摘要
为了控制地效翼船水上降落的垂向速度,降低其入水砰击载荷,提出了一种基于深度强化学习的地效翼船降落纵向控制方法。在传统DQN算法的基础上,提出了基于终止状态评价的TSE-DQN算法。针对地效翼船降落入水垂向速度与俯仰角设计了奖励函数,并通过模拟试验将TSE-DQN算法和传统DQN算法进行了对比。试验结果表明,论文所提出的TSE-DQN算法在地效翼船的降落入水垂向速度优化方面有较大提升,且具有更好的稳定性。TSE-DQN算法对地效翼船降落时垂向速度的控制具有指导意义。
In order to control the vertical velocity of the water landing of the WIG-craft,and reduce its slamming load,a longitudinal control method of the WIG-craft landing based on deep reinforcement learning is proposed.According to the traditional DQN(Deep Q-Learning Network)algorithm,TSE-DQN(Terminal State Evaluation-DQN)algorithm based on terminal state evaluation is proposed.The reward function is designed around the vertical velocity and pitch angle of the WIG-crafts landing on water.The TSE-DQN algorithm and the traditional DQN algorithm are compared through simulation experiments.The experimental results show that the TSE-DQN algorithm proposed in this paper has a greater improvement in the vertical velocity optimization performance of the WIG-crafts,and has better stability.The TSE-DQN algorithm has guiding significance for the vertical velocity control of the landing of WIG-crafts and waterplanes.
作者
张纪
胡唤
张桂勇
张之凡
ZHANG Ji;HU Huan;ZHANG Guiyong;ZHANG Zhifan(State Key Laboratory of Structural Analysis for Industrial Equipment,School of Naval Architecture,Dalian University of Technology,Dalian 116024,China;Collaborative Innovation Center for Advanced Ship and Deep-Sea Exploration,Shanghai 200240,China)
出处
《中国造船》
EI
CSCD
北大核心
2023年第1期215-223,共9页
Shipbuilding of China
基金
国家自然科学基金项目(52061135107)
中央高校基本科研业务费项目(DUT20TD108)
辽宁省兴辽英才计划(XLYC1908027)
大连市重点领域创新团队项目(2020RT03)
关键词
地效翼船
降落纵向控制
垂向速度
深度强化学习
WIG-craft
longitudinal control of landing
vertical velocity
deep reinforcement learning