基于蒙特卡罗搜索树的机动控制研究  

Flight Maneuvering Control Research Based on the Monte Carlo Search Tree

作　　者：魏文领 刘玮[1] 蒋超 WEI Wen-ling;LIU Wei;JIANG Chao(AVIC Xi'an Flight Automatic Control Research Institute,Xi'an 710065,China)

机构地区：[1]航空工业西安飞行自动控制研究所,陕西西安710065

出　　处：《测控技术》2020年第11期126-132,140,共8页Measurement & Control Technology

摘　　要：高效解算飞机机动是无人自主空战中的必备部分,也是战机完成作战任务的重要保证。利用强化学习思想,将蒙特卡罗搜索树(Metro Carlo Search Tree)算法与飞行机动控制相结合,通过反复试错解算出飞机操纵序列,同时应用了多目标优化和滚动时域优化技术,克服了传统机动动作库方法的固有缺陷。最后,以上升转弯、桶滚机动及连续蛇形S机动为例,对某型六自由度飞机模型进行控制实现,证明了所建立的MCTS机动控制算法能够灵活可靠地实现机动控制,最后通过参数摄动实验验证了算法的鲁棒性。The efficient solution is the necessary part of the unmanned air war and the important guarantee of the combat mission.The algorithm of Metro Carlo Search Tree(MCTS)is combined with the flight maneuvering control,and the sequence of aircraft control is calculated by repeated trial and error,and the method of multiobjective optimization and rolling horizon optimization is applied to overcome the inherent defects of the traditional mobile operation library method.Taking up turn,barrel roll maneuvering and continuous serpentine S maneuvering as examples,the control of a type of six degree of freedom aircraft model is carried out,which proves that the MCTS maneuvering control algorithm can realize the maneuvering control flexibly and reliably.Finally the robustness of the algorithm is verified by the parameter perturbation experiment.

关 键 词：机动飞行控制 强化学习 多目标优化 滚动时域优化 

分 类 号：V249[航空宇航科学与技术—飞行器设计] TP273[自动化与计算机技术—检测技术与自动化装置]