SystemIdentification ofUAV Basedon EWC-LMS

Mean-Squared Error(MES)canatbestprovide biased solution inthepresence ofadditive disturbances (both correlated andwhite) ontheinput andtheoutput signals. However, inthesystem identification ofUnmannedAerial Vehicle (UAV),typically, inthetakeoff, theUAV suffers serious disturbance e.g.windgusts, turbulence, ground influence andsampling noise. ForEWC,we cangetthe unbiased parameter estimation inwhitenoise, andthe system identification inthis paperprovethenoise rejection ability ofEWC.EWC (Error Whitening Criterion) wasfirst putforward byMr.Raoin2002andapplied infiltering, andfurther developed theoretically. ButEWC hasnotbeen applied inpractical problems yet. Thispaperapplies EWC inthesystem identification ofreal self-developed UAV'stakeoff motion, andcompares the result ofEWC-LMS withtraditional LMS (Least Mean Square) andTLS(Total LeastSquare). Theresult shows thatincomparison withtraditional LMS and TLS, EWC-LMSimprove theperformance significantly.