Deep reinforcement learning tames spectrum fragmentation in elastic optical network protection

Researchers have proposed a hybrid method that combines deep reinforcement learning with an improved genetic algorithm to manage pre-configured cycle (P-cycle) protection in elastic optical networks (EONs), according to a study published in Scientific Reports. The approach is designed to address two persistent problems in traditional P-cycle protection: spectrum fragmentation and excessive resource redundancy.

In the proposed scheme, the deep reinforcement learning agent learns to identify optimal base cycles, while the improved genetic algorithm refines their combination to achieve efficient, network-wide protection coverage. To further reduce spectrum fragmentation, the team introduced a link-based spectrum partitioning approach integrated with P-cycle configuration, which is intended to enable better organization of spectral resources. A routing mechanism was also designed to select working paths either by prioritizing minimal traffic load or by applying a first-fit strategy, taking both spectrum utilization and path length into account to enhance overall efficiency.

The work is grounded in a theoretical model for assessing the service accessibility of networks that use preset cycles. Compared with earlier work, the authors state that their model can predict service availability more accurately by accounting for the dependency across protection domains as well as all instances where a lightpath is accessible. Building on this availability analysis, they provide an approach for dynamic elastic optical networks that is directed toward service availability throughout the configuration cycle.

Simulation results reported in the study indicate that the suggested deep reinforcement learning method may enhance spectrum resource use while decreasing the likelihood of bandwidth blockage. The proposed approach was compared in simulations with two benchmark-cycle design techniques that utilize the protection effectiveness of p-cycles. According to the reported outcomes, the proposed method can effectively enhance service accessibility in elastic optical networks with near minimal alteration to blocking probability.

Subject of Research: Energy-aware deep reinforcement learning for P-cycle provisioning in elastic optical networks

Article Title: Intelligent P-cycle provisioning in elastic optical networks via energy-aware deep reinforcement learning

Article References: Dave, R., & Thankachan, D. (2026). Intelligent P-cycle provisioning in elastic optical networks via energy-aware deep reinforcement learning. Scientific Reports. https://doi.org/10.1038/s41598-026-74794-9

Image Credits: AI Generated

DOI: 10.1038/s41598-026-74794-9

Keywords: elastic optical networks, P-cycle protection, deep reinforcement learning, genetic algorithm, spectrum fragmentation, service availability, optical networks, bandwidth blocking, Intelligent, P-cycle, provisioning, elastic