Evaluating Roadway Safety Improvement in a Traffic Assignment Framework

In this study, a traffic equilibrium analysis conceptual framework is proposed along with related modeling components pertaining to stochastic capacity due to probabilistic incidents, with a specific focus on crashes. This paper deviates from empirical, regression-based methods to predict crashes. Crash probabilities are first estimated with a conceptual logit model, and the commuters in the network are categorized into two classes: (1) travelers with some degree of knowledge of average travel time and incident conditions across different days, and (2) travelers with access to perfect pre-trip traffic information every day. Within a gap function framework for describing the user equilibrium under different information availability, a static traffic assignment framework is first presented to describe the route choice behavior of the perfect information (PI) and expected travel cost (ETC) user classes under stochastic day-dependent travel time under probabilistic incident traffic conditions. The model is applied to a simple corridor to illustrate the effectiveness of varying safety improvement strategies. A more realistic dynamic traffic assignment framework is further proposed for incorporating safety prediction for large-scale networks with multiplier user classes.