Estimating Roadway Horizontal Alignment from Geographic Information Systems Data: An Artificial Neural Network–Based Approach

Abstract

Estimating horizontal alignment using discretized roadway data points, such as GIS maps, is complicated because the number of curved and tangent segments and their start and end points are not known a priori. This study proposes a two-step approach: The first step estimates the number and type of segments and their start and end points using an artificial neural network (ANN)-based approach. The second step estimates the segment-related attributes such as radii and length by circular curve-fitting. The novelty of this study lies in the simplicity of the input vector to the ANN model, which contains only the latitude and longitude readings of a point and those of its neighboring points. Training and test data were comprised of points extracted from curved and tangent segments of random horizontal alignments, generated synthetically using a computer programming code. The proposed approach was evaluated and compared with other available methods presented in the literature using real roadway horizontal alignment data from one freeway and one rural roadway with a total length of 47 km and 65 curved segments. The analysis results indicated that the proposed approach outperforms other approaches in terms of estimation performance, particularly when the roadway follows a winding alignment.