Overtaking risk modeling in two-lane two-way highway with heterogeneous traffic environment of a low-income country using naturalistic driving dataset.

• A methodological framework to define conflict risk and level of severity during overtaking maneuver using surrogate measures for two-lane two-way highway in non-lane based heterogeneous traffic environment has been outlined. • A number of different models based on the discrete choice modelling approach have been explored to assesses the overtaking conflict risk severity. • The combined role of correlation in the unobserved factors over the alternatives outcomes, fixed threshold, and unobserved heterogeneity effects of different exogenous factors across observations are assessed. • A large set of explanatory variables pertaining to overtaking maneuver in two-lane two-way highway with non-lane based heterogeneous traffic environment are examined; • Several performance measures are used to assist the performance the model evaluation task; and. • A discussion is included with the illustration of the practical applicability of the proposed overtaking maneuver risk modeling framework. Introduction: Driver behavior related to overtaking maneuvers, which are considered a major safety risk determinant on two-lane two-way highway in low- and middle-income countries (LMIC), are an important subject of further analysis. This study evaluates safety risk in terms of nature and severity of probable conflicts during overtaking maneuvers on a bi-directional undivided two-lane highway in a heterogeneous traffic environment of a low-income country. Nature and severity of probable conflicts were defined with the application of surrogate safety proximity indicators in real-world naturalistic driving environment. Method: A risk severity model for overtaking maneuver was developed to better understand the significant factors associated with the probability of conflict and its severity during overtaking maneuver using discrete choice modeling approaches. The relevance of three alternate discrete outcome frameworks, namely multinomial logit (ML), ordered probit (OP), and mixed logit (MXL) models are addressed. The best fitted model is identified and estimated. The impact of the significant attributes was also evaluated. The study collected data from a section of two-lane highway in Bangladesh using naturalistic driving from both observational and computer vision techniques. A total of 46 explanatory variables related to overtaking maneuver are assessed. Results: Speed differential between overtaking and overtaken vehicles have a significant impact on the probability of severe conflicts. Moreover, the presence of a bus as an overtaking vehicle was found to contribute significantly to the severity of conflicts. Conclusions: The study makes substantial research contributions related to overtaking behavior and safety risk evaluation during overtaking in mixed traffic environment in low-income countries. The results can be used as a proactive tool for the evaluation of overtaking maneuvers and associated safety risk, and making policy decisions reducing safety risk during overtaking maneuver as well as overall safety, while acknowledging the limited resources and facilities in low-income countries. [ABSTRACT FROM AUTHOR]