Reinforcement learning call control in variable capacity links

This paper defines a Reinforcement Learning (RL) approach to call control algorithms in links with variable capacity supporting multiple classes of service. The novelties of the document are the following: i) the problem is modeled as a constrained Markov Decision Process (MDP); ii) the constrained MDP is solved via a RL algorithm by using the Lagrangian approach and state aggregation. The proposed approach is capable of controlling class-level quality of service in terms of both blocking and dropping probabilities. Numerical simulations show the effectiveness of the approach.