A Survey of Methods and Architectures for Deep Learning with Optimization-Based Task Scheduling and Computing Resource Allocation for VR Video Services in Advanced 6G Networks

Abstract

The evolution of sixth-generation (6G) networks is expected to enable ultra-low latency, high data rates, and immersive applications such as Virtual Reality (VR) video services. These applications require efficient task scheduling and computing resource allocation to meet stringent Quality of Experience (QoE) requirements. Traditional optimization and heuristic-based approaches are inadequate in handling the highly dynamic, large-scale, and heterogeneous nature of 6G environments. This survey presents a comprehensive review of deep learning-based methods combined with optimization techniques for task scheduling and resource allocation in VR-enabled 6G networks. Deep Reinforcement Learning (DRL), Graph Neural Networks (GNNs), and hybrid AI-optimization frameworks have emerged as powerful approaches for addressing complex scheduling problems. For instance, DRL-based frameworks can dynamically optimize task offloading and resource allocation, significantly reducing latency and energy consumption in edge environments. The paper analyses recent research trends (2020–2023), including edge-cloud collaboration, multi-agent learning, and hybrid optimization frameworks. Furthermore, it highlights key challenges such as computational complexity, scalability, real-time adaptability, and data privacy. The survey concludes that hybrid AI-driven approaches integrating deep learning and optimization techniques offer the most promising solutions for next-generation VR services in 6G networks. Future research directions include lightweight AI models, federated learning, and intelligent edge computing frameworks.