Guaranteeing User Rates With Reinforcement Learning in 5G Radio Access Networks

Abstract

The user experience constitutes an important quality metric when delivering high-definition video services in wireless networks. Failing to provide these services within requested data rates, the user perceived quality is strongly degraded. On the radio interface, the packet scheduler is the key entity designed to satisfy the users' data rates requirements. In this chapter, a novel scheduler is proposed to guarantee the bit rate requirements for different types of services. However, the existing scheduling schemes satisfy the user rate requirements only at some extent because of their inflexibility to adapt for a variety of traffic and network conditions. In this sense, the authors propose an innovative framework able to select each time the most appropriate scheduling scheme. This framework makes use of reinforcement learning and neural network approximations to learn over time the scheduler type to be applied on each momentary state. The simulation results show the effectiveness of the proposed techniques for a variety of data rates' requirements and network conditions.