Recently, the unscented Kalman filter unbiased minimum variance-based (UKF-UMV) has been explored to address state estimation issues with unknown inputs and Gaussian noises. However, the performance of the UKF-UMV will degrade dramatically in the presence of non-Gaussian noises. To tackle this issue, this letter proposes a Gaussian kernel-based estimator called the maximum correntropy unscented simultaneous input and state estimator (MCUSISE). Firstly, the unscented transformation (UT) is applied to handle nonlinear propagation. Drawing on statistical linearization techniques (SLT), the nonlinear measurement function is converted into a linear regression equation (LRE). Then, two optimization problems based on the maximum correntropy criterion (MCC) are formulated to perform simultaneous input and state estimation (SISE). The optimal SISE is achieved by iteratively adjusting the gain matrices for input and state using fixed-point iterative algorithms (FPI). Moreover, no predefined assumptions or prior constraints are imposed on the unknown input, allowing it to take any model. Finally, simulation experiments validate that the proposed MCUSISE achieves improved performance when handling unknown inputs and non-Gaussian noises, particularly impulsive noises.
扫码关注我们
求助内容:
应助结果提醒方式:
