论文部分内容阅读
提出了一种鲁棒化的基于变分贝叶斯的自适应卡尔曼滤波算法.该算法采用具有重尾特性的学生分布取代高斯分布来描述量测模型,减弱系统对于野值的敏感性;再利用变分贝叶斯方法对修正后的模型的时变参数进行逼近推断,在递推地估计状态的同时还能对变化的噪声方差进行跟踪,并更新引入的自由度参数,从而在自适应滤波的同时增强了鲁棒性.仿真实验证明了在野值存在且噪声变化的观测下该算法的自适应与鲁棒性.
A robust adaptive Kalman filter algorithm based on variational Bayesian is proposed.The algorithm replaces Gaussian distribution with the distribution of students with heavy tail to describe the measurement model and weakens the sensitivity of the system to outliers. By using the variant Bayesian method, we approximate the time-varying parameters of the modified model and estimate the state recursively while tracking the variance of the noise variance and updating the parameters of the introduced degrees of freedom The robustness is also improved by adaptive filtering.The simulation results show that the algorithm is adaptive and robust under the presence of outliers and noise changes.