On Hybrid Heuristics for Steiner Trees on the Plane with Obstacles

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Minimal-length Steiner trees in the two-dimensional Euclidean domain are of special interest to enable the efficient coordination of multi-agent and interconnected systems. We propose an approach to compute obstacle-avoiding Steiner trees by using the hybrid between hierarchical optimization of shortest routes through sequential quadratic programming over constrained two-dimensional convex domains, and the gradient-free stochastic optimization algorithms with a convex search space. Our computational experiments involving 3,000 minimal tree planning scenarios in maps with convex and non-convex obstacles show the feasibility and the efficiency of our approach. Also, our comparative study involving relevant classes of gradient-free and nature inspired heuristics has shed light on the robustness of the selective pressure and exploitation abilities of the Dividing Rectangles (DIRECT), the Rank-based Differential Evolution (RBDE) and the Differential Evolution with Successful Parent Selection (DESPS). Our approach offers the cornerstone mechanisms to further advance towards developing efficient network optimization algorithms with flexible and scalable representations.

Original languageEnglish
Title of host publicationEvolutionary Computation in Combinatorial Optimization - 21st European Conference, EvoCOP 2021, Held as Part of EvoStar 2021, Proceedings
EditorsChristine Zarges, Sébastien Verel
PublisherSpringer Science and Business Media Deutschland GmbH
Pages120-135
Number of pages16
ISBN (Print)9783030729035
DOIs
Publication statusPublished - 2021
Event21st European Conference on Evolutionary Computation in Combinatorial Optimization, EvoCOP 2021 Held as Part of EvoStar 2021 - Virtual, Online
Duration: 2021 Apr 72021 Apr 9

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12692 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st European Conference on Evolutionary Computation in Combinatorial Optimization, EvoCOP 2021 Held as Part of EvoStar 2021
CityVirtual, Online
Period21/4/721/4/9

Keywords

  • Minimal trees
  • Optimization
  • Planning
  • Steiner trees

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'On Hybrid Heuristics for Steiner Trees on the Plane with Obstacles'. Together they form a unique fingerprint.

Cite this