简介：Toimprovethereliabilityandaccuracyoftheglobalpositioningsystem(GPS)/microelectromechanicalsystem(MEMS)inertialnavigationsystem(INS)integratednavigationsystem,thispaperproposestwodifferentmethods.Basedonwaveletthresholddenoisingandfunctionalcoefficientautoregressive(FAR)modeling,acombineddataprocessingmethodispresentedforMEMSinertialsensor,andGPSattitudeinformationisalsointroducedtoimprovetheestimationaccuracyofMEMSinertialsensorerrors.ThenthepositioningaccuracyduringGPSsignalshortoutageisenhanced.ToimprovethepositioningaccuracywhenaGPSsignalisblockedforlongtimeandsolvetheproblemofthetraditionaladaptiveneuro-fuzzyinferencesystem(ANFIS)methodwithpoordynamicadaptationandlargecalculationamount,aself-constructiveANFIS(SCANFIS)combinedwiththeextendedKalmanfilter(EKF)isproposedforMEMS-INSerrorsmodelingandpredicting.Experimentalroadtestresultsvalidatetheefficiencyoftheproposedmethods.