自适应高斯粒子滤波在导航中的应用

针对粒子滤波中重要性密度函数难以选取和粒子退化导致的计算精度下降的问题,提出一种新的自适应高斯粒子滤波算法。通过高斯混合密度函数和UT变换来获取状态均值和协方差阵,选择并计算合适的自适应因子来调节均值和方差,在迭代过程中可动态调节重要性密度函数,并用WEM和EM步骤代替重采样,上述滤波算法考虑了最新量测信息的影响,使滤波性能明显改善,能更好地解决非线性非高斯系统模型的抗干扰问题。将提出的算法应用于SINS/GPS组合导航系统跑车试验中,结果表明上述滤波算法能提高导航解算的精度,其性能明显优于已有滤波,同时验证了当系统出现噪声干扰突然变化时提出算法的有效性。

Adaptive Gaussian Particle Filtering for Navigation

It is difficult to select the important density function in particle filter and the computational accuracy decreases due to particle degradation. To solve the problem, this paper presents a new adaptive Gaussian particle filtering algorithm by adopting Gaussian mixture density function and UT transformation to obtain state estimation and covariance, and select suitable adaptive factors to adaptively regulate the estimation and covariance. So the importance density function can be dynamically adjusted in the iteration process. Then it can replace the resampling with the WEM and EM steps. Thus the latest measurement influence was taken account and the particle diversity precision was improved by solving the resistance to interference based on nonlinear and non-Gaussian models. The proposed algorithm was applied to the position error calculation in the SINS/GPS integrated navigation system sports car tests. The filtering algorithm can improve the accuracy of the navigation solution and its performance is obviously better than the existing filtering. At the same time, the effectiveness of the proposed algorithm was verified when the noise disturbance suddenly changes.