It　is　essential　for　lunar　rover　to　know　its　position　in　order　to　fulfill　the　lunar　exploration　mission.　So　the　autonomous　navigation　method　for　lunar　rover　has　become　an　important　topic.　In　this　paper,　a　new　autonomous　navigation　method　for　lunar　rover　based　on　federated　Kalman　filter　was　described.　The　principle　of　this　method　was　introduced,　and　the　system　equation　for　lunar　rover　based　on　the　rover　motion　model　was　developed.　Then　the　celestial　navigation　subsystem　based　on　the　measurement　coming　from　sun　sensor　and　the　DR　navigation　subsystem　based　on　the　measurement　coming　from　velocity　gyro　and　odometer　were　presented.　And　the　federated　Kalman　filter　was　used　to　implement　optimization　estimation　and　information　fusion　in　this　method.　Finally,　the　simulation　of　this　method　was　finished.　A　simulation　results　demonstrated　that　this　method　has　higher　precision　of　position　and　head　angle.　Another　simulation　results　demonstrated　the　reliability,　fault　tolerance　and　feasibility　of　this　method.