Abstract:Robust adaptive filtering supposes that result of robust least square is accurate in current epoch.Through discrepancy between updated kinematic information and the measurements,navigation is adaptive.But in actual car navigation,rate of gross error is high,which makes accuracy of robust result drop.In order to solve this problem,this paper gives a robust adaptive filtering based on moving windows which regard mean error of robust least square solution of previous epochs as the threshold of current epoch.Experiments of road based on GPS/COMPASS realkinematic positioning indicate that this algorithm can improve accuracy of navigation in urban environment.