Semantic-Aware Self-Healing Optical Networks via AI-Driven Cross-Layer Intelligence on the Abilene Internet Backbone

Abstract

The integration of artificial intelligence, semantics-based communication, and photonics-based networking offers a revolutionary solution to optical backbone management in future optical networks. Traditional optical network management systems function based on routing algorithms at the bit level, which lack semantic awareness and fail to consider cross-layer degradation signals from the physical layer.This paper presents Semantic-Aware Self-Healing Optical Networks (SASHON), a novel AI-driven cross-layer intelligence framework for proactive fault prediction, semantic traffic prioritization, and self-healing routing in optical backbone networks. SASHON integrates three tightly coupled components: (1) a Random Forest fault prediction engine trained on eight cross-layer physical and network features achieving 98.2% prediction accuracy on the Abilene Internet Topology Zoo dataset; (2) a semantic compression module that reduces bandwidth demand by 46.1% while preserving task-critical information; and (3) a cross-layer semantic routing cost function that jointly optimizes physical-layer quality, network-layer utilization, and application-layer semantic importance. Evaluated on the Abilene 11-node US backbone topology with 160 heterogeneous traffic requests spanning five semantic classes, SASHON reduces connection blocking probability by 78.6%, improves connection acceptance ratio from 73.8% to 94.4%, reduces average link fault risk by 47.1%, and improves average SNR by 10.6% relative to conventional distance-based baseline routing. These results establish a new performance frontier for AI-assisted proactive optical network management, with direct applicability to 5G/6G transport, smart grid communications, and critical infrastructure interconnects.