This research focuses on traffic parameters estimation based on trajectory data considering an arterial with several signalized intersections. We develop a general framework which can combine various traffic flow models with Bayesian Network (BN) for traffic parameters estimation. The BN is formulated to establish connections between the traffic arrival process, traffic states, traffic flow model parameters and observed vehicle trajectories. More specifically, given traffic arrivals and fundamental diagram parameters (e.g., capacity, jam density, and free flow speed), vehicle trajectories are derived or simulated based on traffic flow modelling (e.g., shockwave analysis (SA), Cell Transmission Model (CTM), and the microscopic traffic simulation model VISSIM). By combining a dynamic traf...[ Read more ]

This research focuses on traffic parameters estimation based on trajectory data considering an arterial with several signalized intersections. We develop a general framework which can combine various traffic flow models with Bayesian Network (BN) for traffic parameters estimation. The BN is formulated to establish connections between the traffic arrival process, traffic states, traffic flow model parameters and observed vehicle trajectories. More specifically, given traffic arrivals and fundamental diagram parameters (e.g., capacity, jam density, and free flow speed), vehicle trajectories are derived or simulated based on traffic flow modelling (e.g., shockwave analysis (SA), Cell Transmission Model (CTM), and the microscopic traffic simulation model VISSIM). By combining a dynamic traffic flow model with Bayesian inference, we develop a framework to establish the learning process for traffic parameters estimation, such as traffic arrivals, traffic states as well as traffic flow model parameters. The expectation-maximization (EM) algorithm is introduced to solve the parameters learning process. The proposed framework is then used to formulate different estimation models by combining different traffic flow models. Specifically, we formulate the SA-BN, CTM-BN and VISSIM-BN models and analyze them accordingly. Additionally, a novel (Deep Neural Network) DNN-BN model has also been developed by replacing the CTM in the conventional CTM-BN model with the DNN. A series of numerical experiments are conducted using these models. It is shown that the performances of all these models are promising. It can be concluded that this framework has the following two features: (1) It is flexible in that any dynamic traffic flow model can be incorporated; (2) by combining the model-based approach with the data-based approach, even with a low penetration of vehicle trajectory data, good accuracy can be achieved for both estimating the traffic parameters as well as the traffic dynamics around signalized intersections along an arterial