QoE Optimisation of Video Streaming Services in RSMA Networks: A Deep Reinforcement Learning Approach : RSMA 네트워크에서 비디오 스트리밍 서비스의 QoE 최적화: 심층 강화 학습 접근법

다국어 초록 (Multilingual Abstract)

The growing popularity and market success of video streaming services need a sophisticated multiple access technique, particularly in view of the predicted surge for beyond fifth generation (B5G) and possibly sixth generation (6G) networks. To this end, rate-splitting multiple access (RSMA) has emerged as a potent contender due to its advanced interference management capabilities. In this study, we explore the application of a downlink RSMA technique in a video streaming system, where a video streaming server situated at a Base Station (BS) sends video data concurrently to users via RSMA. We also investigate enhancing the users' Quality of Experience (QoE) by focusing on two key performance metrics: latency and quality. To accomplish this, we re-formulate the optimisation problem into a Markov Decision Process (MDP) framework and propose a Deep Reinforcement Learning (DRL) approach namely Deep Deterministic Policy Gradient on RSMA-based Video streaming System (DDPG-RMAVS). Our simulation results indicate that the proposed DDPG-RMAVS algorithm outperforms extant state-of-the-art algorithms in various scenarios, signifying a significant breakthrough in video streaming optimisation.

목차 (Table of Contents)

• 1 Introduction 1
• 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
• 1.2 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
• 2 Related Works 7
• 2.1 Deep Reinforcement Learning with RSMA . . . . . . . . . . . . . 7
• 2.2 Deep Reinforcement Learning with Video Streaming . . . . . . . . 9
• 3 Problem Statement 13
• 3.1 Channel Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
• 3.2 Video traffic model . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
• 3.3 Optimisation Problem . . . . . . . . . . . . . . . . . . . . . . . . . 20
• 4 Proposed solution 23
• 4.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
• 4.2 Markov Decision Process framework . . . . . . . . . . . . . . . . . 26
• v
• 5 Performance evaluation 33
• 5.1 Simulation Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
• 5.2 Convergence analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 37
• 5.3 Performance of the Proposed Algorithm . . . . . . . . . . . . . . . 42
• 6 Drawbacks and Future works 51
• 6.1 Drawbacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
• 6.2 Future works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
• 7 Conclusion 55
• Korean Abstract 67
• Acknowledgement 69