The high rate of collisions on horizontal curves compared to other roadway elements makes them one of the most critical elements in a transportation network. In this regard, the current practice to design horizontal alignment needs to be reviewed.The current design guides adopt a deterministic approach to design horizontal curves considering only the driver comfort criterion and using a point mass model in the analysis. Therefore, the current design procedure suffers from several shortcomings. First, it ignores the importance of the probabilistic factors of highway operations and overlooks the variability effects of the factors included in the design. Second, adopting the comfort criterion and ignoring other design criteria, such as vehicle dynamic stability and sight distance, might lead to unsafe design in some cases. Third, using the point mass model in the analysis is questionable since it is insensitive to vehicle characteristics and the road profile. Also, the comfort thresholds used in the design are obtained from experiments conducted in the 1930s. This thesis aims at overcoming some of these issues through developing a new framework to design horizontal curves.
The new design framework adopts the reliability based-design approach and considers four design criteria in the design procedure to accommodate the driver behaviour in a quantitative manner. The naturalistic driving study database is used to examine driving behaviour on horizontal curves and to develop new models to predict the distributions of speed on curves and driver comfort thresholds. In addition, CarSIM software was used to develop models to estimate lateral friction demand with a higher level of accuracy. The first order reliability method is used to evaluate horizontal curve performance. The outputs of the reliability analysis were used as predictors in developing safety performance functions.
Reliability-based design concept is a promising concept that can be implemented in transportation engineering. This concept can provide a quantitative evaluation and can be used as a surrogate measure for safety performance. The research found that evaluation of horizontal curves performance should consider all design criteria simultaneously because changing geometric parameters to improve one of the design criterion, would affect the other design criteria.