深度学习技术在航迹控制系统中的应用  

The application of deep learning technology in trajectory control systems

作　　者：于雷 YU Lei(The Open University of Jilin,Changchun 130022,China)

机构地区：[1]吉林开放大学,吉林长春130022

出　　处：《舰船科学技术》2024年第10期174-177,共4页Ship Science and Technology

基　　金：吉林省教育科学“十四五”规划2021年度课题(GH21505)。

摘　　要：研究深度学习技术在航迹控制系统中的应用,实现航迹智能、精确控制,以适应复杂多变环境和任务需求。利用LOS算法确定船舶航行期望航向角、航迹误差,在融合船舶模型采集的船舶状态信息后,得到航迹跟踪控制的状态变量,将其输入到基于MDP模型的航迹控制器中,将最高长期累积回报作为目标,利用卷积神经网络对控制器训练,学习给定状态到执行动作之间的映射关系,以获得使船舶按照预定航迹行驶的最优舵角动作值,实现航迹精准跟踪控制。实验结果表明,该系统所用控制器经过150回合训练,即可实现航迹数据规律的精准捕捉,具有突出学习能力;干扰工况下,该系统也可使船舶沿期望航迹航行,航迹控制效果显著。Research the application of deep learning technology in trajectory control systems to achieve intelligent and precise trajectory control,in order to adapt to complex and ever-changing environments and task requirements.Using the LOS algorithm to determine the expected heading angle and trajectory error of ship navigation,after fusing the ship state information collected by the ship model,the state variables of trajectory tracking control are obtained,which are input into the trajectory controller based on the MDP model.The highest long-term cumulative return is set as the target,and the controller is trained using convolutional neural networks to learn the mapping relationship between the given state and the executed action,in order to obtain the optimal rudder angle action value that enables the ship to travel along the predete-rmined trajectory and achieve precise trajectory tracking control.The experimental results show that the controller used in the system can achieve precise capture of trajectory data patterns after 150 rounds of training,and has outstanding learning ability;Under interference conditions,the system can also enable the ship to navigate along the desired trajectory,and the trajectory control effect is significant.

关 键 词：深度学习 航迹控制 LOS算法 MDP模型 长期累积回报 卷积神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]