针对粒子滤波应用于GPS/INS组合导航系统时难以保证滤波实时性的问题,提出一种基于线性/非线性结构分解的改进粒子滤波算法.改进算法对状态方程进行线性/非线性结构分解,分别采用重点采样和线性卡尔曼方式进行一步预测递推,充分发挥粒子滤波和卡尔曼滤波的特点,有效降低了粒子滤波的计算量,在保证GPS/INS组合导航系统滤波精度的条件下提高了组合滤波的实时性. Aiming at the problem that the particle filter can not guarantee the real-time performance of the filter when it is applied in GPS / INS integrated navigation system, an improved particle filter based on linear / nonlinear structural decomposition is proposed. The improved algorithm decomposes the state equation linearly / nonlinearly, One-step predictive recursion with key sampling and linear Kalman approach is used to give full play to the characteristics of particle filter and Kalman filter, which effectively reduces the computational load of particle filter and improves the combination of GPS / INS navigation system under the condition of filtering accuracy Real-time filtering.