摘要
针对故障条件下高超声速飞行器的容错制导问题,提出一种基于深度学习的预测校正容错制导算法。在纵向制导律设计中,求解故障下满足配平要求的攻角剖面和升力、阻力系数;构建并训练输入端包含升力、阻力系数变化量的深度神经网络来预测落点,以避免传统预测校正制导算法中大量的积分运算;侧向制导采用基于航向角误差走廊的倾侧角反转逻辑;构造扩张状态观测器对气动参数变化量进行估计,实时输入深度神经网络。仿真结果表明,所设计的容错制导算法制导精度高、实时性好,且在故障和参数摄动条件下能实时解算出满足飞行要求的制导指令。
A deep learning-based predictor-corrector fault-tolerant guidance method is proposed for fault-tolerant guidance of hypersonic vehicles.In the design of longitudinal guidance law,the trimmable angle of attack profile and the coefficients of lift and drag in case of fault are calculated.A deep neural network which inputs contain variations of lift and drag coefficients is developed to predict landing point,thus avoiding large quantities of integral operations in traditional predictor-corrector guidance method.A bank angle reversal logic based on heading angle error corridor is designed for the lateral guidance.Extended state observer is constructed to estimate the variations of lift and drag coefficients,and the estimates are input into the deep neural network.Simulated result shows that the proposed fault-tolerant guidance algorithm has high guidance precision and excellent real-time characteristics and can calculate guidance command which meets flight requirements in real time.
作者
余跃
王宏伦
YU Yue;WANG Honglun(School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, China;Beijing Aerospace Automatic Control Institute, Beijing 100854, China;Science and Technology on Aircraft Control Laboratory, Beihang University, Beijing 100191, China)
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2020年第4期656-669,共14页
Acta Armamentarii
基金
国家自然科学基金项目(61673042)
航空科学基金项目(2014ZA51002、2018ZC51031)。
关键词
高超声速飞行器
容错制导
预测校正制导
深度学习
扩张状态观测器
hypersonic vehicle
fault-tolerant guidance
predictor-corrector guidance
deep learning
extended state observer
