Reproduce “Causal-Aware Intelligent QoE Optimization for VR Interaction with Adaptive Keyframe Extraction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ziru Zhang, Jiadong Yu, Danny H.K. Tsang
Researchers at The Hong Kong University of Science and Technology introduce PS-CDDPG, a Causal Reinforcement Learning algorithm for optimizing Quality of Experience in multi-user VR interactions. This method integrates adaptive keyframe extraction and attention mechanisms, reducing training iterations by approximately 30% compared to DDPG while achieving higher average QoE and improved fairness among users.
Key insights
- Causal inference in Reinforcement Learning for complex systems benefits significantly from focusing only on action-relevant state variables ("Partial State"), as attempting to infer causal influence on all state variables can hinder convergence.
- Leveraging causal influence to guide exploration in DDPG, via a noise-based active exploration strategy, improves training efficiency and helps align learned policies more closely with desired QoE outcomes.
- Intelligent joint optimization of continuous resources (bandwidth, CPU frequency) and discrete parameters (keyframe ratios) is crucial for maximizing QoE and fairness while adapting to dynamic network conditions in VR.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.