top of page


When I started my academic career in 2002, I established a motto for myself – Rigorous, Creative and Relevant. My research principle is to pursue logical reasoning using rigorous scientific approaches, while still thinking outside the box. Ultimately, I want my research to be meaningful and relevant to real-world challenges that affect people’s life.

Through the innovation and achievement milestones chronology chart below and a brief narrative of my research accomplishments in this section, I would like to show how this motto has been present throughout my career. The Rigorous – 95% of my journal publications are Web of Science (WoS) indexed, including with the highly prestigious Transportation Science. The Creative – my research ideas are out-of-the-box, and some of them are patent worthy. The Relevant – all the problems I addressed are motivated by and pertinent to real-world problems or tools used by professional practitioners or academicians.

Since 2002, I consistently devoted myself to advancing the foundational theoretical research and practical applications in network modeling, particularly in the area of dynamic traffic assignment (DTA). These endeavors span three areas of DTA related research: DTA Core, DTA Support, and DTA Application.

The DTA Core area advanced the state-of-the-art of DTA by building a whole new simulation-based DTA (SBDTA) model that has re-engineered the mesoscopic vehicular simulation logic, time-dependent shortest path algorithms, and the vehicle assignment methods. These algorithmic advancements were implemented with the most efficient modern computing architecture. The hallmark accomplishment in this area is the development of the empirically validated Anisotropic Mesoscopic Simulation (AMS) model that exhibits microscopic model like traffic flow properties with superb run-time performance (1E2 – 1E3 times faster than a typical microscopic simulation model). The AMS has been implemented in DynusT and its excellent traffic flow properties, like shockwave propagation and merging and diverging effects consistent with classical traffic flow theories, have been some of the main reasons that DynusT is highly trusted by practitioners who use the model to depict traffic dynamics and congestion patterns truthfully. Another significant advancement in this research area is the development of the Gap-Function Vehicle-based assignment algorithms that effectively identify and improve underperforming vehicles so that dynamic user equilibrium convergence can be enhanced.

The DTA Support area focuses on the supporting functions and methodologies that are critical for making DTA a practical modeling tool. These methods include the calibration of large-scale static and dynamic origin-destination matrices, which need to match both link counts and speed profiles for a regional model with thousands of zones. I proposed a one-norm formulation with linear transformation as a Linear Programming (LP) problem (as opposed to the traditional computationally intractable quadratic formulation). This method immediately opened the possibility of allowing the region-wide DTA model with thousands of zones (consequently millions of LP variables) to be calibrated in a computationally tractable manner. During this period, I also delved into the research area of network partitioning. The goal of network partitioning is to facilitate sub-area analysis and hierarchical traffic controls that seamlessly connect planning oriented regional DTA models to operation focused sub-area models. The OD calibration and network partitioning capabilities were also built into the DynusT software.

While I devoted a significant portion of my energy to DTA algorithmic research, I put the developed algorithms and models into real-life tests by conducting application-oriented studies by myself, as well as by releasing DynusT as an open source software package to the public in 2007. Both activities allowed me to tackle various real-life application challenges beyond what was initially envisioned during the theoretical development phase. The DTA Application area empirically verifies the real-world applicability of the DTA model by examining the validity and reasonableness of the model outputs against real-world observations such as travel counts or travel survey data. The properties of the developed DTA model were also examined through various sensitivity tests and what-if scenarios. Numerous refinements and adjustments to the algorithm and software were made, and this iterative process gradually made DynusT more stable and trustworthy, leading to DynusT quickly becoming a sophisticated professional tool accepted by practitioners. As a result, between 2007 and 2010, DynusT was downloaded more than 400 times internationally and cited or used in 33 published studies.

Another significant body of research that I contributed to is the simulation-optimization area. In this study field, a simulation approach is included in an optimization algorithmic structure to inform the optimization model of more realistic model inputs and parameters that could be either endogenously affected by the optimization model solutions or be more accurately estimated during the solution process. I applied the methodologies along with a stochastic programming approach to the dynamic message signs and refueling stations location problems.

While I made strides in simulation-based DTA research, I also started a line of the investigation in the analytical system-optimal DTA based on the Cell-Transmission Model (CTM) framework. I studied the fundamental properties and proposed the single-destination super sink network transformation to utilize the Linear-Program (LP) formulation to solve multi-dimensional decisions (direction, destination, departure time, and route) for evacuation scenarios in which a ground-zero disaster requires optimal evacuation out of a defined hot zone. Furthermore, I started to work with my students to examine the equivalent properties of earliest arrival and the minimal cost system optimal DTA problem and reached several important conclusions on mathematical properties. This line of research led to several highly cited journal papers in IIE Transactions and Transportation Science.

Since 2010, I continued to deepen the three DTA related areas while expanding to new areas. These emerging areas include integrating DTA with activity-based models (ABM) and agent-based modeling and simulation (ABMS). In the DTA Core area, I focused on researching various computational schemes to advance the computational efficiency of an SBDTA model. I developed the temporal domain decomposition method to allow large-scale SBDTA problems to be solved in a computationally efficient manner. This advancement was later integrated into DynusT and made DynusT arguably the most computationally efficient DTA model in the industry[1]. I also developed the variable time-discretization scheme to reduce the computer memory usage for the time-dependent shortest path algorithm. This scheme further conserved the memory usage for DynusT. One novel concept that I explored in the simulation is the idea of adaptive simulation, in which the simulation time resolution is not predefined but automatically tuned during run time based on the rules and needs. More detailed micro-like simulation temporal resolutions and logics could be triggered in situations where a higher fidelity is necessary. These significant advancements partly contributed to the accelerated number of citations and use of DynusT from 33 (2007-2010) to more than 240 from 2010 to 2017.

In the DTA Support area, I made a significant advancement in dynamic O-D calibration that allows both link traffic counts and speed profiles to be calibrated for a large-scale regional model in a computationally tractable manner. Being able to replicate the real-world congestion pattern in a real-life network is a crucial step for  DTA model deployment. I developed the method in early 2012 but put the methodology into a real-world stress test for five years before publishing it in 2017. This principle takes longer for research work to be published, but it ensures that the calibration method is both theoretically rigorous and practically useful. One exciting research work that I pioneered is the development of the congestion responsive pricing methodology that dynamically determined the toll rate in the express lane in response to the generalized purpose lane congestion situation while maintaining the desired operating speed in the express lane. The unique algorithmic structure allows individual users’ value-of-time to be incorporated into the assignment algorithm without increasing the computation time compared to the uniform value-of-time case. This breakthrough brings forth superb accuracy and realism in predicting volumes and revenue for the express lane facilities. After incorporating this algorithm into DynusT, as shown below, one of the practitioner users reported an excellent back-casting result comparing the model results and the actual observations after the I-70 East express lane was opened in Denver in 2016.

One major movement in transportation modeling is more practical deployments of Activity-based Models (ABM) after nearly four decades of theoretical research. DynusT was selected as the primary DTA model by FHWA/TRB SHRP2 for several high-profile research projects such as three SHRP2 C10 projects that pioneered the integration of ABM and DTA at the regional level. Since 2010 to 2017, DynusT has been integrated with several prominent ABM models such as SACSIM, OpenAMOS, and land use model UrbanSim. One significant achievement is the collaboration with Prof M Hickman and his team on the development of the first simulation-based dynamic transit assignment model FAST-TrIPS. I was part of the initial design team and guided and shaped FAST-TrIPS to achieve the seamless integration of DynusT and DTA in general. Furthermore, I proposed a new approach that exploits the DTA simulation vehicle trajectories to extract and store the zone-to-zone travel times without creating time-dependent skim tables (zone-to-zone travel time matrices) that typically take an enormous amount of computer memory to store. This theoretical approach was successfully adopted by some practitioners in the 2017 FHWA C10 projects for the Atlanta Regional Commission and Ohio Department of Transportations to attain significant computer memory saving.

All the years of my devotion to DTA related research and practical applications gradually raised the awareness of the advantages and benefits of DTA models for transportation agencies and professionals. An increasing number of DTA projects and models have been funded by and adopted by transportation agencies. However, in 2011, I started to question why transportation agencies predominantly consider supply-side strategies without investing enough thinking in how the system can be improved if the users could make small behavior changes in their daily travel decisions. For three decades, all regional planning agencies have a travel demand management (TDM) or commute solution program, but only a handful of TDM programs are considered effective. The increasingly applied congestion pricing strategy can be regarded as a ”stick” approach to penalizing people using the freeway facility during peak hours. This “stick” approach, however, does not help the drivers explore other departure times, routes, and even mode options, so I devised a method using a combination of “carrot incentives and messages to discover and engagement commuters with more diversified mobility options.

After being granted a patent for this approach in 2011, I was encouraged by the University of Arizona tech transfer office to create a university spinoff company called Metropia to roll out a technology product to the market. To avoid conflict of interests and conflict of commitment with my faculty position, I raised startup funding to build a product and business development teams. After three years of product and business development, I finally launched the Metropia Synergy Platform to Tucson as the first market in 2015 and subsequently four more markets over the next 24 months. In 2016, Metropia partnered with the Texas Department of Transportation to win the newly established US-DOT Advanced Transportation Congestion Management and Technology Deployment (ATCMTD) program. The team is one of the eight awardees receiving $9M US-DOT funding ($18M total including local match) to deploy the Metropia powered Mobility-as-a-Service&Tool (MaaS&T) platform to the Houston region from 2018-2021.

The Metropia platform is also designed to be a live experiment platform where we can set up different offerings and incentives to various types of users and observe their responses to the provided information and incentive offers. Just in 2 years, we have gained a tremendous amount of insights about commuters’ attitude and behavior towards their travel choices in the new era. Their decisions are influenced a by a new type of “habit” – desiring to know, seeking instant gratification, favoring variable rewards, caring about social impact. I am currently preparing several papers that highlight the findings from our on-going interaction activities with our users. The goal is to understand our users better and to devise more effective strategies to help our users discover and engage in mobility options by overcoming the stubborn 0ld habits.

My scholarly impacts on the research and professional communities have been broadening and deepening in the last seven years due to my consistent, high-quality research and regular DynusT release; both have helped many researchers worldwide accelerate their studies. In 2017, over 240 publications have either cited or used DynusT as the primary tool for their research. DynusT has been downloaded more than 1000 times worldwide.

My passion for my research and its real-world impacts fueled numerous service activities offered to the communities I serve. Over the ten years of DynusT’s open source release period, I provided more than 1000 hours of personal time helping users install, configure and debug their DynusT work. A large part of the effort went to providing advice and knowledge transfer of the underlying concepts of DTA. To further increase the awareness of the basic concept and benefits of DTA, I volunteered as a task leader for a group of renowned DTA researchers and practitioners in 2009 to start producing a primer for DTA. After two years of solid writing by the co-authors and detailed inspection and validation by the professional community, the quality of this document was recognized by the TRB, and the TRB decided to publish as an official TRB publication in 2011, with the TRB publicly acknowledging my contribution with a certificate of appreciation. This document has also become one of my highly cited papers.

My academic services also include reviewing more than 200 papers for TRB since 2002. I have been serving as a committee member for two TRB technical standing committees – ADB30 Transportation Network Modeling and ABR30 Emergency Evacuation. I have been serving as a reviewer for 110+ submissions for more than a dozen transportation journals including Transportation Science, Transportation Research Part A, B, C, D, E, Networks and Spatial Economics, Journal of Intelligent Transportation Systems, etc. Presently, I am also serving as an associate editor for an emerging international journal titled International Journal of Transportation Science and Technology (IJTST) managed by Elsevier. I chaired more than 20 INFORMS Transportation Systems and Logistics (TSL) Society sessions and served as an Intelligent Transportation System (ITS) special interest group (SIG) Vice Chair and Chair from 2008 to 2010. I acted as the conference chair for the 7th International Symposium on Travel Demand Management in 2015.

I also delivered 69 invited speeches (10 overseas invitations), including 11 at universities, 28 at conferences, eight at US federal governments or foreign ministries, and four at US national laboratories. I will deliver the keynote speech at the 8th International Symposium on Travel Demand Management in Taipei, Taiwan. Sixteen of these talks had the travel paid for by the host. In 2013, I was also invited by the TEDx committee to deliver a TEDx talk on the concept and essence of active demand management. This talk was viewed more than 5300 times.

The objective of my teaching is to help my students gain essential foundational knowledge of the subject matter and how they can relate the course materials to real-world applications via various hands-on activities. My research activities provided great support for this objective. As an example, I regularly taught a semester-long graduate/upper division course in DTA (CE 460/560 Special Topics), covering both theoretical aspects and practical use of DynusT. I often collaborated with the local transportation agencies such as the Pima Associations of Government and let the class students use the learned skills and tools to participate/shadow in on-going projects. One recent example is that we used DynusT to build a regional Tucson simulation model in DynusT and simulated and analyzed the impacts of 20+ new downtown development projects in the next ten years aiming to explore strategies to make downtown Tucson a more vibrant and livable community with multi-modal transportation system offerings over the next decade. This arrangement allowed my research, teaching, and services to the community to be fully integrated and synergized and delivered the maximum values and benefits to all participating parties.

Another unique integrative and synergistic initiative is the Certificate to Master Program (CMP) that I initiated in 2015, which was unprecedented in the US and finally came to fruition in 2017. The CMP program allows the partnering foreign selected universities (presently only Tongji University and Soochow University in China, both ranked within the top 25 universities in China) to allow their qualified senior-year undergraduate students to be enrolled in a CEEM ME/MS program and start taking the graduate-level courses remotely. Once they finish their BS degree and year one CEEM ME courses concurrently, they arrive at the UA Tucson campus to complete their year two courses and receive the ME degree in transportation engineering at UA. This program brings benefits to students, UA, and the professional community. Students cut their total cost by 50% compared to a regular ME program at UA or another equivalent program at other US universities.  CEEM departments enhance their graduate program enrollment with an increased influx of high-quality graduate students, and the partnering professional firms who promise to offer internship opportunities to the CMP program graduate would have first-hand access to outstanding CEEM graduates before they leave the university.

As a senior faculty once said to me “a constant challenge for all faculty researchers is to define what to be devoted to over the next ten years.” We want to spend our time in meaningful, impactful, and high funding potential areas. I am glad that my first ten years at the University of Arizona has been nothing but exciting and I enjoyed fruitful outcomes that fully integrated and synergized my research, teaching and service activities. My ongoing research emphases such as DTA, active demand management, and mobility-as-a-service&Tool will continue to fuel numerous funding and research breakthroughs in mobility decision behavior and mobility management strategies. My quest to fully explore both the demand and supply-side strategies for urban congestion and mobility management is a life-long pursuit, and I am determined to play an instrumental role in pushing the envelope regarding the scientific discoveries and engineering creations in the decades to come. 


[1] The Southern California Associations of Governments (SCAG) conducted its independent research on several prominent DTA models such as DynusT, MATSIM, and TRANSIMS and DynusT was the only model what could run the massive SCAG model with 20+M trips and 80k+ links.

  • DTA Core: MIVA – temporal domain decomposition, variable time-discretized time-dependent shortest path algorithm, adaptive mesoscopic simulation, intermodal assignment, dynamic O-D calibration, constrained time-dependent K shortest path, DTA computational speed-ups.

  • DTA Support: signal optimization, network abstraction, trajectory mining of zone-pair travel time, work zone schedule optimization, congestion-responsive congestion pricing methodology.

  • DTA Application: evacuation network flow optimization, congestion pricing.

  • DTA Integration: OpenAMOS and UrbanSim integration, FAST-TrIPS integration, SACSIM Activity-base Model (ABM) integration, CT-RAMP ABM integration.

  • Urban Mobility: incentive-based active demand management, system optimal behavior strategies, tradable mobility credits, taxi/ridesharing routing behavior, GPS data based driving behavior analysis, network partitioning

  • Driving Behavior: driving risk factors with GPS data, social media traffic report analysis, behaviorally induced system optimal strategies, Bayesian network based driving behavior.

  • Agent-based Simulation: start-up activities, enroute-modeling, dilemma zone analysis, day-to-day learning

  • Flexpress : inter-city high-speed transit demand factors.

bottom of page