Abstract
In unmanned aerial vehicle (UAV) networks, efficient video transmission is crucial for applications such as surveillance, disaster response, and remote sensing, yet it remains challenging due to limited computational resources and energy constraints. This paper presents a novel video transmission framework that simultaneously enhances communication quality and energy efficiency in resource-constrained UAV networks. To address computational limitations while ensuring high-quality video reconstruction, we design a joint source–channel coding (JSCC) scheme with a lightweight encoder that incorporates distributed video coding. To further improve system endurance, we employ deep reinforcement learning (DRL) to dynamically optimize relay node selection. Simulation results demonstrate that the proposed lightweight feature extraction model requires only 1.81% of the complexity of deep contextual video compression (DCVC) and 16.56% of low-complexity deep video compression (L-DVC), while achieving comparable or superior reconstruction performance under diverse channel conditions. Moreover, the proposed scheme significantly improves energy efficiency, thereby extending UAV hover time compared to the baseline.
Funding Information
  • National Natural Science Foundation of China (62293481)
  • National Key Research and Development Program of China (2024YFE0200802)
  • Zhejiang Provincial Natural Science Foundation of China (LZ25F010001)