2023 Volume 79 Issue 20 Article ID: 23-20011
This study proposes a traffic condition estimation method for accidents using sensing data and a traffic flow model. In traffic control, it is important to understand the traffic condition at the time of an accident. Generally, the traffic condition in a target time-space is estimated by using sensing data and a traffic flow model to estimate the complement of unobserved locations. The parameters of the traffic flow model need to be calibrated in advance. However, it is difficult to calibrate the parameters of a traffic flow model in advance because the traffic impact during an accident changes dynamically. In this study, a state-space model is constructed using vehicle detectors, probe data, and variational theory, and simultaneous estimation of parameters and traffic states is attempted. As a result of the validation, it is confirmed that the proposed method can accurately estimate the traffic state at the time of an accident.
