An ML Approach for Crosstalk-Aware Modulation Format Selection in SDM-EONs

In space-division-multiplexed elastic optical networks (SDM-EONs), the routing, modulation, core, and spectrum assignment (RMCSA) problem is a critical lightpath resource assignment problem. Intercore-crosstalk (XT) between cores lowers the quality of parallel transmissions, and the RMCSA algorithm must ensure that XT restrictions are satisfied while optimizing network performance. There is a tradeoff between spectrum efficiency and XT tolerance - higher modulation formats are more efficient in terms of spectrum but are also less forgiving in terms of XT and allow fewer connections on adjacent cores on the overlapping spectrum. XT-aware RMCSA algorithms typically impose an upper limit or threshold on the number of lighted cores on the overlapped spectrum in order to ensure compliance with XT limits. In this paper, we offer a machine learning (ML)-assisted threshold optimization strategy that significantly enhances the performance of XT-aware RMCSA algorithms in terms of bandwidth blocking probability.