基于文本语义感知的无人机轨迹与数据收集联合优化OA

With the rapid advancement of the Internet of Things(IoT),the massive generation of data has imposed higher demands on the quality of wireless communication services.Unmanned Aerial Vehicle(UAV),leveraging their high mobility,rapid deployment capability,and flexible controllability,can serve as aerial mobile edge nodes to achieve efficient data collection and processing from ground-based IoT devices.However,in low signal-to-interference-plus-noise ratio environments,UAV-assisted data collection often faces challenges such as high bit error rates and limited transmission rates.Existing approaches are predominantly based on traditional communication models,often overlook the transmission characteristics and distribution features of semantic information,making it difficult to optimize UAV trajectory and spectrum resource utilization while ensuring semantic reconstruction quality.Therefore,a textual semantic awareness data collection and trajectory optimization method named Semantic Spectral Efficiency Optimization and Trajectory Selection(SSETS)is proposed.By integrating the geographical location,data volume,and semantic information distribution,an algorithm for balancing device clustering is designed.Then,a Multi-agent Deep Reinforcement Learning(MADRL)approach is adopted,where each UAV acts as an independent agent,making intelligent flight action decisions based on local dynamic environmental information to minimize overall energy consumption and maximize the system's Semantic Spectral Efficiency(S-SE).Compared with several baseline methods,the proposed algorithm can significantly reduce system energy consumption and improve the average S-SE.