Estimating Traffic Stream Density Using Connected Vehicle Data
The number of on-road vehicles has increased rapidly over the past few decades, leading to serious traffic congestion in many areas. An efficient way of solving traffic congestion is improving traffic management strategies using advanced technologies and advanced traffic signal control systems that optimize traffic signal timings in real-time. Knowing the number of vehicles on a specific roadway segment is crucial in developing efficient adaptive traffic signal controllers; however, it is difficult to measure traffic density directly in the field.
This research aims to estimate the total number of vehicles on signalized approaches using only connected vehicle (CV) data. The estimate outcomes can be provided to traffic signal controllers to optimally determine the allocation of green time for each traffic signal phase, leading to better intersection performance measures. Different estimators (filters) using CV data will be developed to estimate the total number of vehicles on signalized links, such as Kalman and particle filters. One concern with using CVs is measuring their level of market penetration (LMP). The LMP is defined as the ratio of the total number of CVs to the total number of vehicles. Providing accurate LMP estimates should improve the estimation accuracy of the vehicle counts. Therefore, in this research, a machine-learning model will be developed to provide real-time estimates of the LMP values. Then, the developed filtering model will be combined with the developed machine learning model to improve the vehicles count estimation accuracy. In addition, an adaptive filtering technique will be developed to enable real-time estimates of statistical parameters of the system noise rather than using predefined values for the entire simulation. Finally, this research will examine the impacts of traffic demand level on the estimation model, considering both under- and over-saturated conditions.
Universities and Sponsoring Organizations Involved
Virginia Tech, U.S. Department of Transportation Office of the Secretary Research
Hesham A. Rakha (VT) firstname.lastname@example.org
Hossam M. Abdelghaffar (VT) email@example.com
Funding Sources and Amounts
USDOT: $100,000 (Federal), Virginia Tech: $50,000 (Match)
May 1, 2020
Expected Completion Date
April 30, 2021
Expected Research Outcomes
- Develop a traffic stream density estimation model.
- Develop an adaptive filtering technique to enable real-time estimates of statistical parameters.
- Develop a machine-learning model to provide real-time estimates of the LMP values to improve the accuracy of the density estimation model.
- Evaluate the performance of the developed models at different LMPs.
- Document the developed models in report and in papers submitted to peer-reviewed journals.
Expected Equity Impacts and Benefits of Implementation
The developed real-time estimation model for estimating the number of vehicles on signalized approaches using CV data will be applicable in improving the design and operations of transportation systems, which helps planners develop an efficient adaptive traffic signal controller.
Estimating Traffic Density, Connected and Automated Vehicle, Filtering Techniques, Microscopic Traffic Simulation