带落角约束的均值聚类神经网络滑模制导律研究

Sliding Mode Guidance Law of Means Clustering Neural Network with Falling Angle Constraint

为提高电视末制导炮弹对地面目标的命中精度,减小炮弹的落角误差,同时实现以期望落角命中目标,在分析带落角约束滑模制导律特点的基础上,针对RBF(Radial Basis Function)神经网络滑模制导律难以以期望落角命中目标的不足,提出了一种结合均值聚类与RBF神经网络的滑模制导律,使得神经网络在学习过程中能根据炮弹的实时飞行状态不断调整聚类中心,使中心值始终是目前飞行状态下的最优解,实现制导律的优化。对静目标与动目标算例的数值仿真表明:相比带落角约束的滑模控制,RBF神经网络滑模控制、均值聚类RBF神经网络滑模控制由于均值聚类的加入,求得的切换项增益能使炮弹以期望落角命中目标,且具有较强的鲁棒性。

In order to improve the hit accuracy of TV terminal guidance projectile to the ground target reduce the falling angle error of the projectile and realize the goal of hitting the target at a desired falling angle a sliding mode guidance law combining means clustering with RBF neural network is proposed for which an analysis is made to characteristics of sliding mode guidance law with falling angle constraint and the shortage of Radial Basis Function(RBF)neural network sliding mode guidance law that is difficult to hit the target at the desired falling angle is taken into consideration.The new guidance law enables the neural network to adjust the clustering center and center value continuously according to the real-time flight state of the projectile in the learning process and make the center value be always the optimal solution in the current flight state thus to realize the optimization of guidance law.Numerical simulation is made to static target and moving target.The results show that compared with sliding mode control with falling angle constraint and RBF neural network sliding mode control the proposed guidance law can make the projectile hit the target at the desired falling angle and is more robust.