论文部分内容阅读
阵列测向问题中的极大似然方法(ML)具有利用任意结构的阵列处理相干信号的能力,但由于其代价函数具有多个局部最优点,因此其全局优化问题一直没能彻底解决。模拟退火方法(SA)具有较好的全局收敛能力,但其速度相对于牛顿法等局部优化方法较慢。本文提出了一种解决方案,即SA进行初始化,然后再利用确定性的方法进行加速优化。利用SA寻找ML的代价函数的最优点吸引域时,各种参数的选取对计算效果影响较大,本文对此进行了分析与模拟,对参数进行了优化选取。计算机模拟结果表明,该方案是一种较为有效的折衷。
The ML algorithm in array direction finding problem has the ability to process coherent signals with an array of arbitrary structure. However, its global optimization problem has not been completely solved because its cost function has many local optimum points. The simulated annealing method (SA) has better global convergence ability, but its speed is slower than the local optimization methods such as Newton method. This paper presents a solution that SA initializes and then uses deterministic methods to speed up optimization. When using SA to find the optimal point of ML cost function, the selection of various parameters has a great influence on the computational effect. In this paper, we analyze and simulate it, and optimize the parameters. Computer simulation results show that the program is a more effective compromise.