摘要
飞行器的自主智能避障一直是无人机领域的研究热点。相对于平面运动物体来说,飞行器的空间信息,以及对于避障的态势动作的控制更加复杂。由于飞行器在三维空间的动作的选择都是在一个连续动作空间内,所以本文提出将深度确定性策略梯度(Deep Deterministic Policy Gradient,DDPG)的强化学习方法应用在飞行器自主智能避障场景,并通过可视化仿真模拟算法效果。实验结果表明该方法能够很好地解决连续动作空间的动作选择问题,使得飞行器在障碍环境中探索出完整的避障路径,达到较好的应用效果。
The autonomous and intelligent obstacle avoidance of aircraft has always been a research hotspot in the field of UAVs.Compared with plane moving objects,the space information of the aircraft and the control of obstacle avoidance situational actions are more complicated.Since the selection of aircraft actions in the three-dimensional space is in a continuous action space,the article proposes to apply the Deep Deterministic Policy Gradient(DDPG)reinforcement learning method to the autonomous intelligent obstacle avoidance scene of the aircraft.And through visual simulation,experimental results show that this method can solve the problem of action selection in continuous action space well,enabling the aircraft to explore a complete obstacle avoidance path in an obstacle environment and achieve better application effects.
作者
张仕充
时宏伟
ZHANG Shichong;SHI Hongwei(College of Computer Science,Sichuan University,Chengdu 610065)
出处
《现代计算机》
2021年第13期80-85,共6页
Modern Computer
关键词
智能避障
强化学习
DDPG
Intelligent Obstacle Avoidance
Reinforcement Learning
DDPG