Using Reinforcement Learning to the Priority-Based Routing and Call Admission Control in WDM Networks

Using reinforcement learning (RL), this paper deals with the problem of call admission control (CAC) and routing in differentiating the services of Wavelength Division Multiplexing (WDM) networks to obtain maximized system revenue. The problem is formulated as a finite-state discrete-time dynamic programming problem. Here we adopt the RL method together with a decomposition approach, to solve this problem that is too complex to be solved exactly and demonstrate that it is able to earn significantly higher revenue than the alternatives.