arXiv:2611.05733v1 Announce Type: new Abstract: Given a set of terminals in $\mathbb{R}^d$ and an integer $k\ge0$, the Euclidean bottleneck Steiner tree problem (EBSTP) asks for a tree spanning the terminals with at most $k$ Steiner points that minimizes its longest edge. Although the planar case has been widely studied, exact optimization methods for arbitrary dimension remain scarce. We develop two exact mixed-integer second-order cone programming (MISOCP) formulations for the EBSTP: a single-commodity flow formulation and a formulation with connectivity cuts. Both formulations share a geometric core based on edge-conditional second-order cone constraints and are strengthened by problem-specific preprocessing, edge-dependent big-$M$ constants, symmetry-breaking constraints, constructive upper bounds, and new global lower bounds on the optimal value. We establish the correctness of the formulations and prove the validity of all strengthening techniques. Computational experiments on 1050 benchmark instances in dimensions $d\in\{2,3,4\}$, with up to $50$ terminals and $9$ Steiner points, show that the single-commodity flow formulation consistently outperforms the formulation with connectivity cuts, with an advantage { in solved instances} that grows with dimension. The proposed global lower bounds substantially strengthen the root relaxation, while the single-commodity flow formulation provides the most robust exact solution method across all test sets. These results establish the single-commodity flow formulation as an effective exact approach for the EBSTP in arbitrary fixed dimension.
Continue to arXiv Mathematics →You are being forwarded in 6s — stay on this page