A spanning-tree-based genetic algorithm for some instances of the rectilinear Steiner problem with obstacles

  • Authors:
  • Rita M. Hare;Bryant A. Julstrom

  • Affiliations:
  • St. Cloud State University, St. Cloud, MN;St. Cloud State University, St. Cloud, MN

  • Venue:
  • Proceedings of the 2003 ACM symposium on Applied computing
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Given sets of points and obstacles in the plane, the rectilinear Steiner problem with obstacles seeks to connect the points with a rectilinear Steiner tree---a tree made up of vertical and horizontal line segments---that avoids the obstacles and has minimum total length. We consider only rectangular obstacles and further restrict the problem by requiring that it be possible to connect every point to the tree via exactly one vertical and one horizontal segment. Rectilinear Steiner trees that conform to this restriction can be represented by spanning trees augmented to specify the rectilinear segments. A genetic algorithm that uses a spanning-tree-based coding of rectilinear Steiner trees outperforms a greedy heuristic on 45 instances of the problem, of up to 469 points and 325 obstacles. However, the coding cannot represent arbitrary rectilinear Steiner trees, so it cannot address the unrestricted case, and in the case considered here, it leaves some potentially shorter trees unexamined.