01660962

Component

The “time to normal” is an essential measure of an effective incident management program. Yet, it is not readily available, because typically incident reports only record the time when the crash is cleared. Also, due to queue spillback, upstream traffic may take longer to return to normal conditions than the traffic at the crash site. Additional complications arise when the crash occurs under congested conditions, in which case the crash-induced congestion must be separated from other events’ impacts. To address these concerns, a dynamic approach based on the Kalman Filter algorithm is developed to estimate the crash impact duration using stationary sensor data and incident logs. The proposed method combines real-time conditions and historical patterns to establish travelers’ anticipated travel speeds under incident-free conditions and then employ the difference-in-speed approach to quantify the temporal and spatial extent of the crash impact. Later, the Elastic Net regularization technique is adopted to identify the most influential factors that affect impact duration. The results indicate that speed reductions on the first two upstream segments are the most important indicators of impact duration. In addition, weather and collision type also have significant impacts. The Elastic Net-based duration model outperforms other models including ridge, least absolute shrinkage and selection operator, OLS, decision tree, support vector machine, and neural network. The approach and the results developed in this study can be used to more accurately estimate the impact of crashes and aid in the decision-making process for upgrades that reduce these impacts.

This paper was sponsored by TRB committee AHB20 Standing Committee on Freeway Operations.

18-06127

English

8p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; References; Tables

Crashes; Incident management; Traffic congestion

Data and Information Technology; Highways; Operations and Traffic Management

Transportation Research Board Annual Meeting 2018 Paper #18-06127

TRIS, TRB, ATRI

Jan 8 2018 11:35AM