非线性随机系统的鲁棒自适应平滑估计算法

Robust adaptive smoothing algorithm for nonlinear stochastic systems

针对有色噪声下一类含有未建模动态和扰动的非线性随机系统,提出一种鲁棒自适应平滑估计算法。该算法通过极小化状态平滑估计误差的方差和相邻时刻残差的协方差,在线辨识状态滤波估计误差和残差的方差,实现对未建模动态和扰动的自适应补偿。仿真结果验证了该算法在解决有色噪声下非线性随机系统的时变时滞与参数联合估计问题中的有效性。

Considering a class of nonlinear stochastic systems with colored noises subject to unmodeled dynamics and disturbances, a robust adaptive smoothing algorithm was proposed. The algorithm can compensate such unmodeled dynamics and/or disturbances through online parameters identification and has strong robustness against unmodeled dynamics and disturbances. The optimal parameters, such as variances of state filtering errors and residuals, are determined by minimizing the variances of state smoothing errors and the covariance of the residuals at two adjacent times. The simulation results show the satisfying adaptive ability to track changes of time delay and parameter of a nonlinear stochastic system with colored noises.