Developing an Intelligent Connected Vehicle based Traffic State Estimator
Urban cities are growing and transport infrastructure is being hampered, resulting in congestion, delays, safety problems and increased fuel consumption. One proposed solution is intelligent transport systems that lead to better management of roads and improvements in traffic conditions. Measuring the total number of vehicles approaching an intersection is crucial for the traffic signal performance. Efficient adaptive traffic controls can be developed once accurate measurements are estimated. In this research, various estimators will be developed using the CV data to estimate the total number of vehicles on multi-lane links. Measuring the level of market penetration (LMP) is one of the main concerns for CVs use. By providing accurate LMP estimates, the accuracy of vehicle count estimates should be improved. Therefore, a deep learning model will be developed to provide the LMP values in real time. The developed estimator will then be integrated with the deep learning model developed to improve the accuracy of the vehicle estimates. This research will further study the impacts of traffic demand level, vehicles type, and initial conditions on the performance of the developed estimators.
Universities and Sponsoring Organizations Involved
U.S. DOT Office of the Secretary/Research
Funding Sources and Amounts
USDOT: $80,000 (Federal), Virginia Tech: $40,000 (Match)
January 1, 2021
Expected Completion Date
December 31, 2021
Expected Research Outcomes
Develop linear and non-linear traffic state estimators using CV data.
Develop a deep learning model to provide real-time estimates of the level of market penetration rate of CVs.
Integrate the developed estimator with the developed deep learning model.
Investigate the sensitivity of the developed estimator to a number of factors (e.g., initial conditions, LMP rate of the CVs, traffic demand level)
Expected Equity Impacts and Benefits of Implementation
This work should provide a smart real-time approach to transport agencies to estimate the number of vehicles on roads, leading to improved traffic management and operations. The proposed approach will integrate an estimator with deep learning to improve the state estimation accuracy. The estimate outcomes can be provided to traffic controllers, leading to alleviate the congestion, reduce crashes, fuel consumption, and vehicle emissions, which have a direct positive impact on health and mobility.
Traffic State Estimation, Connected and Automated Vehicles, Deep Learning, Microscopic Traffic Simulation