This workshop aims to provide insight into current developments in transportation network modeling facilitated by the new opportunities and challenges created by emerging trends in the field.
Recent technical advancements, including information, communication and sensing technologies, connected and/or autonomous vehicles, vehicle-to-infrastructure connectivity are a few of the driving factors in transforming today’s transportation systems. The big data that is both generated and required by these emerging technologies has further inspired the development and application of novel analysis approaches, including the use of artificial intelligence (AI) and cloud-based computing, for modeling, prediction and data-driven decision making in transportation networks. Finally, new perspectives in transportation, including shared mobility, mobility-as-a-service and multimodal systems form another driving force in shaping the future transportation networks.
Network modeling, impacts and applications of:
o Stationary and mobile sensing technologies
o Connected and/or autonomous vehicles
o Vehicle-to-infrastructure connectivity
o Vehicle, intersection, and network-level control strategies
Data-driven modeling and analysis as applied to:
o Network performance characterization
o Network performance prediction
o Decision making and control
Advanced modeling of:
o Shared mobility fleets
o Multimodal systems
o Interactions between emerging modes and services
• Monika Filipovska, University of Connecticut
• Hani S. Mahmassani, Northwestern University
• Lili Du, University of Florida
Sponsored by TRB AEP40-5 subcommittee on Emerging Technologies in Network Modeling.
• Joseph Chow and Gyugeun Yoon, New York University
• Sharon Di, Columbia University
• Lili Du, University of Florida
• Monika Filipovska, University of Connecticut
• Michael Hyland, University of California, Irvine
• Michael Levin, University of Minnesota
• Amelia Regan, University of California, Irvine
• Alireza Talebpour, University of Illinois Urbana-Champaign
• Cathy Wu, MIT
More information on speakers and workshop talks coming soon.
The workshop is intended for researchers, practitioners and graduate students interested in transportation network modeling and its relation to emerging technologies, data analytics and trends in the field of transportation. New developments reshaping transportation networks are of interest to transportation planners, mobility providers and city administrators.
The talks will cover some of the key aspects of modeling approaches and applications related to emerging technologies, analytics and perspectives in transportation and demonstrate cutting-edge research. The workshop will be concluded with a 35-minute panel discussion with all of the speakers.
Abstract: The introduction of automated vehicles is likely to change human drivers’
behavior on the road. Such changes can lead to both positive and negative impacts
on traffic flow dynamics. In this presentation, I will talk about our efforts to
characterize the potential changes in human behavior in the presence of automated
vehicles and will present an approach to managing human-automated vehicle interactions
towards improving traffic flow efficiency.
Bio: Dr. Alireza Talebpour received his B.S. and M.S. degrees in Civil Engineering from
Sharif University of Technology, Tehran, Iran, in 2007 and 2009, respectively. He received
his Ph.D. in Civil and Environmental Engineering from Northwestern University, Evanston, IL,
USA, in 2015. He is currently an Assistant Professor in the Department of Civil and
Environmental Engineering at the University of Illinois at Urbana-Champaign. His current
research focuses on theoretical and experimental studies of human-automated vehicle
interactions and he is the vehicle development lead in one of the USDOT Automated Driving
Demonstration (ADS) grants, AVA: Automated Vehicles for All.
Abstract: Recent years, wireless communication, onboard computation facilities (personal digital assistants, smartphones, etc.), and advanced sensor techniques
(loop detector, camera, GPS-based vehicle probe, etc.) have enabled a well-connected and data-rich transportation system, i.e., connected vehicle system (CVS). Even though the CVS has been granted a great potential to smartly route
travelers (or CVs) to avoid traffic congestion, scholars have recognized that as the majority vehicles become CVs, current real-time information provision and routing methods may worsen traffic congestion, given each CV still selfishly and
independently chooses its own shortest path (independent routing). This inherent deficiency of the current routing methods is rooted in the inconsistency between system performance (system-optimality) and individual vehicles' route choice behavior (user-optimality).
By viewing this, our studies have introduced a novel Coordinated In-Vehicle Routing Mechanism (CRM), which coordinates the routing decisions of a group of CVs en route to address the overreaction phenomenon and the inability to control system performance. While the CRM is shown to outperform independent routing methods with respect to system performance, it is not designed to ensure a specific
level of system performance. To boost the performance of the CRM, we further investigate information perturbation strategies for the CRM. It is an information provision scheme that strategically alters the traffic information sent to CVs opting in the CRM to mitigate network congestion. Our studies proved that the system performance improvement thanks to the information perturbation is more significant than the corresponding average user optimality loss.
Bio: Dr. Du is an associate professor in the Department of Civil and Coastal Engineering, University of Florida. She also worked as an assistant and then an associate professor at Illinois Institute of Technology from 2012-2017, and as a Post-doctoral Research Associate for NEXTRANS at Purdue University from 2008 to 2012. Dr. Du received her Ph.D. degree in Decision Sciences and Engineering Systems with a
minor in Operations Research and Statistics from Rensselaer Polytechnic Institute in 2008. Dr. Du also received her MS degree in Industrial Engineering from Tsinghua University in 2003 and BS degree in Mechanical Engineering from Xi’an Jiaotong University in 1998. Dr. Du’s research is characterized by integrating operations research, network modeling, game theory, control theory, machine learning and statistical methods into transportation system analysis and network modeling. Her current research mainly focuses on AV/CV/CAV impacts, mobility on demand, resilience, big data analytics in traffic flow and network analysis.
Abstract: If a transit planner has limited knowledge about the prevailing demand, there exists a risk of implementing a route set inefficient to serve the demand. It is more likely to happen when introducing a novel mobility system that the public is not familiar with. Therefore, this proposed methodology adapts the Knowledge gradient, one of the optimal learning policies, to sequentially expand a transit route system. The algorithm evaluates alternative links to which a route can extend based on prior knowledge about the reward, demand coverage. It examines deterministic reward and knowledge gradient, which represents the value of exploring a link. Two examples are provided: grid network with artificial demand data and simplified graph of Public Use Microdata Areas in New York City with household travel survey.
Bio: Gyugeun Yoon is a Ph.D. candidate in Transportation Planning and Engineering at New York University. His research interests include transit route planning, urban mobility, reinforcement learning, and sequential decision process.
Co-author: Joseph Cho, New York University
Abstract: As this era’s biggest game-changer, autonomous vehicles (AV) are expected to exhibit new driving and travel behaviors, thanks to their sensing, communication, and computational capabilities. However, a majority of studies assume AVs are essentially human drivers but react faster, “see” farther, and “know” the road environment better. We believe AVs’ most disruptive characteristic lies in its intelligent goal-seeking and adapting behavior. Building on this understanding, we propose a dynamic game-based control leveraging the notion of mean-field games (MFG). I will first introduce how MFG can be applied to the decision-making process of a large number of AVs. To illustrate the potential advantage that AVs may bring to stabilize traffic, I will then introduce a multi-class game where AVs are modeled as intelligent game-players and HVs are modeled using a classical non-equilibrium traffic flow model. Last but not the least, I will talk about how the MFG-based control is generalized to road networks, in which the optimal controls of both velocity and route choice need to be solved for AVs, by resorting to nonlinear complementarity problems.
Bio: Dr. Sharon Di is an Assistant Professor of Civil Engineering and Engineering Mechanics at Columbia University. Dr. Di applies optimization, game theory, and data analytics to large data collected from various types of traffic sensors, including individual tracing devices such as GPDs. Her studies of travel behavior focus on such factors as travel demand, high-occupancy travel lanes, and the effects of ride-hailing services like Uber, as well as on the future role of connected and automated vehicles. Dr. Di is also a committee member of the Center for Smart Cities, at Columbia’s Data Science Institute.
She received a BS in traffic engineering, summa cum laude, in 2005 and an MA in transportation information and control engineering in 2008 from Tongji University, China. She received a PhD in civil, environmental, and geo-engineering from the University of Minnesota, Twin Cities, in 2014. Dr. Di received a Chan Wui & Yunyin Rising Star Workshop Fellowship for Early Career Professionals from the Transportation Research Board in 2016.
Abstract: Ride-hailing services have brought significant mobility benefits to travelers; however, they have also increased vehicle miles traveled (VMT), congestion, and vehicle emissions. Shared-ride or ride-splitting services are seen by some as having the potential to provide similar mobility benefits to travelers with significantly lower VMT, congestion, and emissions levels. Unfortunately, shared-ride service options still represent a small share of all vehicle-based shared mobility trips in most cities. While the reasons for this are manifold, undoubtedly operational inefficiencies play an important role.
The shared-ride mobility service operational problem has traditionally been modeled as a combined dynamic traveler-vehicle assignment and routing (i.e., sequencing of pickups and drop-offs) problem, wherein vehicles are assumed to take the shortest path between each pair of pickup/drop-off locations. In our study, we explicitly incorporate path-finding into the operational problem and consider both travel time and proximity to future demands on paths. This non-myopic approach implicitly incorporates future information about the system into the control function.
The presentation will discuss the underlying operational problem, modeling approach, computational results, implications for mobility service providers and transportation planners/policymakers, and future research directions. The computational results compare the proposed path-finding approach with the conventional shortest path approach in terms of user wait times and in-vehicle travel times along with fleet VMT and fleet utilization.
Bio: Dr. Michael Hyland is an Assistant Professor of Civil and Environmental Engineering (CEE) at the University of California-Irvine, where he is affiliated with the Institute of Transportation Studies. Dr. Hyland works to improve the modeling, analysis, planning, and control of urban transportation systems to help create smarter (i.e., more efficient, sustainable, and affordable) cities through research and teaching. His research interests include emerging transportation systems such as bikesharing, ridesharing, and shared-use automated vehicle mobility services, as well as the integration of these emerging services with existing transportation modes. Before joining the faculty at UC Irvine, he earned his PhD in CEE from Northwestern University and his B.S. and Master’s degrees in CEE from Cornell University. Dr. Hyland is a two-time recipient of the Dwight David Eisenhower Transportation Fellowship and was named one of the Top 20 Future Leaders in Transportation by the Eno Center for Transportation in 2016.
Co-authors: Dingtong Yang and Navjyoth Sarma, University of California-Irvine
Abstract: This paper presents a matching mechanism for assigning drivers to routes where the drivers pay a toll for the marginal delay they impose on other drivers. The simple matching mechanism is derived from the deterministic algorithm for online bipartite matching. The toll, which is anticipatory in design, is an adaption of one proposed by Dong et al. Our research proves that the matching mechanism proposed here is Pareto user-optimal, that is, it is fair to all drivers and achieves a competitive ratio of 1 + log(m), where m is the number of available routes, when applied with a goal of minimizing total network travel cost.
Bio: Dr. Amelia Regan is a Professor of Computer Science and Transportation Systems Engineering at the University of California-Irvine. The overarching themes of her work are in efficiency, sustainability, and security in transportation systems. Her research interests include cyber physical transportation systems, dynamic and stochastic network optimization, combinatorial optimization, warehouse logistics, optimal contracting, freight transportation planning, technology adoption in transportation, machine learning tools for temporal-spatial data analysis, congestion pricing, technologies to improve the safety, comfort and convenience of pedestrians and disabled travelers and matching markets for urban resource allocation. Since 1997, Dr. Regan’s research has been supported by various sources including the National Science Foundation, the Transportation Research Board and JB Hunt Inc., and has been published in more than 160 refereed journal articles and conference proceedings papers in Journals including (among others) IEEE Transactions on Intelligent Transportation Systems, IEEE Network, IEEE Transactions on Computers, IEEE/ACM Transactions on Networking, Transportation Research (A, B, C and E), Transportation Science, Operations Research and INFOR. Dr. Regan received an NSF CAREER award in 1999.
Co-author: J. Ceasar Aguma, University of California-Irvine
Abstract: Modeling travel time variability in transportation networks gives more comprehensive knowledge of the network state and allows for risk-based decision-making. In stochastic dynamic networks with dependencies, past and current network state information can be used to adjust the knowledge of future network states, which may lead to changes in corresponding optimal routing solutions. This study considers the problem of optimal reliable routing in stochastic dynamic networks, adaptive to en-route in-vehicle information or information from a connected vehicle (CV) environment. An approach is presented for proactive and reactive routing solutions with access to en-route in vehicle and CV information, respectively. Approximations targeted at improving computational efficiency for large-scale implementations of the approaches are also presented.
Bio: Dr. Monika Filipovska is an Assistant Professor of Civil and Environmental Engineering (CEE) at the University of Connecticut, affiliated with the Connecticut Transportation Institute. She holds PhD and MS degrees in CEE from Northwestern University and a dual B.S. degree in Urban Systems Engineering and Mathematics from New York University. Her research is in the areas of dynamic transportation networks and traffic flow modeling, focused on applications of emerging technologies and data-driven analytics approaches for reliability modeling, prediction, and routing problems.
Co-author: Hani S. Mahmassani, Northwestern University
Abstract: Max-pressure traffic signal control has many desirable properties. It is analytically proven to maximize network throughput if demand could be served by any signal control. Despite its network-level stability properties, the control itself is decentralized and therefore easily computed by individual intersection controllers. Discussions with city engineers have suggested that a major barrier to implementation in practice is the non-cyclical phase actuation of max-pressure control, which can actuate any phase, in arbitrary order, to serve the queue(s) with highest pressure. This arbitrary phase selection may be confusing to travelers expecting a signal cycle and is therefore unacceptable to some city traffic engineers. This paper revises the original max-pressure control to include a signal cycle constraint. The max-pressure control must actuate an exogenous set of phases in order, with each phase actuated at least one time step per cycle. Each cycle has a maximum length, but the length can be reduced if desired. Within those constraints, we define a modified max-pressure control and prove its maximum stability property. The revised max-pressure control takes the form of a model predictive control with a one cycle lookahead, but we prove that the optimal solution can be easily found by enumerating over phases. The policy is still decentralized. Numerical results show that as expected, the cyclical max-pressure control performs slightly worse than the original max-pressure control due to the additional constraints, but with the advantage of greater palatability for implementation in practice.
Bio: Dr. Michael Levin is an Assistant Professor of Civil, Environmental and Geo- Engineering at the University of Minnesota. His research focuses on modeling connected autonomous vehicles (CAVs) and intelligent transportation systems to predict and optimize how these future technologies will affect travel demand and traffic flow. He received his PhD and M.S.E in Civil Engineering in 2017 and 2015, respectively, and his B.S. in Computer Science in 2013 from the University of Texas at Austin.
Abstract: The mixture of automated and human driven vehicles poses a multitude of technical challenges, but holds opportunities for congestion mitigation, safety, and environmental impacts. Motivated by recent successes achieving system-level congestion elimination on small, idealized traffic settings using deep reinforcement learning, this talk addresses two major remaining obstacles: scaling to a diverse array of traffic scenarios, and near-term concerns around the robustness and trust of automated vehicles. To overcome these challenges, the work studies: 1) the potential of transfer learning for generalizing knowledge across diverse traffic scenarios, and 2) the potential for human drivers to emulate learned congestion-mitigating control laws. These results have implications for sustainability and the near-term impact of mixed fleets.
Bio: Cathy Wu is an Assistant Professor at MIT in LIDS, CEE, and IDSS. She holds a PhD from UC Berkeley, and B.S. and M.Eng from MIT, all in EECS, and completed a Postdoc at Microsoft Research. Her interests are broadly in machine learning and mobility. She studies the technical challenges surrounding the integration of autonomy into societal systems. Her work has been acknowledged by several awards, including the 2019 IEEE ITSS Best Ph.D. Dissertation Award, 2019 Microsoft Location Summit Hall of Fame, 2018 Milton Pikarsky Memorial Dissertation Award, the 2016 IEEE ITSC Best Paper Award, and numerous fellowships, and appeared in the press, including Wired and Science Magazine.
Co-author: Zhongxia Yan, Massachusetts Institute of Technology