Estimation of incident delay and its variability in freeway networks

Date of Award




Degree Name

Doctor of Philosophy (Ph.D.)


Civil and Architectural Engineering

First Committee Member

Chang-Jen Lan - Committee Chair


Estimation and prediction of the delay due to incidents in freeway networks is an important task for efficient incident management. Traditional deterministic delay estimation methods can not account for the stochastic attributes of dynamic traffic network. Modeling the major influencing factors in delay estimation as random variables, which include incident duration, remaining capacity and traffic flow rate, can better explain the stochastic characteristics of dynamic networks. Separate stochastic incident delay estimation models for the following purposes are presented: (1) transportation system performance evaluation, which estimates the average delay, the standard deviation of delay and the expected total delay due to one isolated incident; and (2) real-time incident management, which estimates the arrival-time dependent delay and its variability due to an isolated incident. It is found that deterministic incident delay model based on deterministic queuing theory can estimate the average incident delay comparably with the stochastic delay model, but it seriously underestimates the variability of incident delay, the expected total delay, and the variability of arrival-time dependent delay due to negligence of the variability of remaining capacity and traffic flow rate in stochastic network.In the meantime, hazard-based incident duration models are proposed to identify the important explanatory factors for incident duration, among which truck involvement plays a significant role. The Weibull model with heterogeneity performs the best. Two statistical models for estimating remaining capacity are presented based on Normal and Beta distributional assumptions. Both models have high accuracy predictability. The Normal model performs slightly better in terms of accuracy, while the Beta model is more efficient. The Beta model also produce more reasonable prediction bounds and is therefore considered a better estimator overall.Finally, the effects of incident duration uncertainty on diversion strategies and signal timing plans during incidents are investigated. It is concluded that from the perspective of system total travel time minimization, as incident duration and/or traffic demand increases, the optimal diversion rate increases; as the uncertainty in incident duration increases, the optimal diversion rate decreases.


Engineering, Civil

Link to Full Text


Link to Full Text