NeuroRisk: Physics-Informed Neural Optimization for Risk-Aware Traffic Engineering

2026-05-12

AI Summary

This work addresses the challenge of correlated failures in wide-area networks, which often cause severe availability degradation. Conventional traffic engineering (TE) struggles to simultaneously achieve high link utilization and stringent availability guarantees due to its reliance on excessive safety margins. To overcome this limitation, the authors propose a physics-informed deep unfolding optimizer that unifies risk-aware TE within a Sort-and-Select framework for the first time. The approach introduces a gated edge-local reservation mechanism and a permutation-invariant gradient alignment strategy, enabling efficient optimization across hundreds of probabilistic failure scenarios. By integrating physics-informed neural networks, deep unfolding, and permutation-invariant representations, the method achieves solution speeds 10²–10⁵ times faster than traditional solvers on production-scale WANs, with negligible risk optimization gaps and superior nominal throughput compared to existing neural baselines.

Abstract

In production Wide-Area Networks (WANs), correlated failures dominate availability losses, forcing operators to reserve large safety margins that leave substantial capacity underutilized. Achieving high utilization under strict availability targets therefore requires risk-aware Traffic Engineering (TE) over dozens to hundreds of probabilistic failure scenarios-yet solving this problem at operational timescales remains elusive. We demonstrate that existing risk-aware formulations can be unified under an embedded Sort-and-Select structure, exposing a fundamental trade-off between expressiveness and tractability: classical optimizers either restrict scenario selection for efficiency or incur prohibitive decomposition costs. While deep learning appears promising, prior Deep TE methods mainly target maximum link utilization and rely on scaling-based feasibility, which fundamentally breaks under explicit capacity constraints and scenario-dependent risk. We present NeuroRisk, a physics-informed deep unrolled optimizer that exploits the structure of Sort-and-Select. NeuroRisk enforces feasibility via gated edge-local reservations and represents scenario sets through permutation-invariant, gradient-aligned cues. Evaluations on production-style WANs show that NeuroRisk achieves small optimality gaps relative to the solver with orders of magnitude speedup $(10^2- 10^5 \times)$ on risk objectives, while outperforming neural baselines on nominal throughput.

Problem

• risk-aware traffic engineering
• correlated failures
• Wide-Area Networks
• probabilistic failure scenarios
• capacity underutilization

Innovation

• Physics-Informed Neural Optimization
• Risk-Aware Traffic Engineering
• Sort-and-Select Structure
• Permutation-Invariant Representation
• Deep Unrolled Optimizer