Connected & Autonomous Transportation Systems

Driven by recent booming transportation technologies (e.g., vehicle automation and electrification) and operational models (e.g., crowdsourcing, car sharing), transportation is undergoing a systematic revolution faster than ever expected. While we see that entrepreneurs have been aggressively turning these innovations into productivity and wealth, transportation researchers shall also ride on these trends: we need to plot a comprehensive blueprint of the best next generation transportation systems and identify a technically sound and socially beneficial pathway to get there from where we are today. The goal of our research group is to lead these academic undertakings by developing new theories, modeling methods and pilot technologies.

We deem that the utmost distinction of the next generation transportation systems lies in the unprecedented connectivity of subject matters ranging from individual vehicles to infrastructure systems. Building upon our recent successes, our future research will focus on two of the foremost challenges in this transportation revolution process. First, we will test whether self-driving technologies can be organized into connected self-driving traffic that can organize individual vehicles into smart traffic streams to achieve the system optimum. Further, instead of viewing transportation as an isolated system, we will model it as an integrated component of a nexus of connected infrastructure systems that provide heterogeneous resources and services (rather than aggregated mobility only) to our society.

Connected Self-Driving Vehicles

"It's too dangerous," Elon Musk said, "like you can't have a person driving a two-ton death machine". Whether agree with him or not, we have to recognize that the movements of current highway traffic are indeed hampered by a number of human drivers' limits, such as uncertain and error-prone maneuvers, slow reaction time, limited information processing capability, and uncooperative driving behavior. With these limits, it is almost impossible to precisely formulate fundamental human driving behavior, not to mention how to exactly improve it. While precise quantitative methods work well in systems governed by clear physical laws (e.g., power grid, water supply, communication networks), they have only limited successes in observing and controlling a highway transportation system due to elusive human driving behavior. As a result, the existing highway traffic management paradigm mainly relies on scattered, aggregated, reactive and empirical-based control measures from limited infrastructure facilities. Consequently, today's highway traffic frequently induces driving behaviors leaving everyone worse off, often amplifies local perturbations into systematic breakdowns, and is struggling to cope with unacceptably high incident rates.

Fortunately, the advent of connected vehicle and self-driving car technologies brings hope for transforming this traditional infrastructure-based traffic management paradigm into a proactive, disaggregated and cooperative vehicle-based control paradigm. Advanced sensing, communication and control technologies are on the verge of turning traditional manual vehicles into connected self-driving robots. It is expected that driverless vehicles will be widely deployed in a few years to serve various transportation activities ranging from daily commutes to package deliveries. These technological innovations will fundamentally transform highway vehicle behaviors and traffic operation paradigms. The pivoting factor is that while human driving behaviors can be hardly improved, controlled or even accurately observed, the driving rules of robot vehicles are precisely predictable, fully controllable and arbitrarily programmable. With these advantages, the new paradigm can directly control the trajectories of self-driving vehicles to simultaneously achieve both the optimal system performance and the smoothest individual driving experience. Even when self-driving vehicles only composes a small percentage of the total traffic, controlling their trajectories can still affect the surrounding manual vehicles and shift the traffic much closer to the system optimum.

Shooting Heuristic

we made promising theoretical findings and algorithmic developments in using connected self-driving vehicles to improve traffic performance of both signalized arterials and uninterrupted freeways. For example, we have proposed a parsimonious "shooting heuristic" method that can efficiently optimize multiple trajectories at a signalized highway segment. With knowledge borrowed from other fields (e.g., time geography), we show that this proposed method has profound theoretical properties (e.g., in relation to classic traffic flow models) and superior computational efficiency and optimality. This development has been adapted to freeway speed harmonization to mitigate stop-and-go traffic and optimize freeway performance in mixed traffic. we will continue to propose fundamental theories and methods and develop real-time control algorithms for establishing the new vehicle-based control paradigm. To estimate the full potential of this new paradigm, we will continue to analyze the utopian scenario where all vehicles are compliant robots but for more complex geometries, such as full intersections, multi-lane corridors, and complete networks. Then we will identify feasible pathways from today's traffic to the ideal scenario by investigating different market penetrations, different compliance rates, and both centralized and distributed control models. In particular, shared mobility and vehicle electrification will explored jointly with vehicle automation. Field experiments will be also conducted to demonstrate the effectiveness of these developments at a manageable yet realistic scale.These research initiatives are being partially supported by the following active research grants.

  • CAREER: Pathway to a Driverless Highway Transportation System: A Behavior Analysis and Trajectory Control Approach. Sponsored by the U.S. National Science Foundation. PI.
  • Speed Harmonization Fundamental Research: Phase II. Sponsored by USDOT, Federal Highway Administration (subcontracted from Leidos Inc). PI.

  • Connected Infrastructure Systems
    Interdependent Systems

    While transportation is traditionally viewed as an isolated system functioning by itself, it is actually intertwined with a number of other infrastructure systems (e.g., power grids, fuel supplies, sensor networks, communication systems) through physical, resource, information, and social interactions. With the rapid changes of transportation technologies and operational models in the recent years, the transportation system becomes increasingly interdependent with other infrastructure systems. Only by working together can these infrastructure systems provide resilient resources and services to sustain human society's day-to-day operations and long-term growth.

    The complicated interdependence between different infrastructure systems increases the vulnerability of each system component. An infrastructure system may be disrupted by not only problems occurring within this system but also the ripple effects from failures of other inter-connected systems. And local malfunctions may be cascaded and amplified through the inter-system connections to cause a global failure. Therefore, the traditional single-system operation paradigm cannot ensure infrastructure resilience against inter-system disruptions.

    This research direction aims to generalize traditional transportation system models into an interdependent system design framework.We have investigated resilient design of a number of infrastructure systems, including electric vehicle sharing, surveillance sensors, fuel supply, freight logistics, etc. More importantly, we have laid out a methodological foundation for modeling inter-connected infrastructure systems through a number of novel structures, such as supporting station decomposition and heterogeneous resource flow. The future research will focus on the behaviors and mechanisms at system borders. We will propose a series of modeling components and analysis methods to describe information sensing, cyber communications, physical operations, and social impacts of interdependent systems. Optimization models will be built to determine reliable deployment of surveillance sensors, mobile probes and information hubs under uncertain environments. Stochastic infrastructure disruption mechanisms are modeled to capture the risks from both normal operational perturbations and extreme catastrophic disasters. Both static and dynamic modeling methods will be adopted to investigate interdependent system operations at various time and spatial scales. Both centralized and decentralized management strategies will be tested to examine different institutional and organizational structures. This research direction benefits from existing analytical theories and numerical methods in graph theories, mathematical programming, distributed system control, and economic analysis. We embrace collaboration opportunities with interdisciplinary scholars from engineering, cyber-physical systems and social sciences to assure that the developments are sound at both physical and cyber layers and relevant to pressing needs from society. These efforts are partially suppoted by the following active research grants.

  • Collaborative Research: CRISP Type 1: Self-Organized Infrastructure-Population Nexus: A Distributed Heterogeneous Flow-based Modeling Framework. Sponsored by the U.S. National Science Foundation. Lead PI.
  • Collaborative Research: Planning Reliable and Resilient Transportation Networks against Correlated Infrastructure Disruptions. Sponsored by the U.S. National Science Foundation. Lead PI.
  • Intermodal Logistic System Network Design with Expedited Transportation Services. Sponsored by National Center for Intermodal Transportation for Economic Competitiveness (Tier I UTC). PI.