基于神经网络的再入飞行器可达域生成方法

Generation of Entry Vehicle Reachable Domain Based on Neural Network

针对再入飞行器可达域计算实时性需求,提出了一种换极坐标系下的神经网络模型求解方案。参数化设计飞行倾侧角和攻角剖面,在换极坐标系下构建弹道落点快速预示神经网络模型,提高神经网络模型任务适应能力,采用贝叶斯正则化方法训练模型获得网络权重。基于弹道落点快速预示模型快速获得落点集,采用直线边界和椭圆边界近似方法拟合获得再入飞行器可达域。仿真结果验证表明本文提出的方法实时性和适应性强,能够满足在线任务规划的需求。

In order to meet the real-time requirements of the entry domain generation of hypersonic vehicles,a neural network model solution scheme in a pole-changing coordinate system is proposed in this paper.The bank angle profile and attack angle profile were parameterized,and the rapid prediction neural network model of the landing point was constructed in the pole-changing coordinate system to improve the adaptability of the neural network model.The Bayesian regularization method was used to obtain the network weight.Based on the rapid prediction model of ballistic landing points,the reentry vehicle reachable domain was obtained by using linear boundary and elliptic boundary approximation methods.Simulation results show that the proposed method has strong real-time performance and adaptability and enables online task planning.