• ASL360 introduces AI-driven scheduler for 360° video over UAV-assisted 5G networks. • Uses deep reinforcement learning (PPO) to select base and enhancement layers, maximizing QoE. • Video encoded into dependent layers and tiled segments, enabling fine-grained download scheduling. • Dynamic cost adjustment balances video quality, buffer occupancy, and quality variation in real time. • Achieves ~2 dB higher average quality, 80% less rebuffering, 57% lower quality variation vs baselines. • Demonstrates significant QoE gains for immersive VR streaming in challenging network conditions.

Article Summaries:

  • ASL360: AI‑Enabled Adaptive Streaming of Layered 360° Video over UAV‑Assisted Wireless Networks

Researchers introduce ASL360, a deep‑reinforcement‑learning scheduler that optimizes on‑demand 360° VR video delivery in next‑generation UAV‑assisted 5G networks. The system combines a macro base station and a UAV‑mounted mm‑Wave base station to transmit layered, tile‑segmented video. Scheduling decisions are modeled as a constrained Markov decision process, solved with a Proximal Policy Optimization (PPO) agent that balances video quality, buffer occupancy, and quality variation through dynamic cost adjustment. Experimental results show ASL360 improves average video quality by ~2 dB, reduces rebuffering time by 80 %, and lowers quality variation by 57 % compared to baseline methods, demonstrating its effectiveness in dynamic network environments.

Sources: