【摘要】 针对粒子滤波的退化现象、样本贫化问题以及标准无迹粒子滤波(UPF)算法计算量偏大的缺陷,利用基于超球面采样变换(SSUT)的UKF算法产生重要性概率密度函数,与序贯重要性再采样(SIR)结合,并引入粒子群优化,形成一种新的粒子滤波算法.对称分布UT变换的sigma点为2n+1个,而SSUT变换为n+2个.新算法利用SSUT变换减少了采样点的个数,通过混合建议分布进一步减少了计算量,使计算效率得到了明显的改善.仿真结果表明,该算法滤波精度优于扩展卡尔曼粒子滤波,而与标准UPF相当,计算效率明显高于标准UPF算法.更多还原

【Abstract】 As an important nonlinear filtration method,particle filters have become an area of intense study by domestic and foreign researchers.A standard unscented particle filter(UPF)suffers from degeneracy phenomenon and sample impoverishment problem as well as requiring an immense amount of calculation.Application of the unscented Kalman filter algorithm,based on a spherical simplex unscented transformation(SSUT),created a function with an importance density.In combination with sequential importance resampling(SIR)and the introduction of particle swarm optimization,this formed a new algorithm for particle filters.The standard unscented transformation adopts a symmetric set of 2n +1 sigma points,and the SSUT has n +2 sigma points.With the SSUT cutting down the number of sigma points and the proposed combined distribution cutting down the calculation requirements still further,the efficiency of computation was considerably improved,while retaining filter accuracy equal to that of standard UPF.Simulation results indicated that the new algorithm enhances filter precision compared with the extended Kalman particle filter.Results should be similar to those of a standard UPF,yet it has greater calculating efficiency.更多还原