基于UPF的INS非高斯噪声故障检测

UPF-based Fault Detection Method of Integrated Navigation System with Non-Gaussian Noises

由于实际组合导航系统很可能受到非高斯噪声的影响,而传统的故障检测方法对非高斯噪声情况讨论较少。基于粒子滤波的故障检测技术非常适合处理非线性、非高斯问题,并且有效地克服了传统方法的不足,但是普通的粒子滤波器存在粒子"退化"等问题。为此,本文提出了一种基于UPF滤波器的导航系统故障检测方法,通过在普通粒子滤波中引入UKF产生建议分布及重采样的方法,有效抑制了普通粒子滤波器粒子"退化"的问题。并针对噪声的非高斯特性,将似然检测方法与粒子滤波可以估计似然函数的特点相结合,提出了一种基于UPF的故障检测方法。通过GPS/SINS组合导航系统在噪声服从瑞利分布情况下的故障检测仿真实例,表明此方法适用于在非高斯噪声情况下的导航系统故障检测。

As the actual navigation systems are likely to be affected by non - Gaussian noises, while the traditional fault detection meth- ods had little discussion on this situation. Because the particle filter - based fault detection is appropriate for non - linear and non - Gaussian problems, it has effectively overcome the deficiencies of traditional methods. But there are particle degeneration drawbacks of the common particle filter. The paper presents an unscented particle filter （UPF） - based fault detection method of navigation systems with non - Gaussian noises. By using the UKF to generate proposal distributions and the re - sampling methods within the particle filter, the drawbacks of particle ＂ degeneracy＂ of the generic particle filter can be effectively compensated. Considering the characteristics of the particle filter which can estimate the likelihood function, a UPF - based fault detection method is presented. The fault detection sim- ulation examples of the GPS/SINS integrated navigation system with Rayleigh noises show that this method is applicable to the fault de- tection of navigation systems in the case of non - Gaussian noises.