The　aim　is　to　gain　a　rational　estimator　of　monotone　regression　function　with　randomly　right　censored　data.　Using　the　theory　of　the　greatest　convex　minorant　and　the　statistical　idea　of　padding　the　uncensored　data　highly,　a　new　estimator　is　proposed.　The　estimator　can　guarantee　monotone　property,　which　is　superior　to　traditional　kernel　estimators.　Thus,　in　practical　applications,　such　as　establishing　children＇s　growth　curve,　when　right　censored　data　comes　up　and　the　regression　function　can　be　concluded　to　be　monotone　by　historical　experience　or　common　sense,　this　estimation　method　is　more　natural.　Under　assumed　conditions,　the　asymptotic　distribution　of　the　estimator　is　found　and　the　estimator　is　shown　to　be　consistent,　indicating　that　the　estimator　has　good　limit　behaviors.