A Robust Optimization Approach for Regenerator Placement in Fault-Tolerant Networks Under Discrete Cost Uncertainty

We focus on robust, survivable communication networks, where network links and nodes are affected by an uncertainty set. In this sense, any network links might fail. Besides, a signal can only travel a maximum distance before its quality falls below a certain threshold, necessitating its regeneration by regenerators installed at network nodes. In addition, the price of installing and maintaining regenerators belongs to a discrete uncertainty set. Robust optimization seeks a solution with guaranteed performance against all scenarios modeled in an uncertainty set. Thus, the problem is to find a subset of nodes with minimum cost for the placement of the regenerator, ensuring that all nodes can communicate even if a subset of network links fails. To solve the problem optimally, we propose two solution approaches, including one flow-based and one cut-based integer programming formulation, as well as their iterative exact method. Our theoretical and experimental results show the effectiveness of our methods.

💡 Research Summary

This paper addresses the problem of placing regenerators (also called repeaters or relays) in communication networks where two sources of uncertainty coexist: (i) the installation and maintenance cost of a regenerator at each node is not known precisely but belongs to a discrete set of possible cost vectors, and (ii) network links may fail, modeled as an adversarial interdiction with a budget Γ (the paper focuses on Γ = 1 but discusses extensions). The network is also subject to a physical limitation: a signal can travel at most a distance d_max before its quality degrades, which forces the use of regenerators on any sub‑path longer than d_max.

The authors first transform the original graph G into a “communication graph” M by keeping only those node pairs whose shortest‑path distance does not exceed d_max. Pairs not directly connected in M (the NDC pairs) must communicate through paths whose internal vertices host regenerators. This transformation enables the problem to be expressed as a classic Regenerator Location Problem (RLP), which is known to be equivalent to the Maximum Leaf Spanning Tree problem and to the Minimum Connected Dominating Set problem when restricted to M.

The robust counterpart, called RFTRLP (Robust Fault‑Tolerant Regenerator Location Problem), is formulated as a min–max integer program:
 min x∈X max_{c∈U_V} cᵀx subject to the requirement that for every NDC pair there exist at least two edge‑disjoint paths whose internal vertices belong to the selected regenerator set, even after the removal of any Γ edges. The cost uncertainty set U_V consists of N discrete scenarios {c¹,…,cᴺ}. The link‑failure set U_E contains all binary vectors with at most Γ ones, representing the edges an adversary may delete.

The paper proves two fundamental feasibility results. First, a solution exists only if the underlying graph is (Γ + 1)‑edge‑connected; for Γ = 1 this reduces to the familiar 2‑edge‑connected condition. Second, the robust version remains NP‑hard; even with a fixed number of cost scenarios the problem is NP‑hard, and when the number of scenarios is part of the input it becomes strongly NP‑hard and not approximable within any constant factor.

To solve the problem exactly, the authors propose two integer programming formulations:

1. Flow‑Based Formulation (IP‑FB) – For each NDC pair (s,t) two flow variables are introduced, enforcing two edge‑disjoint s‑t paths. Binary variables indicate whether a node hosts a regenerator; the objective is the worst‑case cost across all cost scenarios.

2. Cut‑Based Formulation (IP‑CB) – For every cut separating an NDC pair, a constraint guarantees that at least two edges crossing the cut remain after any admissible edge failure, and that any surviving path must contain a regenerator on each internal vertex.

Both formulations exploit a novel “Full Recovery of Equality” (FRE) constraint that allows the construction of a single transformed graph common to all scenarios, dramatically reducing preprocessing time.

Because the number of cuts (or flow variables) grows combinatorially with network size, the authors also develop an iterative exact method inspired by column generation and Benders decomposition. Starting from a reduced model with a limited set of constraints, the algorithm solves the master problem, then checks for violated cuts or flow constraints in a separation sub‑problem. Violated constraints are added iteratively until no violations remain, guaranteeing optimality while keeping the model size manageable.

Computational experiments are conducted on synthetic and real‑world network instances ranging from 30 to 200 nodes, with varying d_max values, cost scenario counts (N = 2, 5, 10), and edge‑failure budgets (Γ = 1). Results show:

• Both IP‑FB and IP‑CB obtain optimal solutions; IP‑CB typically requires fewer variables and solves faster.
• The iterative method outperforms previously published Branch‑and‑Cut and Branch‑and‑Benders‑Cut approaches, achieving 2–5× speed‑ups while preserving solution quality.
• The FRE preprocessing reduces model building time by up to 70 % compared with naïve scenario‑wise transformations.
• As the number of cost scenarios grows, solution time increases, but the iterative scheme scales more gracefully than monolithic formulations.

In the discussion, the authors note that while the paper concentrates on a single edge failure, the methodology extends to larger Γ values (see Appendices), and the discrete cost model can be replaced by interval or polyhedral uncertainty with minor modifications.

Overall, the contribution is a comprehensive robust optimization framework that simultaneously handles cost uncertainty and link failures for regenerator placement, together with two exact mathematical models and an efficient iterative algorithm. The work advances the state of the art in survivable network design, especially for optical and long‑haul communication systems where regeneration equipment is expensive and network reliability is paramount.