基于ATSUKF算法的卫星姿控系统故障估计

针对卫星姿态控制过程中可能发生的执行机构或敏感器故障，提出了一种基于无损卡尔曼滤波（UKF）及偏差分离原理的自适应二阶无损卡尔曼滤波（ATSUKF）算法。首先，提出TSUKF算法，通过UKF处理姿态机动时的非线性并通过偏差分离原理将非线性系统的状态及故障分别估计，避免非线性模型的线性化过程同时降低了计算过程中的矩阵维度。然后，在TSUKF算法的基础上提出了ATSUKF算法，通过滑动窗口内的残差计算自适应矩阵，使滤波器在统计特性不准确的情况下仍然具有较快的收敛速度，特别适用于卫星快速机动过程中的姿态与故障估计。数值仿真结果表明，ATSUKF算法相较于TSUKF算法能有效降低统计特性不准对系统造成的不利影响，实现卫星姿态、执行机构/敏感器故障的快速估计。

Fault estimation in satellite attitude control system based on ATSUKF algorithm

Based on the bias-separate principle and the Unscented Kalman Filter (UKF), an Adaptive Two-Stage Unscented Kalman Filter (ATSUKF) is proposed to address the fault estimation of actuators/sensors in the satellite attitude control system. First, the TSUKF is presented. The decoupled state and faults are estimated by applying the bias-separate principle and using the nonlinearity by the UKF when the attitude is maneuvered without linearization of the system model. This avoids lowering the dimension of the matrix during the estimating process. Based on the TSUKF, an adaptive version is proposed. The adaptive matrices are calculated by the residuals in the sliding data window to make the noise covariance matrices change adaptively. With the ATSUKF estimator, the convergence rate is enhanced when the prior knowledge of the noise covariance matrices are inaccurate or the fault changes. Numerical simulations demonstrate the effectiveness of the approach proposed.