抗差自适应无味粒子滤波

ROBUST ADAPTIVE SCENTLESS PARTICLE FILTER

针对系统观测值含有粗差或者存在状态异常的情况,基于无味粒子滤波算法(UPF)和抗差自适应滤波原理,提出了一种新的抗差自适应无味粒子滤波算法(AR-UPF)。该方法采用方差膨胀模型抑制观测粗差的影响;利用自适应UKF算法来实时调整状态参数预报值的协方差,控制状态异常的影响。算例分析表明,新方法能够有效地控制观测粗差与状态异常,提高了滤波精度。

In order to resist the bad influence of measurement outliers and state model errors, a new robust adaptive unscented particle filter is proposed, which is based on unscented particle filter algorithm and adaptively robust filter theory. First, the new method uses the variance expanding model to resist measurement outliers. Second, the adaptive unscented Kalman filter（UKF） is used to adjust the covariance of state predicted value, and to reduce the influence of state model errors. The examples have proved that the new procedure of the robust adaptive un- scented panicle filter is feasible and efficient.