【作者】 李彩菊； 李亚安；

【Author】 LI Cai-ju~1,LI Ya-an~2 (Collage of Marine Engineering,Northwestern Polytechnical University,Xi’an,710072,China)

【机构】 西北工业大学航海学院；

【摘要】 在各种非线性滤波技术中,扩展卡尔曼滤波(EKF)是一种最简单的算法,它将卡尔曼滤波(KF)局部线性化,适用于弱非线性、高斯环境下。Unscented卡尔曼滤波(UKF)用一系列确定样本来逼近状态的后验概率密度,适用于高斯环境下的任何非线性系统。粒子滤波(PF)用随机样本来近似状态的后验概率密度,适用于任何非线性非高斯环境,但有时选择的重性分布函数与真实后验有较大差异,从而导致滤波结果存在较大误差,而Unscented粒子滤波(UPF)正好克服了这一不足,它先通过UKF产生重要性分布,再运用PF算法。通过仿真实验,对四者的性能进行比较,结果证明,在非线性非高斯环境下,UPF的性能明显优于另外三种滤波器。更多还原

【Abstract】 In all kinds of nonlinear filter algorithms,Extended Kalman Filter that based on local linearization of KF,and has a good performance in Gaussian and mild nonlinear environment is one of the simplest algorithms.Unscented Kalman Filter (UKF) utilizes a set of definite samplings to approximate.posterior probability density(PDF) function,while Particle Filter(PF) uses random particles.Hence,UKF is suitable for any nonlinear but Gaussian environment,but PF don’t have the limitation of Gaussian.Sometimes,the performance of PF will be bad when the importance proposal distribution has a large difference with the true PDF.The Unscented Particle Filter(UPF) solved the shortage of the PF above,which first uses an UKF to generate the importance proposal distribution,and then follow the PF algorithm.By simulation experiments,their performances are compared.The results prove the performance of UPF is much better than another three filters in nonlinear and non-Ganssian environment.更多还原