/2018

Project Summary

Congestion on urban arterials is largely centered around intersection control. Traditional traffic signal schemes are limited in their ability to adapt in real time to traffic conditions or by their ability to coordinate with each other to ensure adequate performance. Specifically, there is a tension between adaptivity (as with actuated signals) and coordination through pre-timed signals (signal progression). We propose to investigate whether routing protocols in telecommunications networks can be applied to resolve these problems. Specifically, the backpressure algorithm of Tassiulas & Emphremides (1992) can ensure system stability through decentralized control under relatively weak regularity conditions. It is as yet unknown whether this algorithm can be adapted to traffic signal systems, and if so, what modifications are needed. Traffic systems differ in several significant ways from telecommunication networks: each intersection approach has relatively few queues (lanes) that must be shared among traffic to various definitions. First-in, first-out constraints lead to head-of-line blocking effects, traffic waves move at a much slower speed than data packets, and traffic queues are tightly limited by physical space (finite buffers). Determining whether (and how) the backpressure concept can be adapted to traffic networks requires significant research, and has the potential to dramatically improve signal performance.

Objective

Internet of Moving Things using Full Duplex Mesh Networks

Project ID: 132
University: University of Texas at Austin
Principal Investigator: Sriram Vishwanath
Start Date: 5/1/16
Expected Completion Date: 12/31/2018

Project Summary

Through years of research, we have developed true full duplex communication systems (transmission and reception in the same band at the same time) using novel off the shelf components. Such radios are unique in their ability to listen while transmitting at the same frequency at the same time. Although other full duplex technologies exist (from work conducted at Stanford, Rice and Columbia), these technologies are typically antenna or custom-chip based. Our solution is unique in that it is based on off-the-shelf discrete components together with software. This ability to build a software-centric full duplex solution has many advantages, including low-cost, rapid reconfigurability, and agility. Full duplex radios are able to listen and talk simultaneously, making them ideally suited for mesh networking applications. Conventional mesh networking is highly prone to poor performance due to massive overheads and rigidity. Full duplex radios are much more flexible and adaptable, and can perform tasks such as handoff and scheduling in a low-overhead, rapid manner. This makes them ideally suited to be the basis for the Internet of Moving Things (IoMT). IoMT aims to connect all moving (and static) objects with one another—buses, cars, people, even their pets—without using a cellular or satellite backbone. It enables vehicle to vehicle (V2V) and vehicle to infrastructure (V2I) connectivity in a much more seamless fashion than currently thought possible. Full duplex based IoMT will provide low-cost connectivity between people, while helping us understand how people and vehicles move.

Objective

Spatial Correlation Estimation of Millimeter Vehicular Communication Channels Using Out-of-Band Information

Project ID: 136
University: University of Texas at Austin
Principal Investigator: Robert Heath
Start Date:
5/1/2017
Expected Completion Date: 5/31/2019

Project Summary

Wireless channels in vehicular communications systems rapidly vary due to the fast changes of their topology. Obtaining reliable instantaneous information about the propagation channel is invariably important in wireless communications. It is more challenging in vehicular communication systems especially at millimeter wave (mmWave) bands since the problem is further exacerbated by the hardware constraints required for mmWave systems. For example, a small number of RF chains and low-resolution ADCs enforce to limit to the number of measurements for and the direct access to the channels between transceivers.

Compared to the instantaneous channel information, the second order statistics (or the spatial correlation) of the channels vary slowly so it is relatively not hard to obtain them in general. In the case of mmWave systems, however, even acquiring the second order statics is still difficult because of the lack of direct access to the channel and possibly many transmit/receive antennas with hybrid beamforming architectures – inducing high training overhead. Therefore, to overcome the problem, we aim at developing a framework to leverage the second order statistics of out-of-band channels and to estimate mmWave channel correlations by using them. Specifically, algorithms will be developed to fetch out-of-band information from sub-6 GHz channels use this information for mmWave channel correlation estimation.

Objective

Joint Millimeter-Wave Communication and Radar for Automotive Applications

Project ID: 137
University: University of Texas at Austin
Principal Investigator: Robert Heath
Start Date:
5/1/2017
Expected Completion Date: 5/31/2019

Project Summary

Vehicular radar and communication are the two primary means of using radio frequency (RF) signals in transportation systems. Automotive radars provide high-resolution sensing using proprietary waveforms in millimeter wave (mmWave) bands and vehicular communications allow vehicles to exchange safety messages or raw sensor data. Both the techniques can be used for applications such as forward collision warning, cooperative adaptive cruise control, and pre-crash applications.

The project aims at developing a mmWave vehicular joint system that exploits the same waveform for both communication and radar operations, thus allowing hardware reuse. The use of a standard mmWave waveform, which provides access to a large bandwidth, will lead to significant advantages in terms of higher data rates for communication and better accuracy/resolution for radar operation compared with approaches based on sub-6 GHz frequencies. Our initial work was to propose the idea of using IEEE 802.11ad for a joint vehicular communication and radar system. This allows us to exploit the same spectrum and to leverage shared hardware based on the mmWave consumer WLAN standard. The approach is reasonable because the most prevalent vehicular communication standard, dedicated short-range communications (DSRC), is based on the WLAN standard. In this project, we propose to refine previously developed algorithms and develop a further framework and associated algorithm to better achieve vehicular radar and communication objectives. These algorithms shall be used for designing modified IEEE 802.11ad and associated processing techniques (e.g., by exploiting sparsity) that permit both radar and communication functionalities in a multi-target scenario. The results will allow insights about how to design an optimal joint waveform for different vehicular scenarios to meet the continuously growing performance requirements of a more advanced assisted driving and future autonomous driving.

Objective

Improved Models for Managed Lane Operations

Project ID: 140
University: University of Texas at Austin
Principal Investigator: Stephen Boyles
Start Date: 9/1/2017
Expected Completion Date: 8/31/2018

Project Summary

Managed lanes (ML) are increasingly being considered as a tool to mitigate congestion on highways with limited areas for capacity expansion. Managed lanes are dynamically priced based on the congestion level, and can be set either with the objective of maximum utilization (e.g., a public operator) or profit maximization (e.g., a private operator). Optimization models for determining these pricing policies make restrictive assumptions about the layout of these corridors (often a single entrance and exit) or knowledge of traveler characteristics on behalf of the modeler (e.g., distribution of willingness to pay). Developing new models to address these issues would allow for better utilization of these facilities.

Objective

In this project, we will study dynamic pricing models for managed lanes. We will focus on two aspects of dynamic pricing: (a) utilizing real time traffic measurements to inform parameters of the pricing model, and (b) developing a optimal pricing formulation for managed lanes with multiple entrances and exits.

V2I Managed Lanes Test Bed

Project ID: 142
University: University of Texas at Austin
Principal Investigator: James Kuhr
Start Date: 9/1/2017
Expected Completion Date: 09/30/2018

Project Summary

Vehicle-to-everything (V2X) technology connects vehicles, infrastructure, and any other communicating devices. By sending and receiving short communications, known as Basic Safety Messages (BSMs), vehicles will receive key information about their surroundings that can be relayed to their drivers to aid safe operations. The potential for safety applications under a mature V2X system is tremendous: through use of a mature system, NHTSA studies indicate that 81 percent of all unimpaired crashes could be avoided.

Research is critical to better understand how communication technologies may support or limit the potential impacts of connected vehicles. The project will investigate vehicle-to-infrastructure (V2I) communication technology to be installed along connected roadways, including DSRC, LTE and potential 5G technologies, and investigate V2I privacy and security threats.

Objective

Transit Policy in the Context of New Transportation Paradigms

Project ID: 150
University: University of Texas at Austin
Principal Investigator: James KuhrChandra Bhat, and Natalia Ruiz Juri
Start Date: 9/1/2017
Expected Completion Date: 09/30/2018

Project Summary

With changing transportation paradigms, there is significant potential for a shift in the balance between the overall population use of, and reliance on, ridesharing services versus traditional transportation options such as personal car ownership or transit use. This shift could lead to a realignment of the bulk of the responsibility for mobility to private entities and away from individual citizens and public entities. Today, as supplemental to the multitude of transportation options that are available, the availability, or lack thereof, of ridesharing services produces low to minimal risk to the traveling public. However, in a future in which ridesharing is optimally (widely) employed, the current independent nature of ridesharing services will influence wider community transit services. This problem statement explores the effects of new types of transportation on transit through the creation of several plausible future scenarios, and what policy decisions could potentially be made to ensure that transit is optimally employed.

Objective

Utilizing other D-STOP supported surveys (and potentially conducting our own as needed) that will identify attitudes towards ridesharing and various other methods of carpooling/car sharing that could serve as a competitor to transit services, we will work to analyze the potential effects of new transportation paradigms on transit system ridership. We will then identify key policy considerations for transit service providers in the next 10 years.

Video Data Analytics for Safer and More Efficient Mobility

Project ID: 151
University: University of Texas at Austin
Principal Investigator: Natalia Ruiz Juri
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary
Identifying and tracking objects in video data frames has become significantly easier in recent years, thanks to advances in artificial intelligence and greater access to increased computing power. Some commercially available products allow transportation agencies to leverage new video data analysis capabilities to support traffic analysis, operations, and safety applications. However, such products often require the use of dedicated hardware which is typically costly.

However, many transportation agencies own extensive networks of pan-zoom-tilt cameras, used mostly for traffic monitoring. The analysis of the traffic streams produced by such cameras can potentially yield valuable data sets for a number of planning and operation applications, and support analyses which are not possible using data from simpler sensors.

Data-driven, Real-time Traffic Signal Optimization: A Distributed Approach

Project ID: 152
University: University of Texas at Austin
Principal Investigator:  Stephen Boyles
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary
The vehicular transportation system is essentially a distributed computing system, but one with very limited communication and coordination abilities at present. These limitations are especially acute in light of the ambiguities of human interaction on the road and as the current standard system for signal coordination. We propose to build on prior development of an improved system for traffic control, based on improved, reliable communication between a traffic signal controller and vehicles. We also aim to study how, in both partially and fully connected environments, the concepts of distributed computer operating systems, communication networks, and optimization algorithms can be leveraged to improve such a control system. The goal is a highly abstracted and modular system that can be implemented by any city to optimize their traffic control systems using a high-performance computing cluster. In this regard, we aim to create an adaptive intersection control system that greatly improves upon traditional adaptive systems: one which leverages the power of computing to provide performance optimizations humans could never deliver.

Real-time, Targeted Incentives for Strategic Travelers

Project ID: 153
University: University of Texas at Austin
Principal Investigator:  Stephen Boyles
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary
New communication technologies allow targeted network control aimed at reducing congestion by identifying, in real time, alternative routes for travelers. In current practice, such measures involve messages broadcast to all (or many) travelers, or tolls which are assessed identically among vehicles. A more finessed approach has the potential of reducing congestion while alleviating some of the social equity concerns with tolling, or concerns about over-reaction to real-time travel information.

Tight-coupling of Vision, Radar, and Carrier-phase Differential GNSS for Robust All-weather Positioning

Project ID: 154
University: University of Texas at Austin
Principal Investigator:  Todd Humphreys
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary
Advanced driver assistance systems (ADAS) are a key technology for improving road safety. But both current and proposed ADAS are limited in important ways. Vision- and lidar-based ADAS performs poorly in heavy rain, snow, or fog. Lack of vehicle situational awareness due to these sensing limitations will unfortunately be the cause of many accidents, including fatalities, for connected and automated vehicles in the years to come.  The goal of this research is to develop and test a sensing strategy with robust perception: No blind spots, applicable to all driveable environments, and available in all weather conditions. We believe there are three key requirements for collaborative all-weather sensing:
– Precise vehicle positioning within a common reference frame
– Decimeter-accurate vision and radar mapping
– A means of quantifying the benefits of collaborative sensing

Modeling Willingness-to-share Trips in an Autonomous Vehicle Future: A Stochastic Psychological Latent Construct Approach

Project ID: 155
University: University of Texas at Austin
Principal Investigator:  Chandra Bhat
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary

Shared autonomous vehicle (SAV) systems, in which autonomous vehicles (AVs) are owned by transportation network companies that offer Mobility as a Service (MaaS) to customers, are gaining considerable research attention. SAVs have the potential to reduce vehicle ownership and parking requirements, improve traffic conditions, and minimize empty-vehicle travel. However, the extent to which each of these potentials can be achieved depends on consumers’ willingness to adopt such services as well as their disposition to sharing rides. While both MaaS and SAV adoption have received some attention in the literature (for example, Krueger et al., 2016; Rayle et al., 2016; Clewlow and Mishra, 2017; Dias et al., 2017; Lavieri et al., 2017), there is no specific discussion and measurement about willingness-to-share (WTS). The objective of this project is to develop this notion of WTS in the transportation context and propose a measuring and modeling approach to this concept. The developed modeling framework will be used to investigate variations in WTS and value of travel time (VTT) across distinct population segments for different trip purposes. Outcomes from this investigation can contribute to SAV adoption forecasts and guide SAV demand assumptions in traffic models.

Emerging Transportation Mobility Options and Technologies: A Comprehensive Analysis of Consumer Preferences using Survey and Supplementary Data

Project ID: 156
University: University of Texas at Austin
Principal Investigator:  Chandra Bhat
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary

The new emerging transportation landscape includes new mobility options and technologies, such as autonomous vehicles (AVs) and ride-hailing services (like Uber and Lyft). In this project, we will undertake a revealed preference survey of individuals’ current travel choices (including ridehailing) and obtain stated preferences/choices about individuals’ potential responses to new mobility options. The survey data will be combined with other sources of land-use and transportation network data to project travel demand into the future, and develop policy recommendations on how best to harness the potential of new mobility options to make transportation in urban areas safer, more reliable, and more resilient. In particular, the survey’s main objective is to collect a rich set of data that includes information about people’s travel behavior and their perceptions and attitudes towards advanced transportation technologies and mobility options. This survey will be a coordinated effort among four different universities and will collect data in five metropolitan areas (Phoenix, Atlanta, Austin, Houston, and Tampa) regarding users’ perceptions on new transportation technologies and services, allowing for cross-city comparisons. The D-STOP project will be responsible for all aspects of data collection in the Austin and Houston areas, and will contribute to data analysis across all five cities. The research will inform the development of behavioral models of technology adoption capable of reflecting the impacts of these disruptive forces on traveler behavior and values, and provide guidance to policy makers and urban transportation planning agencies in terms of regulations and preparing for the new era of transportation services.

Sensing and Communications in V2V and V2I Settings

Project ID: 157
University: University of Texas at Austin
Principal Investigator:  Sanjay Shakkottai 
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary

We consider the wireless scheduling problem of jointly scheduling resources (activating/de-activating and communications/sensing mode selection) at various locations on a vehicle, in order to optimize for both communications and sensing (e.g. through in-band radar). The resources could be antenna resources located at various points on a car (e.g. sides, back-bumper, front). An example scenario would be a car simultaneously engaged in V2V communications (with another car), V2I communications with a base-station, and sensing the environment (for location, obstacles, etc.). The resource allocation task here would be to dynamically select each of the resources for communications or sensing. When considering infrastructure nodes, one could also consider turning off resources to save energy. We build on our earlier research to study a few important new directions:

  1. Dynamic scheduling algorithms that use queue-lengths, channel state and current sensed-state in order to result in smaller delays and improved sensing fidelity. These algorithms would dynamically move modalities (communications/radar) across multiple antenna resources (e.g. different arrays located on multiple locations on a car) to optimize for both communications and sensing.
  2. Online-learning based algorithms to optimize the use of various antenna resources (e.g. learning-based beam pattern optimization for improved sensing).

Online Matching, Black-box Optimization and Hyper-parameter Tuning

Project ID: 158
University: University of Texas at Austin
Principal Investigator:  Sanjay Shakkottai 
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary

Machine learning algorithms form an integral part of modern data-driven platforms and systems. In the vehicular setting, examples range from platforms for matching — allocating passengers to vehicles, matching cargo freight carriers – to onboard deep-learning based algorithms for driver-assist. While these algorithms adapt a range of parameters based on new information, what is common is that they typically need certain parameters to be fixed (the hyper-parameters) and are outside the learning framework. Due the high-dimensionality of the parameter space, hyper-parameter tuning (i.e. selecting these hyper-parameters) is a major hurdle in deploying algorithms. We propose a principled approach for search and optimization of hyper-parameters.

Solving Perception Challenges for Autonomous Vehicles Using SGD

Project ID: 159
University: University of Texas at Austin
Principal Investigator:  Constantine Caramanis
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary

All levels of autonomy in vehicles faces a critical problem due to the fragility and lack of robustness of state of the art image classifiers to perturbations in the input image. Specifically, it has been repeatedly shown that classifiers that enjoy extremely high accuracy on test sets and challenge sets, are remarkably susceptible to misclassifying images that have small, but planted, perturbations. Stop signs can be misclassified as yield signs, with modifications that are imperceptible to the casual human observer.

We plan to use techniques and ideas from natural scene statistics, manifold learning and convex optimization, in order to build in robustness to non-perceptually-significant perturbations to images. A perturbation that morphs one image into another (according to human perception), is naturally a perturbation on the manifold of natural images. Adversarial attacks, such as those described above, are off of this manifold. A main challenge is that the desciprtion, while perhaps compelling, is purely at the conceptual level. We will use techniques for learning a compact and usable description of this manifold. We will develop ideas from dictionary learning, convex optimization, manifold learning and natural scene statistics.

Large-scale Optimization with Small-scale Data

Project ID: 160
University: University of Texas at Austin
Principal Investigator:  Constantine Caramanis
Start Date: 9/1/2018
Expected Completion Date: 08/31/2020

Project Summary

This project takes the same approach as Project 159, but addresses the same pressing problem from a completely different perspective, developing completely different ideas and methodological tools. All levels of autonomy in vehicles face a critical problem due to the fragility and lack of robustness of state of the art image classifiers to perturbations in the input image. Specifically, it has been repeatedly shown that classifiers that enjoy extremely high accuracy on test sets and challenge sets, are remarkably susceptible to misclassifying images that have small, but planted, perturbations. Stop signs can be misclassified as yield signs, with modifications that are imperceptible to the casual human observer.

Many researchers have identified this as a critical problem for neural networks. Our approach grows directly out of a previously funded D-STOP project. In that work, we have been working on developing fast SGD-based algorithms for large scale inference. Our preliminary experiments have demonstrated that those ideas can in fact be used to defend against these so-called adversarial attacks. Our proposal is to study how the natural exploration of the sample space that is drive by our SGD approach can be mapped to implicitly define the natural manifold of images. We conjecture that this is the critical concept for defending against adversarial attacks.