A Value-Aware Transport Abstraction: Cross-Layer Dynamic Transmission Control for Real-Time Applications
Masato Segawa, Yuzo Taenaka, Kazuya Tsukamoto · IEEE Access · 2026
Wireless communication quality is subject to frequent fluctuations due to physical factors and network congestion, which severely degrades the performance of real-time applications such as video streaming. However, a mismatch exists between network layers: conventional applications transmit data without considering dynamically fluctuating wireless conditions, which worsens congestion, while physical and MAC layers transmit all packets uniformly without considering the importance of application data. To address this mismatch, this research proposes a “value-aware transport abstraction”—a general-purpose architecture that mitigates application quality degradation while ensuring efficient utilization of wireless resources. In this architecture, the application simply defines the purpose-specific “value” of its data, and the transport layer dynamically selects and transmits data in real-time. We implement this architecture through two approaches: Structural Value based on MPEG frame dependencies and Semantic Value derived from object detection in surveillance scenarios. By implementing this architecture on the QUIC transport protocol, the system dynamically selects transmission data—discarding low-priority frames or switching to object-only transmission—based on real-time network conditions, specifically the congestion window (CWND). Evaluation results demonstrate that the Structural Approach reduces delay by approximately 36% during congestion. Furthermore, in the Semantic Approach, under severe bandwidth constraints (10 Mbps), the proposed method reduces transmission data by approximately 94% and decreases delay from hundreds of seconds to 0.15 seconds, successfully transmitting over 97% of critical objects. Additionally, in dynamic bandwidth and cross-traffic scenarios, the method demonstrates high adaptability and coexistence capabilities, significantly reducing the impact on background traffic while maintaining real-time performance. These results demonstrate that the proposed value-aware transport abstraction effectively maintains real-time application performance under dynamic wireless conditions.