距离观测方程参数估计的随机搜索法

Random Search Method for Estimating Parameters of Observation Equation

针对传统非线性模型参数估计方法(牛顿法、高斯-牛顿法、最速下降法、阻尼最小二乘法等)过度依赖初始值精度、只能获取局部最优解的缺点,利用遗传算法全局搜索能力的特点,得到距离观测方程参数估计解。最后利用短程测距数据和水下定位数据,验证了该算法的参数估计解。结果表明:该算法的收敛解优于线性平差估计解且明显提高解的精度。

Aiming at the disadvantages of the traditional nonlinear model parameter estimation methods(such as Newton method,Gauss-Newton method,maximum descent method,damped least square method,etc.),which are excessively dependent on the initial value accuracy and can only obtain the local optimal solution,this paper applied the global searching ability of genetic algorithm to obtain the parameter estimation solution.Finally,the parameter estimation solution of the algorithm was verified by using the short-range ranging data and the underwater positioning data.The results showed that the convergence solution of the algorithm was better than the linear adjustment estimation solution and the accuracy of the solution was obviously improved.