自适应噪声定界的改进集员辨识算法

Improved set-membership identification algorithm with adaptive noise bounding

集员辨识所需的系统噪声边界在现实应用中往往难于精确确定,通常采取的过估边界将导致算法性能的退化.本文针对缺乏足够先验噪声边界知识下的集员辨识问题进行了相应的研究,通过对输入干扰和测量误差的有界假设,将系统噪声边界建模为一个依赖于模型参数的时变量,由此提出了一种根据估计参数自适应调定噪声边界的改进最优定界椭球辨识算法,避免了过估噪声边界假设引起的保守性增大的缺陷,提高了算法的收敛速度.仿真中将本文提出的改进算法和带固定过估噪声边界的原始算法进行了比较,表明了该方法的有效性.

In the setmembership identification （SMI）, it is difficult to precisely determine the bounds of the system noise in most real applications. The widely used overestimated bounds will deteriorate the performance of the algorithm. We investigate this problem when the a priori knowledge of the noise bound is insufficient. Under the assumptions of bounded system inputs and measurement errors, we model the noise bound as a timevarying variable depending on some model parameters. We propose an enhanced optimal bounding ellipsoid （OBE） identification algorithm with adaptive boundtuning to adjust the noise bound based on the estimated parameters, which prevents the increased conservation from the overestimated bound assumption and improves the convergence rate of the algorithm. Simulation results show higher effectiveness of the proposed algorithm than that of the conventional algorithm with fixed overestimated noise bound.