用于最大似然2-DOA估计的裂变粒子群算法

FPSO algorithm for the ML estimation of 2-DOA

最大似然2-DOA估计是一种抗干扰能力强、精确度高的测向算法。对其目标函数求解时容易收敛到局部最优解,最终影响最大似然2-DOA估计的性能。考虑到粒子群(PSO)算法具有全局化寻优能力及计算量巨大的特点,本文提出了一种裂变粒子群(FPSO)算法,在保证全局化寻优能力的同时大大降低计算量。但当2-DOA估计环境恶劣时FPSO算法仍然需要较大的迭代次数才能达到一定的2-DOA估计精度。为了进一步降低计算量,提出了改进的FPSO算法。其采用AP算法初始化全局最优粒子,精确了初始化精度从而降低精确2-DOA估计所需的迭代次数。仿真证明,FP-SO算法和改进的FPSO算法能够分别在AWGN信道和短波信道下以较低的计算量实现精确的2-DOA估计,其均方根误差逼近其CRB。

the Maximum Likelihood 2-DOA estimation is a kind of method which is high accurate and applicable in situation with coherent signals and low SNR. The root of its ML function is difficult to get, which is likely to converge to local optimized root. Allo- wing for that PSO method has characteristics of converging to global optimized root and enormous computational complexity, a new meth- od named FPSO is proposed in this paper, which greatly lower the computational complexity and keep the property of global conver- gence. However, if FPSO algorithm wants to arrive at highly 2-DOA estimation accurate when signals suffer form muhi-path effect, it needs to larger its iteration number. In order to get much lower computation, a modified FPSO method is put forward. It costs lower computation in above bad condition because of the more accurate initialization of global optimum particle get from AP algorithm. Simula- tions prove that FPSO and modified FPSO algorithm could accomplish accurate 2-DOA estimation at a cost of low computational complex- ity respectively in AWGN and HF channel. Meanwhile, their RMSE approximate each CRB with increase of SNR mid the number of iteration.