Abstract: Traditional particle filter are not used in practice mainly because they cannot satisfy real-time requirements.So this paper presents an efficient implementation model for multi-unit particle filter.In this model，the multiple units of particle data are divided into many groups，and then selected maximum value to the sampling data in each group.This method is effective in reducing re-sampling number of particles，and can improve operation efficiency.The experimental results prove that the model has better operation efficiency，higher accuracy and it easily satisfies real-time requirements.

Key words: particle filter, re-sampling, maximum value, average mean-square errors

摘要： 针对粒子滤波重采样时运算量大的问题，提出一种改进的多单元粒子滤波算法，该算法在多单元粒子滤波的理论基础上，通过加入数据采样步骤，对多个单元的粒子数据进行分组，对其采样选取各组数据的最大值，有效减少系统重采样时的粒子数，提高系统运算效率。实验结果表明：在相同的条件下，相对于多单元粒子滤波，该算法运行时间减少10倍，精度提高20倍，能够满足实时应用要求。

关键词: 粒子滤波, 重采样, 最大值, 平均均方误差