【摘要】 为了更加有效地对符号间干扰进行自适应补偿,本文提出了一种基于均方误差准则的集粒子云算法EPSA(ensemble particle swarm algorithm).在本算法中,自适应均衡器的每个抽头权值向量被看作问题空间中的一个粒子,所有权值向量对应于粒子云.根据均方误差准则,集粒子云算法以多维问题空间中飞行的粒子位置的集平均误差作为适应值,评估对应抽头权值向量的性能,从而改变粒子飞行的方向和速度,调整粒子位置的变化量.粒子在空中飞行由"互飞行"和"自飞行"构成,其中"互飞行"是通过粒子间的相互作用表征的,而"自飞行"则是通过粒子的自我调整实现的.粒子按各自的加速系数依次在不同模式中飞行,由于粒子间彼此互惠作用的影响,集粒子云算法的收敛速度得到了提高.理论分析和仿真结果均表明,本文提出的算法性能明显优于定步长或变步长的LMS算法,同时算法的复杂度几乎没有明显的增长.更多还原

【Abstract】 An effective ensemble particle swarm algorithm based on mean-squared error (MSE) criterion is proposed to compensate inter-symbol interference (ISI) adaptively. The position of each particle in the swarm, which is flying through the multidimensional problem space, is corresponding to a tap weight vector candidate. Based on MSE, each position is scored to obtain an ensemble average error as its fitness value. According to these values, the performance of the corresponding tap weight vector is evaluated and the direction and the velocity of each particle are adjusted so as to modify the increment of each particle’s position. The flying of the particles consists of two different flying patterns, mutual flying and self flying, which are distinguished by respective acceleration coefficients. Mutual flying is characterized by the interaction of the particles, while self flying is performed by each particle itself. Theoretical analysis and computer simulations prove that the new algorithm achieves better performance with no observable increase of complexity compared with the LMS method with fixed or variable step-size.更多还原