BTT导弹神经网络直接逆控制算法的FPGA实现研究

The Implementation of BTT Missile Neural Network Direct Inverse Control Algorithm Based on FPGA

针对BTT导弹控制系统中导弹模型的非线性和强耦合的特点,应用逆系统和神经网络相结合的方法建立了导弹的逆模型。并以俯仰通道为例,设计了逆控制算法的BP神经网络结构,采用VerilogHDL编写了BP神经网络各个功能模块,并将其在FPGA上实现。通过神经网络逆控制算法在FPGA硬件实现和通用计算机上软件实现的对比,表明该方案能够满足BTT导弹对控制算法运算速度的要求。文中还对硬件实现神经网络的性能进行了分析。

Regarding the nonlinear strong coupling characteristics of missile model in BTT missile control system,an inverse missile model is built by combining neural network with inverse system.Taking the pitch channel for instance,the structure of BP neural network inverse control algorithm is designed.And the Verilog HDL is used to implement each module of BP neural network and then realized in FPGA.According to the comparison of the implementation of neural network inverse control methods between FPGA and general computer software,the scheme can meet requirement of the BTT missile control algorithm for computing speed.The performance of neural network hardware implementation is also analyzed in this paper.