Modelling the impact of traffic incidents on travel time reliability

Abstract

Assessing and prioritising cost-effective strategies to mitigate the impacts of non-recurrent congestion on major roads, due to traffic incidents, represents a significant challenge for road network managers. In addition, traffic incidents impose significant unexpected changes in the travel time of travellers.Travel Time Reliability (TTR) has become one of the most important indicators of transport performance. In this regard, the impact of traffic incidents on TTR is crucial for both operators and travellers. An extensive literature review indicated that there is a lack of relevant research in this area. To address and improve the knowledge in this area, this thesis established a four-stage logical framework to achieve the main objective of this research. The research was aimed at modelling traffic incident impacts and quantifying the effects of traffic incidents on TTR on freeways.Based on historical data from an 'integrated database', a robust methodology to identify recurrent and non-recurrent congestion and to recognise traffic incident related congestion was proposed. Further, a number of attributes relating to traffic incidents and traffic measures for both recurrent and non-recurrent congestion were extracted to quantify the impacts of traffic incidents, including incident duration and buffer time as a measure of TTR. This facilitated insight into the factors that affect these two important impacts of incidents.Extra Buffer Time was defined to calculate the extra travel time caused by traffic incidents.This reliability measure indicates how much time is required by travellers to arrive at their destination on time with 95% certainty, in the case of an incident, over and above the travel time that would have been taken under recurrent conditions. The Extra Buffer Time Index (EBTI) was calculated to show the ratio of changes in TTR in the case of an incident.A hazard-based duration modelling approach was considered to model incident duration.Parametric accelerated failure time survival models were developed to consider heterogeneity across duration in the hazard function and in the explanatory variables. In addition, a new approach was proposed to model EBTI using a Tobit model to understand the factors affecting EBTI. Moreover, this research considered both fixed and random parameters to capture unobserved heterogeneity across data. A validation process was performed to assess the level of accuracy of the results of the prediction models.ii An integrated database was established using historical data from different sources including 12 months of incident data, 18 months of traffic data, and weather data for the same period for an Australian freeway network. The results of the analysis showed that the significant variables affecting incident duration include characteristics of the incidents (severity, type, injury, medical required, etc.), infrastructure characteristics (shoulder availability), location, time of day, and traffic characteristics. Moreover, the findings revealed no significant effect...

Factorial design–machine learning approach for predicting incident durations

“…In this study, AFT incident duration survival models were established. These models were improved later by utilizing the same modeling approaches but considering more factors, including incident characteristics, infrastructure characteristics, time of day, and traffic characteristics (Hojati et al., 2014; Tavassoli et al., 2016). Additionally, they assessed the influence of traffic incidents on travel time reliability, where additional buffer time was presented as the extra delay resulting from traffic incidents.…”