Holographic viewpoint rotation prediction-based resource-efficient holographic-type communication service provision in an EON-enabled NG-RAN

Abstract

Holographic-type communication (HTC) services, driven by six degrees of freedom (6DoF)-enabled holographic viewpoints and multisensory media (e.g., visual, auditory, and olfactory) data, offer ultra-immersive realism but introduce significant challenges. Frequent and proactive user interactions in HTC exacerbate dynamic bandwidth demands and lead to redundant hologram transmissions, as users engage with only a small portion of the hologram at any time. Additionally, ultra-low latency and precise synchronization requirements across concurrent data flows carrying HTC services further complicate the resource-efficient multisensory data distribution. This paper investigates resource-efficient HTC service provision over an elastic optical network (EON)-enabled next-generation radio access network (NG-RAN), focusing on DU-CU deployment, holographic routing, and spectrum allocation for the transmission of holograms and multisensory media data. To address these issues, we propose a mixed-integer linear programming (MILP) model and a holographic viewpoint rotation prediction-based graph neural network with an edge-node switch convolution-enhanced deep reinforcement learning (VRP-GENSC-DRL) algorithm. The proposed algorithm integrates viewpoint rotation prediction (VRP) using long short-term memory (LSTM) to convert dynamic bandwidth demands into static requirements, a graph neural network (GNN) with edge-node switch convolution for precise feature extraction, and deep reinforcement learning (DRL) for optimized baseband function deployment and holographic RSA. Simulation results indicate that VRP-GENSC-DRL reduces the total number of active processing nodes and consumption of spectrum bandwidth by approximately 50% compared to benchmarks without VRP, effectively addressing the challenges of HTC service delivery.