粒子滤波算法在非线性系统中得到了广泛的应用，其精度取决于目标概率函数和重要性函数是否相近，并且样本退化问题也影响了算法的性能。针对粒子滤波算法中样本退化的问题，本文提出一种基于奇异值分解的粒子滤波算法。该算法通过使用奇异值分解方法得到的重要性概率密度函数更接近于目标概率分布，降低样本退化的影响，提高了滤波器的精度，然后在列车组合定位系统的数学模型中应用该算法进行仿真实验。 -Particle filter algorithm for nonlinear systems is widely used, its accuracy depends on whether the target probability function and importance of the functions are similar, and sample degradation also affects the performance of the algorithm. For particle filter algorithm sample degradation problem, this paper proposes a particle filter algorithm based on singular value decomposition. The algorithm by using singular value decomposition of the importance of the probability density function method to get closer to the target probability distribution, reduce sample degradation, improve the accuracy of the filter, and then apply the algorithm in combination mathematical model train positioning system simulation experiments. The simulation