SLAM　has　become　the　core　method　of　robot　positioning,　but　SLAM　based　on　EKF　method　cannot　solve　the　filter　sensitivity　problem　caused　by　uncertain　parameters.　This　paper　proposed　a　self-adjusting　desensitized　EKF-SLAM　system　based　on　the　Desensitised　EKF　which　momentarily　adjusts　the　emphasis　on　the　sensitivity　in　the　cost　function.　This　algorithm　utilizes　the　comparison　of　systematic　theoretical　residual　and　practical　innovation　to　construct　the　weight　factor　of　the　sensitivity　in　the　cost　function,　which　adjusts　the　sensitivity　weight　online　so　as　to　realize　the　adaptive　estimation　of　the　desensitized　Kalman　filter.　This　method　is　applied　to　the　EKF-SLAM　system　to　overcome　the　degradation　of　system　performance　caused　by　systemic　sensitivity　to　uncertain　parameters　from　the　odometer.　Finally,　the　results　from　the　outdoorsy　experiment　demonstrate　the　performance　of　the　proposed　desensitized　EKF-SLAM　system.　©　2018　IEEE.