Vertical grades and vertical curvature significantly influence traffic safety. However, obtaining accurate and large-scale data on roadway vertical alignment remains a major challenge. This paper presents a cost-effective and efficient method for estimating roadway vertical alignment using publicly available aerial LiDAR data provided by the United States Geological Survey. An Artificial Neural Network (ANN) model was proposed to predict whether a LiDAR point belongs to a vertical curve or a tangent segment. Due to the limited availability of actual roadway vertical alignment data and the substantial data requirements of machine learning models, a synthetic training dataset was generated by systematically varying road grades and segment lengths to represent realistic combinations of tangents, crest and sag curves. This approach ensured that the model was exposed to a wide range of geometric configurations and allowed it to learn generalized relationships between vertical alignment features and their corresponding geometric parameters. The model was then independently evaluated by comparing the vertical alignment estimated from the extracted aerial LiDAR data for two-lane two-way rural roadways, Route 152 in New Jersey and Route 299 in California, with their corresponding actual vertical alignment data. In addition, a case study was conducted on another rural two-lane highway in which the model was used to compute safe speeds for each roadway segment. The resulting speeds were then compared with the posted speed limits along the corridor. The satisfactory estimation results of this study indicate that the proposed approach can be used for conducting large-scale analyses to estimate vertical alignment using publicly available LiDAR data.
Jami et al. (Fri,) studied this question.