Abstract: It is difficult for UPF（Unscented Particle Filter） applying to the GPS/INS（Global Positioning System/Inertial Navigation System） integrated navigation because of its large calculation, so the method of proposal distribution generation for UPF based on global sampling is presented. Combining the latest measured values, it uses UKF to generate a proposal distribution for the particle set, eliminating the need for the UPF which applies UKF（Unscented Kalman Filter） link to each particle for circulation, eliminating the resembling step, reducing the computation of the UPF. Simulations use the pseudo range as the observations in nonlinear model which has 10-dimensional state variables. The results show that compared with PF, improved UPF has a higher accuracy in estimation, compared with UPF, it has smaller computation. Therefore it can solve the problem that it is difficult to apply to the GPS/INS integrated navigation for UPF because of its large calculation.

摘要： 针对UPF（Unscented Particle Filter）由于计算量大而难以应用于GPS/INS（Global Positioning System/Inertial Navigation System）组合导航中的问题，提出一种基于全局采样的UPF算法。结合了最新的量测值，对粒子集整体利用一次UKF（Unscented Kalman Filter）算法产生建议分布，免去了UPF中对每个粒子循环套用UKF的环节，省去了重采样步骤，减少了UPF的计算量。仿真实验采用以伪距为观测量的状态变量为10维的非线性模型，结果表明，改进的UPF与PF相比，具有更高的估计精度，与UPF相比，具有更小的计算量，能够解决UPF难以应用于GPS/INS组合导航中的问题。

关键词: 无迹粒子滤波算法（UPF）, 建议分布, 全局采样, 全球定位系统/惯性导航系统（GPS/INS）, 组合导航