System Identification of UAV Based on EWC-LMS

Abstract

Mean-squared error (MES) can at best provide biased solution in the presence of additive disturbances (both correlated and white) on the input and the output signals. However, in the system identification of unmanned aerial vehicle (UAV), typically, in the takeoff, the UAV suffers serious disturbance e.g. wind gusts, turbulence, ground influence and sampling noise. For EWC, we can get the unbiased parameter estimation in white noise, and the system identification in this paper prove the noise rejection ability of EWC. EWC (error whitening criterion) was first put forward by Mr. Rao in 2002 and applied in filtering, and further developed theoretically. But EWC has not been applied in practical problems yet. This paper applies EWC in the system identification of real self-developed UAV's takeoff motion, and compares the result of EWC-LMS with traditional LMS (least mean square) and TLS (total least square). The result shows that in comparison with traditional LMS and TLS, EWC-LMS improve the performance significantly