In　order　to　resolve　the　shortcoming　of　the　traditional　federal　Kalman　filtering,　a　new　method　is　presented　in　which　the　fuzzy　reasoning　system　is　combined　with　the　traditional　Kalman　technology.　This　method　can　modify　the　statistical　characteristic　of　noises　on　real　time.　It　can　not　only　modify　the　local　filter　but　also　bring　forward　a　bran-new　information　fusion　arithmetic　which　is　applied　to　central　filter.　A　fuzzy　reasoning　system　is　built　to　obtain　different　weights　of　the　all　states　estimate　values　that　are　based　on　the　actual　circumstance　and　fuse　these　values.　Then　the　complicated　calculation　is　avoided　and　precision　is　advanced.　At　last　the　new　means　is　applied　in　the　INS/GPS　Integrated　Navigation　system　and　the　result　of　computer　simulation　indicates　that　it　is　useful　with　high　effectiveness　and　practical.