Reasearch On Uav-Assited Localization Enhancement Method In Forest Environment

In forest areas,timely,accurate,and robust localization services are essential for enhancing life security and improving emergency rescue efficiency.Unfortunately,damaged localization facilities in disaster areas and the susceptibility of GPS equipment to interference and damage make it difficult to meet localization service requirements during emergency rescue operations.One potential solution is the use of UAVs to assist in emergency localization tasks.However,due to the limited energy consumption of a single UAV and the weak localization accuracy and robustness of a single localization method,cooperation with multiple UAVs is necessary to minimize the localization errors of victims and to integrate visual information obtained by cameras to improve the accuracy and stability of signal localization.This paper focuses on the development of trajectory planning and signal localization algorithms that integrate visual information in the context of multi-UAV-assisted emergency localization in forest areas.The research results are then used to design and implement a UAV localization task management platform for use in these areas.Firstly,to provide timely and accurate localization services for victims,this paper proposes the use of multiple UAVs as aerial anchors.A two-stage multiUAV trajectory planning algorithm based on multi-agent deep reinforcement learning(MA-DRL)is proposed for cooperative localization.The algorithm optimizes the flight path of multiple UAVs by considering the forest channel,UAV energy consumption,and localization model,to minimize the localization error quickly.In the pre-fly stage,coarse-grained flight is performed by multiple UAVs to obtain the number of victims and their rough location.In the formal-fly stage,the trajectory is intelligently planned using the MA-DRL algorithm,based on prior knowledge and data obtained in the previous stage.The simulation results verify the timeliness and accuracy of the proposed algorithm.Nextly,to address the issue of limited robustness of single localization services during emergency scenarios in forest areas,this research focuses on signal localization that integrates visual information from the forest environment.Specifically,to improve localization accuracy and exclude non-line-ofsight signals received by UAVs from victims,a visual line-of-sight signal classifier based on feedforward neural networks is proposed.Additionally,a twostage fusion algorithm based on attention mechanism and attention deep neural network is proposed to fuse visual data and signal localization data and reduce the impact of noise in data.Simulation results demonstrate the effectiveness of the proposed algorithm in improving the accuracy and robustness of the localization algorithm.Finally,based on the research on UAV-assisted localization algorithms presented in the previous sections,this paper developed a UAV localization task management platform for localization tasks in forest areas.The platform was built through demand analysis,detailed design,and coding,and enables the configuration,management,and visualization operation of multi-UAV localization services.The platform demonstrate that the platform is effective in verifying the research content and enhances the digitization of emergency rescue.