云南师范大学学报（自然科学版）2025，Vol.45Issue(3)：43-48,6.DOI:10.7699/j.ynnu.ns-2025-031
一种基于EDDPG算法的无人机自主导航方法研究
Research on an Autonomous Navigation Method for Unmanned Aerial Vehicles Based on EDDPG Algorithm
摘要
Abstract
A self navigation framework based on enhanced depth deterministic policy gradient(EDDPG)algorithm is proposed to address the issues of low efficiency and insufficient safety in autonomous nav-igation and obstacle avoidance of drones in complex dynamic environments.Firstly,the 3D path plan-ning problem is transformed into a collaborative optimization task of multi-source perception and rein-forcement learning.By integrating LiDAR point cloud data,ontology motion parameters,and target o-rientation information,a 34 dimensional state space is constructed to comprehensively characterize the dynamic characteristics of the environment.On this basis,a dual critic network architecture is designed to suppress value estimation bias,combined with a layered experience replay mechanism to improve key sample utilization,and an adaptive noise injection strategy is introduced to dynamically adjust the balance between exploration and utilization.The experimental results show that the training perform-ance parameters and flight performance indicators of EDDPG algorithm are significantly improved compared to other algorithms.关键词
无人机/自主导航/路径规划/LiDARKey words
Autonomous navigation/Unmanned aerial vehicles/LiDAR分类
信息技术与安全科学引用本文复制引用
张明航,韩翃,杨依凡..一种基于EDDPG算法的无人机自主导航方法研究[J].云南师范大学学报（自然科学版）,2025,45(3):43-48,6.基金项目
国家自然科学基金资助项目(U1802257). （U1802257）
