基于GPU的并行遗传算法在时频差估计中的应用

互模糊函数可以估计时频差参数，但在弱信号条件下，需要大量采样点才能获得较好的估计结果，面临巨大的计算压力，现有算法大都基于遍历思想进行时频二维搜索，实时性较差。针对此问题，提出基于GPU加速的并行遗传算法进行时频差快速估计，该算法针对互模糊函数的特点，结合GPU设计高速并行的遗传进化架构，通过对适应度函数的并行化计算，选择、交叉、变异的并行化操作，提升算法的执行效率。实验表明，文章设计的GPU加速算法能够带来较大的速度提升，可以快速得到时频差估计结果.

Application of GPU Based Parallel Genetic Algorithm in TDOA and FDOA

There is a huge calculation cost to estimate TDOA and FDOA with cross ambiguity function under weak signal conditions, which needs a large number of sampling points in order to obtain better estimation results. Existing algorithms based on the ergodic theory have poor real time performance. To solve this problem, the parallel genetic algorithm on GPU is proposed based on the characteristics of cross ambiguity function. Due to the high speed parallel architecture of GPU, the fitness function computing, selection, crossover, and mutation can be operated in parallel to enhance the algorithm execution efficiency. Experimental results show that the computational efficiency of the proposed GPU accelerated algorithm is greatly improved and the estimation results of TDOA and FDOA can quickly be obtained.