基于最大似然估计和混合梯度优化的射手模型辨识

光纤制导过程中射手对于光电显示的响应及其控制行为将直接影响弹药的制导控制性能。针对此问题,将最大似然估计法应用于射手模型辨识研究中。为解决辨识过程中遇到的非线性优化问题，采用遗传算法和高斯-牛顿优化算法混合策略提高寻找全局最优解的概率，并使用单纯形法提高算法鲁棒性。基于交叉原理提出适用于导引头回路的精确模型，设计仿真实验平台并进行了多轮次人在回路实验，将输出误差框架下的最大似然估计方法成功应用于实验数据中。结果表明：混合梯度优化算法能够找到全局最优值，辨识模型能够准确反映射手本身的动态特性，辨识方法和辨识的射手模型对于光纤寻的弹药制导控制系统设计具有一定的实际意义。

Identification of Shooter Model Using Maximum Likelihood Estimation and Hybrid Gradient Optimization

The response of the shooter to the photoelectric display and his control behavior during fiber optical guidance have direct effect on the guidance performance of missile. The maximum likelihood estimation method is used in the identification of shooter model. For the nonlinear optimization in the identification process, a hybrid optimization strategy, which is combined of genetic algorithm and Gauss-Newton optimization, is used to increase the probability of finding the global optimal solution, and the robustness of strategy is enhanced with simplex method. An accurate model for seeker control based on crossover principle is proposed, a simulator is designed to perform multiple human-in-the-loop experiments, and the maximum likelihood estimation is successfully applied to the test data in terms of output error. The results shows that the hybrid optimization algorithm can be used to find the global optimum, and the accurate estimates of shooter model can be obtained.