Hierarchical Adaptive Transmission for Point Cloud Video Streaming Based on Reinforcement Learning

With the rapid advancement of 3D acquisition and reconstruction technologies, point cloud video (PCV) has emerged as a promising immersive media format transitioning toward commercial applications. However, PCV exhibits inherent challenges for streaming transmission due to its unstructured data structure, massive data volume, and sparse/non-uniform spatial distribution characteristics. Existing viewport-prediction-driven tile-based adaptive streaming schemes suffer from three critical limitations in long-term sequence prediction: (1) unstable Quality of Experience (QoE) modeling, (2) accuracy degradation over extended prediction horizons, and (3) suboptimal long-term decision-making, which hinder stable and efficient streaming performance. To address these challenges, this study proposes a hierarchical adaptive transmission method. First, a standardized proportional mapping-based QoE assessment framework is designed to mitigate quality evaluation fluctuations caused by inter-frame structural heterogeneity. Second, a hierarchical transfer framework is established to enhance the transfer robustness under long-term viewport prediction. Finally, the HPAT(Hierarchical Policy-based Adaptive Transmission algorithm) is introduced, which combines an Actor-Critic reinforcement learning architecture to achieve hierarchical tile bitrate allocation driven by joint bandwidth and viewport information. Experimental results demonstrate that the proposed method enables more efficient utilization of network resources while ensuring optimal user experience.