Dynamic Transportation Network Modeling, Analysis, Simulation, and Control (DTN-MASC)

DTN-MASC is to understand the behavior and interactions of major components/players in a multi-modal transportation network system, develop mathematical paradigms and tools to model such interactions and behavior, and design efficient solution methods. The main purpose is to develop novel management policies to manage the multi-modal transportation system more effectively and efficiently. Important applications include emergency evacuation planning and modeling, congestion pricing, urban traffic control, among others. This is particularly critical now in the era of connected / automated vehicles (CAVs), electric vehicles / buses, new mobility systems (ridehailing, ridesharing, carsharing, bike/scooter sharing) , and AI / big data analytics in transportation. The topics we are currently investigating include the network effect of new mobility services, the transit first-mile/last-mile problem using new mobility services, urban traffic control with CAVs, and optimizing the charging infrastructure of electric buses.

Below are some specific examples of DTN-MAS related research topics and results iUTS has conducted in the past.

Dynamic User Equilibrium (DUE) and applications

Our iUTS team has worked on developing dynamic user equilibrium (DUE) methods, models, and algorithms for both the continuous-time problems and the discrete-time problems. By collaborating with optimization experts and mathematicians, we were able to apply a new mathematical paradigm, called differential variational inequality (DVI) and its special form of differential complementarity systems (DCS), to model DUE to capture its two major and distinct components, i.e., (i) drivers’ choice behavior (such as departure time, route, and mode choices); and (ii) the network traffic flow dynamics. The former is usually formulated as an optimization problem or an equilibrium problem, while the latter is often formulated as an ordinary differential equation (ODE) or a partial differential equation (PDE). DVI and DCS integrates these two (mathematically very different) components into one consistent and coherent mathematical framework and gives them a formal treatment in terms of solution existence, uniqueness, stability, and solution methods. In particular, we reformulated the Vickrey-type point queue model as a DCS and applied/extended the double queue model for link level traffic flow dynamics. However, DUE problems are inherently more challenging than regular DVI and DCS, mainly due to the time-delayed terms in some of the major components (such as route choice and flow propagations). This requires new developments of the standard DVI/DCS theories, which are challenging yet very interesting problems both mathematically and practically.

    

Dynamic System Optimum (DSO) and Applications

Dynamic system optimum (DSO) describes and predicts the dynamic traffic network flow from the system perspective by assuming that all drivers are fully cooperative to minimize the total cost (e.g., travel time) of the system. It was widely believed in the past that DSO is less realistic than DUE, which however could provide a benchmark to develop and compare network-wide transportation management strategies such as congestion pricing. The iUTS team has developed continuous-time DSO models that integrates the link-based double queue model. In particular, we showed the solution properties and existence of free-flow DSO models in which all drivers would wait at the origins instead of in the network. Such free-flow DSO solutions are easier to compute and may play some important role in developing future dynamic traffic network management strategies especially when connected and automated vehicles (CAVs) are widely deployed. We also applied the DSO models to network-wide emission pricing and control, and the emergency evaluation planning for Lower Manhattan of the New York City.

     

Modeling the Network Effect of Shared Mobility

Emerging shared mobility services such as e-hailing, transportation network companies (TNCs), and ridesourcing and ridesharing, are rapidly changing the way how people travel in urban areas. There are however uncertainty regarding how these new services will impact congestion, energy use, and emissions at a network level, as well as how they will compete or complement the public transportation system. The iUTS team has worked on developing network models to capture and quantify those network effects of shared mobility services, including network-level congestion and how they may cooperate with transit to better solve the first-mile and last-mile transit problems.

   

Traffic Signal Control and Optimization with Connected and Automated Vehicles

The wide deployment of CAVs may profoundly change the look of urban traffic and the way how it is controlled. There are also uncertainties about how CAV technology may evolve and deployed, as well as challenges on understanding the dynamics/interactions of CAVs and human driven vehicles (HDVs). iUTS recently investigated the modeling of CAVs and HDVs on transportation networks, developed reinforcement-learning based eco-driving methods for a single CAV to save energy, and synthesized recent CAV-based urban traffic control studies and summarized the most critical gaps for future research.

By collaborating with vehicle control experts, iUTS developed CAV-based methods to optimize traffic signal timing plans by considering the driving and fuel consumption characteristics of individual vehicles. Some key features of the study include: (i) fixed cycle length so that signal coordination can be done readily for multiple intersection; (ii) different types of vehicles (such as gasoline cars and trucks, electric vehicles, buses, etc.) with their distinct fuel consumption characteristics; (iii) an optimization model that can be approximated as a dynamic programming (DP) problem, with a two-step method to guarantee the fixed-cycle-length solution.

When all vehicles are automated vehicles, the iUTS team, via collaborations with vehicle experts from Tsinghua University, worked on developing a cooperative method for the simultaneous optimization of traffic signal timing (macro level) and vehicle control (micro level), by considering two objectives: transportation efficiency and vehicle fuel consumption. We consider transportation efficiency at the macro signal timing control level and fuel economy at the micro vehicle control level. This also implies that the primary goal of the proposed method is to ensure the efficiency of all vehicles, while at the same time to minimize vehicle fuel consumption. Such consideration helps decompose the method into two interactive components, which makes the cooperative method easier to construct and solve. The results of this research received the Best Paper Award (2nd Prize) from the IEEE Intelligent Vehicles Symposium 2017. The paper is titled “V2I Based Cooperation between Traffic Signal and Approaching Automated Vehicles”, and is one of the two papers selected from over 300 papers submitted to the Symposium.

 

Multiscale Traffic Control with Mixed CAVs and HDVs

The concept of Multiscale traffic control (MTC) is illustrated in Fig. 1, with distinct temporal and spatial scales at the vehicle, intersection, corridor, sub-network, and global network levels. Signal-vehicle coupled control (SVCC), as shown in Fig. 2, is a special case of (a two-scale) MTC, dealing with the joint optimization of intersection signal timing and the passing of its surrounding vehicles.

Fig. 1. Multiscale Traffic Control, from vehicle level (fastest temporal scale and smallest spatial scale) to global network-wide control (slowest temporal scale and largest spatial scale), with a few intermediate scales in between (intersection, corridor, sub-network); see Guo and Ban (2023).
Fig. 2. Signal-vehicle coupled control (SVCC). Two-scale coupled control of intersection signals (slower) and vehicles (faster). SVCC can be formulated as a multi-scale control problem. A model predictive control (MPC) algorithm is developed to decompose the problem into a slower-scale signal control problem and a faster-scale vehicle (platoon) control problem. Consistency schemes are defined and imposed to the MPC algorithm, which guarantee the stability of the control methods.
MPC-based solution method and Consistency between scales: Due to the multiple time scales involved in the MTC formulation, directly solving the MTC model is challenging. Model predictive control (MPC)-based approach can be applied by decomposing the multiscale problem into multiple single-scale problems. For SVCC, this includes 1) a slower-scale signal timing optimization problem to optimize for mobility (i.e., to minimize the total travel time of all vehicles), and 2) a faster vehicle (or vehicle platoon) control problem to optimize for energy use (i.e., to minimize fuel consumption or energy use). The key here is a consistency scheme that ensure state consistency between signal control and vehicle control, enabling the stability of the MPC method. Consistency for SVCC specifies that the vehicle position specified by the slower-scale signal control problem at the coarse time scale should be satisfied by the faster-vehicle vehicle control at the same coarse time scale. Given this consistency condition, the stability result can be established for the MPC method. The MPC method and the consistency definition also ensure the implementation principle of MTC to satisfy multiple key objectives: i) safety by design to include traffic signals and follow the dual-diagram signal design scheme; ii) mobility by signal timing optimization to minimize the total time of all passing vehicles of the intersection; and iii) sustainability by vehicle control to minimize the energy use of individual vehicle (or vehicle platoon).

Simulation testing results in Seattle. SVCC was first tested in simulation using real-world traffic intersections. The signalized intersection of Fairview Ave & Denny Way in downtown Seattle was selected, as shown in Fig. 3. Fig. 4 shows the vehicle trajectories when passing through the intersection under the SVCC (MTC) control method. It was observed that the control method tends to form and maintain vehicle platoons when passing the intersection. Table 1 shows more detailed testing results and comparisons (with actuated signal control) under different demand levels and symmetry settings.

Fig. 3. The signalized intersection of Fairview Ave & Denny Way in downtown Seattle.
Fig. 4 Vehicle trajectories passing the intersection, under the MTC control method. Noteworthy is that the control method tends to devise signal timing so that vehicle platoons are formed and maintained when passing the intersection.
Table 1. The experimental results with varying demand levels and demand symmetry. The improvements of most metrics over actuated signal control range from 15% to 50%

Field testing on Mcity 2.0 testbed. The SVCC model was also tested on Mcity 2.0 testbed, a mixed reality testbed, as shown below in Fig.5. Mcity 2.0 provides remote access APIs to integrate SVCC algorithms with Mcity hardware systems (CAVs, signal controllers) and digital infrastructure (digital maps, simulation models, data center) and a visualization tool (as shown in Fig. 5) to monitor the testing performance.

Fig. 5. SVCC testing on Mcity 2.0 testbed. Mcity 2.0 is a mixed reality

Further simulation testing shows the performance of SVCC under different CAV penetration levels (25%, 50%, 75%, 100%), and comparisons with two benchmark traffic control methods (fixed-time control and actuated control). Details are shown in Fig. 6.

Fig. 6. SVCC performances and comparisons (*: for number of conflicts, the time-to-collision (TIC) threshold is 3 seconds). Left: for 100% CAV penetration, multiscale SVCC control outperforms fixed-time control and actuated control in terms of mobility (queue length, waiting time, etc.), and energy (fuel consumption). Note that SVCC may produce more conflicts (related to safety) if the TIC threshold is set as 3 seconds, as shown in the figure. If TIC is 2 seconds, no conflicts will be observed for any control method. Right: when CAV penetration rate reduces from 100% to 75%, 50%, and 25%, SVCC still outperforms the benchmark control methods with slightly smaller improvements except the queue length and number of conflicts.

MTC on urban traffic networks with mixed CAVs and HDVs. As shown in Fig. 7, when extending the single-intersection SVCC method to a network of traffic signals in urban areas, information sharing is assumed among neighboring intersections. This allows the prediction of traffic from neighboring intersections, thus the ability to control each intersection in a distributed manner. Under similar consistency condition (as to the single-intersection SVCC), stability can also be established for network MTC.

Fig. 7. Illustration of MTC on an urban traffic network.

The network MTC model and algorithm were tested on multiple networks in simulation, including two hypothetical networks and a 4 by 6 downtown Seattle network shown in Fig. 8. Testing results show sizable improvements of the network MTC compared with actuated signal control by a few percentages to 25%. More details can be found in Guo and Ban (2024).

Fig. 8. A 4-by-6 network in downtown Seattle

Publications

  1. Guo, Q., Ban, X., 2024. Network Multi-scale Urban Traffic Control with Mixed Traffic Flow. Transportation Research Part B 185, 102963.
  2. Guo, Q., Ban, X., 2023. A multiscale control framework for urban traffic control with connected and automated vehicles. Transportation Research Part B 175, 102787.
  3. Zhang, M., Li, L., Ban, X., 2024. Unleashing the Two-dimensional Benefits of Connected and Automated Vehicles via Dedicated Intersections in Mixed Traffic. Transportation Research Part C 160, 104501.
  4. Lin, Y., Devasia, S., Fabien, B., Ban, X., 2023. Increasing traffic capacity of mixed traffic at signalized traffic intersections using delayed self-reinforcement. Transportation Research Part C 157, 104403.
  5. Zhang, J., Pei, H., Ban, X., Li, L., 2021. Analysis of cooperative driving strategies at road network level with macroscopic fundamental diagram, 2022. Transportation Research Part C 135, 103503.
  6. Guo, Q*, Ban, X., Aziz, H.M.A., 2021. Mixed traffic flow of human driven vehicles and connected/automated vehicles on a dynamic transportation network. Transportation Research Part C 128, 103159 (https://doi.org/10.1016/j.trc.2021.103159); To be presented at the International Symposium on Transportation and Traffic Theory (ISTTT) 2022.
  7. Yang, X., Ma, R., Yang. P., Ban, X., 2021. Link Travel Time Estimation in Double-Queue-Based Traffic Models. Journal Promet – Traffic&Transportation, accepted.
  8. Guo, Q.*, Angah, O.1, Liu, Z.2, Ban, X., 2021. Hybrid deep reinforcement learning based eco-driving for low-level connected and automated vehicles. Transportation Research Part C 124, 102980.
  9. Guo, Q.*, Ban, X., 2020. Macroscopic fundamental diagram based perimeter control considering dynamic user equilibrium. Transportation Research Part B 136, 87-109.
  10. Li, W.1, Ban, X., Zheng, J., Liu, H., Cheng, G., Li, Y., 2020. A deep learning approach for real-time traffic volume prediction at signalized intersection. Journal of Transportation Engineering, 146(8): 04020081.
  11. Li, W. *, Ban, X., 2020. Connected vehicle based traffic signal coordination. Engineering 6(12), 1463-1472.
  12. Li, W.*, Wang, J. *, Fan, R. *, Guo, Q. *, Zhang, Y. *, Siddique, N. *, Ban, X., 2020. Short-term traffic state prediction from latent structures: accuracy vs. efficiency. Transportation Research Part C 111, 72-90.
  13. Ban, X., Dessouky, M., Pang, J.S., Fan, R.*, 2019. A general equilibrium model for transportation systems with e-hailing services and flow congestion. Transportation Research Part B 129, 273-304.
  14. Di, X., Ban, X., 2019. A mixed link-node and path formulation for equilibrium of new mobility systems. Transportation Research Part B 129, 50-78.
  15. Wang, J.P.*, Huang, H.J., Ban, X., 2019. Optimal capacity allocation for high occupancy vehicle (HOV) lane in morning commute. Physica A 524, 354-361.
  16. Guo, Q.*, Li, L., Ban, X., 2019. Urban traffic signal control with connected and automated vehicles: A survey. Transportation Research Part C 101, 313-334.
  17. Wang, J.P.*, Ban, X., Huang, H.J., 2019. Dynamic ridesharing with variable-ratio charging-compensation scheme for morning commute. Transportation Research Part B 122, 390-415.
  18. Li, W.*, Ban, X., 2019. Connected vehicle based traffic signal timing optimization. IEEE Transactions on Intelligent Transportation Systems 20(12), 4354-4366.
  19. Xu, B.*, Li, S.E., Bian, Y., Li, S., Ban, X., Wang, J., Li, K., 2018. Distributed conflict-free cooperation for multiple connected vehicles at unsignalized intersections. Transportation Research Part C 93, 322-334.
  20. Xu, B.*, Ban, X., Bian, Y., Li, W.*, Wang, J., Li, K., 2018. Cooperative method of traffic signal optimization and speed control of connected vehicles at isolated intersections. IEEE Transactions on Intelligent Transportation Systems 20 (4), 1390-1403.
  21. Di., X., Ma, R., Liu, X., Ban, X., Yang, H., 2018. Network design for ridesharing user equilibrium. Transportation Research Part B 112, 230-255.
  22. Yang, X.*, Ban, X., Mitchell, J., 2018. Modeling multimodal transportation network emergency evacuation considering evacuees’ cooperative behavior. Transportation Research Part A 114(B), 380-397.
  23. Ma, R.*, Ban, X., Pang, J.S., 2018. A link-based dynamic complementarity system formulation for continuous-time dynamic user equilibria with queue spillbacks. Transportation Science 52(3), 564-592.
  24. Ji. X.F.*, Ban, X., Zhang, J., Ran, B., 2017. Subjective-utility travel time budget modeling in the stochastic traffic network assignment. Journal of Intelligent Transportation Systems, Accepted.
  25. Yang, X.*, Ban, X., Mitchell, J., 2017. Modeling multimodal transportation network emergency evacuation considering evacuees’ cooperative behavior. Transportation Research, Part A, Accepted.
  26. Ma, R.*, Ban, X., Pang, J.S., 2017. A Link-Based Dynamic Complementarity System Formulation for Continuous-time Dynamic User Equilibria with Queue Spillbacks. Transportation Science, accepted.
  27. Di, X., Liu, H, Ban, X., Yang, H., 2017. Ridersharing user equilibrium and its implications for High-Occupancy-Toll lane pricing. Transportation Research Record, accepted.
  28. Ji. X.F.*, Ban, X., Li, M., Zhang, J., Ran, B., 2017. Non-expected route choice model under risk on stochastic traffic networks. Networks and Spatial Economics, in press. DOI: 10.1007/s11067-017-9344-3.
  29. Yang, X.*, Ban, X., Ma, R.*, 2017. Mixed equilibria with common constraints on transportation networks. Networks and Spatial Economics 17(2), 547-579.
  30. Ma, R.*, Ban, X., Szeto, W.Y., 2017. Emission modeling and pricing on single-destination dynamic traffic networks. Transportation Research Part B 100, 255-283.
  31. Luo, L., Ge, Y., Zhang, F., Ban, X., 2016. Real-Time Route Diversion Control in a Model Predictive Control Framework with Multiple Objectives: Traffic Efficiency, Emission Reduction and Fuel Economy. Transportation Research Part D 48, 332-356.
  32. Di, X., Liu, H., Ban, X., 2016. Second best toll pricing within the framework of bounded rationality. Transportation Research Part B 83, 74-90.
  33. Zhao, J., Li, W.*, Wang, J., Ban, X., 2016. Dynamic Traffic Signal Timing Optimization Strategy Incorporating Various Vehicle Fuel Consumption Characteristics. IEEE Transactions on Vehicular Technology 65 (6), 3874-3887.
  34. Sánchez-Díaz, I., Holguin-Veras, J., Ban, X., 2015. A time-dependent freight tour synthesis model. Transportation Research Part B, 78, 144-168.
  35. Sun, Z.*, Ban, X., Hao, P.*, Yang, D., 2015. Trajectory-based vehicle energy/emission estimation for signalized arterials using mobile sensing data. Transportation Research Part D 34, 27-40.
  36. Ma, R.*, Ban, X., Pang, J.S., Liu, X., 2015. Approximating time delays in solving continuous-time dynamic user equilibria. Networks and Spatial Economics 15(3), 443-463.
  37. Ma, R.*, Ban, X., Pang, J.S., Liu, X., 2015. Time discretization of continuous-time dynamic network loading models. Networks and Spatial Economics 15(3), 419-441.
  38. Yushimito, W.*, Ban, X., Holguin-Veras, J., 2015. Correcting the market failure in work trips with work rescheduling: an analysis using bi-level models for the firm-workers interplay, Networks and Spatial Economics 15(3), 883-915.
  39. Ge, Y.E., Stewart, K., Sun, B., Ban, X., Zhang, S., 2014. Investigating undesired spatial and temporal boundary effects of congestion charging. Transportmetrica B: Dynamics, in press.
  40. Ma, R.*, Ban, X., Pang, J.S., 2014. Continuous-time dynamic system optimal for single-destination traffic networks with queue spillbacks. Transportation Research Part B 68, 98-122.
  41. Yushimito, W.*, Ban, X., and Holguin-Veras, J., 2014. A two stage optimization model for staggered work hours. Journal of Intelligent Transportation Systems 18(4), 410-425.
  42. Sanchez, I., Holguin-Veras, J., and Ban, X., 2014. A time-dependent freight tour synthesis model. In Proceedings of the 93rd Annual Meeting of Transportation Research Board, Washington, DC.
  43. Yushimito, W.*, Ban, X., Holguin-Veras, J., 2013. Correcting the market failure in work trips with work rescheduling: an analysis using bi-level models for the firm-workers interplay, Networks and Spatial Economics, in press. DOI: 10.1007/s11067-013-9213-7.
  44. Di, X., Liu, H., Ban, X., and Yu, J.W., 2013. One the stability of a boundedly rational day to day dynamic. Networks and Spatial Economics, in press. DOI: 10.1007/s11067-014-9233-y.
  45. Di, X., Liu, H., Pang, J.S., and Ban, X., 2013. Boundedly rational user equilibria (BRUE): Mathematical formulation and solution sets. Transportation Research Part B 57, 300-313.
  46. Ban, X., Ferris, M.C., Tang, L., and Lu, S., 2013. Risk-neutral second best toll pricing. Transportation Research Part B, 48(2), 67-87.
  47. Ban, X., Pang, J.S., Liu, X., and Ma, R.*, 2012. Continuous-time Point-Queue Models in Dynamic Network Loading. Transportation Research Part B, 46(3), 360-380.
  48. Ban, X., Pang, J.S., Liu, X., and Ma, R.*, 2012. Modeling and Solution of Continuous-Time Instantaneous Dynamic User Equilibria: A Differential Complementarity Systems Approach. Transportation Research Part B, 46(3), 389-408.
  49. Ban, X., Ferris, M., Liu, H., 2010. Numerical studies on reformulation techniques for continuous network design problems with asymmetric user equilibrium. International Journal of Operations Research and Information Systems, 1(1), 52-72.
  50. Yushimito, W.F., Ban, X., and Holguin-Veras, J., 2010. Staggered work hours: a bi-level model and the role of incentives. In Proceedings of the 3rd International Symposium on Dynamic Traffic Assignment.
  51. Ban, X., and Liu, H., 2009. A link-node discrete-time dynamic second best toll pricing model with a relaxation solution algorithm. Networks and Spatial Economics 9(2), 243-267.
  52. Ban, X., Liu, H., Ferris, M.C., and Ran, B., 2008. A link-node complementarity model and solution algorithm for dynamic user equilibria with exact flow propagations. Transportation Research, part B, 42(9), 823-842.
  53. Ban, X., and Liu, H., 2007. A link-node discrete-time dynamic second best toll pricing model with a relaxation solution algorithm. Presented at the 86th Transportation Research Board Annual Meeting and submitted for publication .
  54. Ban, X., Liu, H., and Ran, B., 2005. A link based quasi-variational inequality model for dynamic user equilibria, towards real time traffic operations. In Proceedings of the 8th IEEE International Conference on Intelligent Transportation Systems (CD-ROM).
  55. Yang, F., Liu, H., H, R., Ban, X, and Ran, B., 2003. Bi-level formulation for optimal traffic information dissemination. Transportation Research Record 1836, 21-28.
  56. Liu, H., Ban, X., Ran, B., and Mirchandani, P., 2003. Formulation and solution algorithm for fuzzy dynamic traffic assignment model. Transportation Research Record 1854, 114-123.
  57. Liu, H., Ban, X., Ran, B., and Mirchandani, P., 2002. Analytical dynamic traffic assignment model with probabilistic network and travelers’ perceptions. Transportation Research Record 1783, 125-133.

Notes: * indicates graduate students Dr. Ban has advised or visiting students he has supervised.

Close Menu