Accelerating Network Slice Embedding with Reinforcement Learning

Network slicing plays an important role in next-generation communication systems (5G and beyond). With network slicing, mobile network operator is able to create various logical networks (slices) that are tailored to support specific services. One of the most challenging problems in network slicing is how to share the physical resources across multiple slices in an efficient way. This paper proposes some heuristics based on reinforcement learning to effectively solve this sharing problem. Simulations have been performed to evaluate the proposed methods, in comparison with exact solutions obtained from solving an integer linear program. Numerical results demonstrate the effectiveness of the proposed learning-based algorithms in solving the problem of network slice embedding.