A　novel　evaluation　method　of　granularity　partition　for　functional　modules　based　on　robot　technology　middleware　(RTM)　is　proposed.　Firstly,　the　comprehensive　weights　for　pertinence　indexes　of　structures　and　functions　of　modules　are　calculated　by　fuzzy　analytical　hierarchy　process　(FAHP),　and　correlation　matrix　of　the　system　is　established.　A　fuzzy　dendrogram　clustering　algorithm　is　proposed　to　obtain　the　module　partition　schemes　for　the　robot　system　under　different　granularities.　To　construct　multi-attribute　decision　matrix,　the　models　of　cohesion　and　coupling　for　each　scheme　are　structured　as　two　sources　of　evidences　for　DS　(Dempster-Shafer)　evidence　theory　based　on　the　principle　of　module　independence.　Then　the　trust　intervals　of　every　decision　scheme　are　sorted　by　a　preference　ordering　method　for　intervals　to　obtain　the　optimal　module　partition　scheme　for　the　robot　system.　The　evaluation　method　is　verified　by　applying　it　to　the　robot　3D　mapping　system.　The　system　implementation　and　results　show　that　the　method　is　effective　and　feasible.　©　2016,　Science　Press.　All　right　reserved.