Deep Reinforcement Learning-Based RMSA Policy Distillation for Elastic Optical Networks

Abstract

The reinforcement learning-based routing, modulation, and spectrum assignment has been regarded as an emerging paradigm for resource allocation in the elastic optical networks. One limitation is that the learning process is highly dependent on the training environment, such as the traffic pattern or the optical network topology. Therefore, re-training is required in case of network topology or traffic pattern variations, which consumes a great amount of computation power and time. To ease the requirement of re-training, we propose a policy distillation scheme, which distills knowledge from a well-trained teacher model and then transfers the knowledge to the to-be-trained student model, so that the training of the latter can be accelerated. Specifically, the teacher model is trained for one training environment (e.g., the topology and traffic pattern) and the student model is for another training environment. The simulation results indicate that our proposed method can effectively speed up the training process of the student model, and it even leads to a lower blocking probability, compared with the case that the student model is trained without knowledge distillation.

Formation of methodological approaches to increasing the efficiency of optical networks

“…The influence of this moment becomes more significant the more voluminous and wide the analyzed optical network. It should be noted that the formulas proposed by Augusta Fresnel for calculating the amplitude and intensity, refracted and reflected electromagnetic waves when passing through a flat interface between two media with different refractive indices, to a large extent, inversely correlate with the correctness and quality of installation of a fiberoptic network -the greater the installation error -the lower the validity of the results obtained [13]. Thus, it should be emphasized that the polarization for which the vector of the electric field strength of an electromagnetic wave is perpendicular to the plane of incidence is calculated using the following mathematical expressions and formulas:…”

“…In practice, this technology allows you to add up to 15 km of effective data transmission with a signal in a 100 Gbit system, overcoming its natural OSNR threshold of 40 km. The use of a more stringent approach in organizing the activity of an optical network makes it possible to further increase intermediate control procedures due to the fact that the decoder not only determines the value of the binary signal, but also guarantees the reliability of the decision made [12][13].…”

Routing, Modulation Level, and Spectrum Assignment in Elastic Optical Networks—A Serial Stage Approach with Multiple Sub-Sets of Requests Based on Integer Linear Programming

“…The inequalities (37) and (38) ensure that the slot block assigned to an incoming request s as well as an already installed s do not overlap each other.…”

“…It is worth mentioning that optimization approaches can also be classified according to the number of objective functions simultaneously optimized [34], considering (a) mono-objective approaches [5,6,8,[14][15][16][17]19,22,23,[26][27][28][29][35][36][37] and (b) multi-objective ones [13,18,20,21,24,25,33,38]. Note that the multi-objective optimization approach, taking into account the Pareto context, was reported for Routing and Spectrum Allocation (RSA) [39,40], but not for RMLSA.…”

Performance Analysis of Automatic Hidden Ligthpaths in Multi-Layer Networks

“…It is because RWA solutions do not consider the contiguous allocation constraint and different modulation formats. Therefore, new algorithms are introduced; namely, routing and spectrum allocation (RSA) algorithms [24,25]. Additionally, when an RSA algorithm considers the selection of different modulation formats, then this is called a routing modulation format and spectrum allocation (RMSA) algorithm.…”