粒子滤波器及其在卫星姿态估计中的应用

Particle Filter and Its Application in Satellite Attitude Estimation

粒子退化和计算量较大是限制粒子滤波应用的主要问题,常规的重采样方法虽然可以缓解粒子退化,但却容易导致粒子枯竭,且计算量较大,因此本文提出了基于混沌摄动的均值逼近粒子滤波器。按权值大小将粒子分组后,用均值替换权值较小的粒子,可使粒子从低似然区向高似然区域逼近。用Kullback信息描述均值逼近产生的粒子分布与似然分布的差别,通过迭代发现Kullback信息是递减的,从而证明该算法是合理的。混沌摄动重采样算法,用类似载波的方法将具有全局遍历性的混沌变量引入,更增加了粒子的多样性。另外,将本算法应用于某型导弹的姿态估计问题中,仿真结果显示了新算法的有效性。

Degeneracy phenomenon and large calculation are the main disadvantages of particle filtering,which restricts its application.Common re-sampling methods can resolve the degeneracy phenomenon,but the sample impoverishment and large calculation is a secondary result.So a mean approaching particle filter based on chaotic perturbation is proposed to resolve the above problem.The particles are divided into two groups according to its weight,and then the small weight particles are replaced by the mean so that the particles can approach from the low likelihood region to the high likelihood region.Moreover,the difference between the particle distribution produced by the behavior of mean approaching and the likelihood distribution is described by the Kullback information.The Kullback information decreases with the increasing iteration degree,which proves that this algorithm is rational.Furthermore,a chaotic perturbation re-sampling method is presented to reduce calculation,which imports the chaotic variable that has the property of stochastics similar to the carrier wave in order to ameliorate the diversity of samples.In addition,this algorithm is applied to a certain type of missile attitude estimation.The simulation results demonstrate that the new algorithm is feasible.