Perception Enhanced Multimodal Multitask Semantic Communication and Resource Management for UAV-Assisted ISAC Systems

Recent advances in integrated sensing and communication (ISAC) unmanned aerial vehicles (UAVs) have enabled their widespread deployment in critical applications such as emergency management. Existing approaches still face challenges in energy efficiency, real-time adaptability, and service latency. To address these limitations, a perception enhanced multimodal multitask semantic communication (PE-MMSC) framework is first proposed, and a task-oriented resource optimization model is established on this foundation to maximize comprehensive quality of experience (QoE) in this paper. Specifically, PE-MMSC leverages a perception-enhanced attention mechanism to achieve semantic enhancement of critical features and more effective multimodal fusion guided by preliminary coarse classification results. For the resource optimization model, reinforcement learning is employed to address two subproblems: trajectory planning and resource management, and a semantic importance (SI) feedback mechanism is proposed to develop SI SI-guided double deep Q-network(SI-DDQN), which can make adaptive resource allocation decisions for important data with semantic awareness. Experimental results demonstrate that the proposed PE-MMSC system achieves 5%–10% higher target classification accuracy compared to conventional systems while maintaining comparable data reconstruction quality with acceptable computational overheads, and the SI-DDQN achieves nearly two to three times greater QoE improvement than traditional algorithms.