Optimizing mountain railway alignments with a potential field guided 3D-RRT-star algorithm
Xinjie Wan, Hao Pu, Taoran Song, Paul Schonfeld, T.Y. Yang, Lihui Peng
Abstract
Developing railway alignments becomes increasingly challenging in undulating terrain with densely-distributed obstacles. Traditional alignment optimization methods, which employ a design process characterized as “search for alignment-favorable environments based on trial alignments”, struggle to efficiently generate optimized alignments under these conditions. To address this problem, an environmental suitability analysis is first implemented by abstracting the study area as a set of voxels within various structural layers. The Environmental Suitability (ES) for each voxel is then formulated based on its location in different structural layers. Then, by considering the spatial distribution and the ES values, alignment-favorable regions and alignment-unfavorable regions are identified through a kernel density analysis and k-mean clustering. These regions are further represented by Potential Fields (PFields), which are integrated with a 3D Rapidly-exploring Random Tree-star (3D-RRT-star) to create a PField-RRT-star search method. Through application to a real-world mountain railway case, the PField-RRT-star method demonstrates improved search efficiency and solution quality.