改进的无迹粒子滤波在组合导航中的应用研究

Improved Unscented Particle Filtering for Integrated Navigation System

粒子滤波算法是一种适用于非线性非正态约束的统计滤波算法.针对粒子滤波存在退化现象,从围绕增加粒子的多样性和重要性分布函数的选择出发,提出了一种改进的无迹粒子滤波算法.该算法是利用无迹卡尔曼滤波产生的近似高斯分布作为重要性密度函数,在每次迭代中,结合马尔科夫链蒙特卡洛使粒子能够移动到不同地方,从而可以避免贫化现象.将这种算法应用到GPS/DR组合导航系统中,仿真结果证明了采用改进的无迹粒子滤波方法能达到很好的跟踪效果.

Particle filter is statistical filter arithmetic based on analog used in the non linearity and non normal restriction.There is the degenerate problem in particle filter.A improved particle filter method is given according to increasing particle diversity and choosing important distribution function.The approximate Gaussian distribution generated Unscented Kalman filter is used important distribution function in the method.The particle is moved different place using MCMC in each iteration to avoid degenerate problem.The method is used in GPS/DR integrated navigation system.The result of simulation has shown that the method have better tracking capability using improved UPF.