DHRA: DQN Based Deterministic Heterogeneous Resource Allocation for Multi-user Holographic Conferencing

Holographic-type communication (HTC) services, such as multi-user holographic conferencing, have emerged as a promising direction for future immersive communication. However, point cloud, a mainstream format in HTC, is characterized by its large and fluctuating data size, which poses a challenge to ultra-low latency transmission. In this paper, we propose a DQN-based deterministic heterogeneous resource allocation system for multi-user holographic conferencing (DHRA). The DHRA system introduces a Point Cloud Shaper (PCS) and a Hierarchical Queuing and Forwarding (HQF) mechanism to extract core structural points from dynamically fluctuating point cloud frames and ensure their deterministic transmission. Furthermore, a heterogeneous resource scheduling model and a Multi-user Resource Scheduling Algorithm (MRSA) are designed to maximize user access. Experimental results demonstrate that the DHRA prototype can support multi-user holographic conferencing with 2 speakers and 3 audience members while maintaining a display frame rate of over 20 FPS. Additionally, MRSA achieves an 87% scheduling success rate with 100 users, outperforming baseline algorithms by a maximum of 27.9%.