The paper presents a Q-learning based dynamic routing algorithm for C+L band elastic optical networks (EONs) considering fiber impairments such as cross-phase modulation (XPM), self-phase modulation (SPM), amplified spontaneous emission (ASE), and inter-channel stimulated Raman scattering (ISRS). The effect of fragmentation is considered in the Q-learning process in addition to considering constraints related to spectrum continuity, contiguity, and non-overlapping. Three classical spectrum allocation strategies, first-fit, last-fit, and exact-fit are used after the Q-learning routing algorithm. The proposed routing, modulation, and spectrum allocation (RMSA) approach is shown to have a lower blocking probability compared with using K-shortest path routing combined with the three classical spectrum allocation strategies.

本文提出了一种基于 Q-learning 的 C+L 波段弹性光网络 (EON) 的动态路由算法，该算法考虑了诸如交叉相位调制 (XPM)、自相位调制 (SPM)、放大自发发射 (ASE) 等光纤损伤，和通道间受激拉曼散射（ISRS）。除了考虑与频谱连续性、邻接性和非重叠相关的约束之外，在 Q 学习过程中还考虑了碎片的影响。在 Q-learning 路由算法之后使用了三种经典的频谱分配策略，首次拟合、最后拟合和精确拟合。与使用结合三种经典频谱分配策略的 K 最短路径路由相比，所提出的路由、调制和频谱分配 (RMSA) 方法被证明具有较低的阻塞概率。