Deep Reinforcement Learning for QoS-Constrained Resource Allocation in Multiservice Networks

Main Article Content

Juno Vitorino Saraiva
Iran Mesquita Braga Junior
Victor Farias Monteiro
Franciso Rafael Marques Lima
Tarcisio Ferreira Maciel
Walter da Cruz Freitas Junior
Francisco Rodrigo Porto Cavalcanti


In this article, we study a Radio Resource Allocation (RRA) that was formulated as a non-convex optimization problem whose main aim is to maximize the spectral efficiency subject to satisfaction guarantees in multiservice wireless systems. This problem has already been previously investigated in the literature and efficient heuristics have been proposed. However, in order to assess the performance of Machine Learning (ML) algorithms when solving optimization problems in the context of RRA, we revisit that problem and propose a solution based on a Reinforcement Learning (RL) framework. Specifically, a distributed optimization method based on multi-agent deep RL is developed, where each agent makes its decisions to find a policy by interacting with the local environment, until reaching convergence. Thus, this article focuses on an application of RL and our main proposal consists in a new deep RL based approach to jointly deal with RRA, satisfaction guarantees and Quality of Service (QoS) constraints in multiservice celular networks. Lastly, through computational simulations we compare the state-of-art solutions of the literature with our proposal and we show a near optimal performance of the latter in terms of throughput and outage rate.

Article Details

How to Cite
Saraiva, J. V., Braga Junior, I. M., Monteiro, V. F., Lima, F. R. M., Maciel, T. F., Freitas Junior, W. da C., & Cavalcanti, F. R. P. (2020). Deep Reinforcement Learning for QoS-Constrained Resource Allocation in Multiservice Networks. Journal of Communication and Information Systems, 35(1), 66–76.
Regular Papers

Most read articles by the same author(s)