Deep Learning and Optimization Approaches in Task Scheduling and Computing Resource Allocation for VR Video Services in Advanced 6G Networks: A Review

Abstract

The rapid advancement of sixth-generation (6G) wireless networks is expected to enable immersive applications such as Virtual Reality (VR) video services, which require ultra-low latency, high bandwidth, and efficient resource utilization. Traditional cloud-centric architectures struggle to meet these stringent requirements due to communication delays and limited scalability. As a result, integrating deep learning with optimization techniques in edge computing environments has emerged as a promising solution for efficient task scheduling and computing resource allocation. Deep learning models, particularly Deep Reinforcement Learning (DRL), have demonstrated strong capabilities in handling dynamic and complex decision-making problems. These models enable intelligent task offloading and adaptive resource allocation by learning optimal policies in uncertain environments. Recent studies show that DRL-based scheduling frameworks outperform traditional heuristic and greedy algorithms, significantly reducing task completion time and improving system efficiency. Optimization techniques such as Lyapunov optimization and convex optimization further enhance system performance by providing theoretical guarantees for stability and resource efficiency. These approaches enable balancing multiple objectives, including latency, energy consumption, and throughput. Moreover, hybrid models combining deep learning and optimization techniques have been shown to improve scalability and adaptability in edge-cloud environments. VR video services introduce unique challenges such as real-time rendering, high data transmission rates, and strict Quality of Experience (QoE) requirements. Efficient task scheduling and resource allocation are therefore critical for ensuring seamless service delivery. This review systematically analyses recent advancements in deep learning and optimization-based approaches, compares their effectiveness, and identifies key research challenges.