摘要
无人水面艇在动态环境下的路径规划是无人船导航的难点。在传统的Q学习算法的基础上,采用了一种新的搜索策略,该策略结合了贪婪搜索与玻尔兹曼搜索,平衡了搜索的随机性和目的性,减少了搜索陷入局部最优的可能性,同时提出了一种新的奖惩函数,综合了无人水面艇航行时导航和避障两种行为,更加符合无人水面艇实际航行环境。整个方法贴近实际情况,成功实现了水面无人艇的动态环境下路径规划。仿真模拟结果和实船航行都验证了所提出方法的有效性。
The path planning of Unmanned Surface Vehicle(USV) in dynamic environment is the difficulty of unmanned ship navigation. In this paper, a dynamic environmental path planning method for USV based on improved Q-learning algorithm is proposed. And a new search strategy is adopted, which combines greedy search with Boltzmann search, balances the randomness and purpose of search, reduces the possibility of falling into the local optimization searching, and proposes a new reward function, which combines navigation with obstacle avoidance of USV. It is more in line with the actual navigation environment of USV. The whole method is close to the actual situation and successfully realizes the path planning in the dynamic environment of USV. The simulation results and real ship navigation verify the effectiveness of the proposed method.
作者
王猛
李民强
余道洋
WANG Meng;LI Minqiang;YU Daoyang(Institute of Intelligent Machines,Chinese Academy of Sciences,Hefei 230031,China;Department of Automation,University of Science and Technology of China,Hefei 230026,China)
出处
《仪表技术》
2020年第4期17-20,36,共5页
Instrumentation Technology
基金
国家自然科学基金面上项目(61873253)
安徽省重点研究和开发计划(201904a07020056)
安徽省科技重大专项(18030801104)。
关键词
无人水面艇
动态环境
Q学习算法
路径规划
unmanned surface vehicle
dynamic environment
Q-learning algorithm
path planning
