基于贪婪DDPG的无人机智能避障算法  

UAV Intelligent obstacle avoidance algorithm based on greedy DDPG

作　　者：展望晨 郭乐江[1] 许世佳 刘重阳[1] ZHAN Wangchen;GUO Lejiang;XU Shijia;LIU Chongyang(Air Force EarlyWarning Academy,Wuhan 430019,China)

机构地区：[1]空军预警学院,武汉430019

出　　处：《空天预警研究学报》2024年第5期342-346,共5页JOURNAL OF AIR & SPACE EARLY WARNING RESEARCH

基　　金：湖北省自然科学基金项目(2023AFB1028);2024年空军军事理论课题(24KJ3C1-0090R);2024年全军军事理论课题。

摘　　要：针对传统无人机DDPG避障算法难以解决无人机在未知条件或复杂环境中飞行时的智能避障问题,提出一种基于贪婪DDPG无人机智能避障算法.在传统DDPG算法基础上,通过引入动态调整贪婪度方法和高斯噪声策略,使无人机在探索环境和制定避障策略时,能够更加高效地平衡探索和利用的关系,从而提高学习效率和避障性能.实验结果表明,贪婪DDPG算法在训练效率和泛化能力上均优于传统DDPG算法,显示出良好的鲁棒性.In response to the difficulty of traditional UAV DDPG obstacle avoidance algorithm in addressing intelligent obstacle avoidance for UAVs flying in unknown conditions or complex environments,this paper proposes a greedy DDPG-based intelligent obstacle avoidance algorithm.Based on the traditional DDPG algorithm,the dynamic greedy degree adjustment method and Gaussian noise strategy are introduced to enable UAVs to balance the relationship between exploration and utilization more efficiently in time of exploring the environment and developing obstacle avoidance strategies,so as to improve the learning efficiency and obstacle avoidance performance.Experimental results demonstrate that the greedy DDPG algorithm outperforms the traditional DDPG algorithm in terms of training efficiency and generalization ability,showing good robustness.

关 键 词：无人机避障 贪婪DDPG算法 探索策略 

分 类 号：V279[航空宇航科学与技术—飞行器设计] TP183[自动化与计算机技术—控制理论与控制工程]