On Hybrid Heuristics for Steiner Trees on the Plane with Obstacles

研究成果: Conference contribution

2 被引用数 (Scopus)

抄録

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.

本文言語English
ホスト出版物のタイトルEvolutionary Computation in Combinatorial Optimization - 21st European Conference, EvoCOP 2021, Held as Part of EvoStar 2021, Proceedings
編集者Christine Zarges, Sébastien Verel
出版社Springer Science and Business Media Deutschland GmbH
ページ120-135
ページ数16
ISBN(印刷版)9783030729035
DOI
出版ステータスPublished - 2021
イベント21st European Conference on Evolutionary Computation in Combinatorial Optimization, EvoCOP 2021 Held as Part of EvoStar 2021 - Virtual, Online
継続期間: 2021 4 72021 4 9

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
12692 LNCS
ISSN(印刷版)0302-9743
ISSN(電子版)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

ASJC Scopus subject areas

  • 理論的コンピュータサイエンス
  • コンピュータ サイエンス(全般)

フィンガープリント

「On Hybrid Heuristics for Steiner Trees on the Plane with Obstacles」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル