UD Decomposition Based Adaptive UKF for Nonlinear Estimation of States and Parameters
作者单位：College of Artificial IntelligenceJianghan University
会议名称：《第40届中国控制会议》
会议日期：2021年
学科分类：080902[工学-电路与系统] 0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学]
关 键 词：adaptive UKF nonlinear estimation UD decomposition
摘 要：The unscented Kalman filter（UKF） has been a helpful technique which is widely used for nonlinear estimation of states and parameters, fusion of information, target tracking, and fault diagnosis, etc. However, in the case of unknown covariance matrices of process noises and measurement noises, its convergence and performance cannot be guaranteed. To overcome this problem, this paper proposed an adaptive UKF, which is based on the principle of match of covariance between the theoretical matrix and the estimated matrix. A threshold condition is given out to trigger the adaptive process, thus the original stability of UKF will not be decreased. What’s more, a UD decomposition technique is introduced to improve the numerical stability of the adaptive UKF. To compare the performance of the adaptive method and UD decomposition based UKF, an index of degree of mismatch is proposed alongside with root mean square error. Simulation results verify the better accuracy and robustness of the proposed filter.