As　an　important　type　of　associative　learning,　operation　conditioning　and　its　mathematical　models　have　been　studied　a　lot.　The　recent　trend　is　to　introduce　intrinsic　motivation　in　operant　conditioning　to　expand　the　search　space.　However,　traditional　curiosity-based　intrinsic　motivation　models　have　a　strong　preference　for　those　states　seldomly　visited.　As　a　result,　they　intend　to　ignore　the　states　most　possibly　leading　to　target,　which　may　decrease　the　efficiency.　Aiming　to　solve　the　problem,　we　propose　a　task-oriented　curiosity　based　intrinsic　motivation　model　(TOCIM).　The　model　is　described　as　a　tuple　consisting　of　8　elements,　including　state　space　S,　action　space　A,　orientation　matrix　O,　orientation　function　V,　access　number　matrix　N,　curiosity　matrix　C,　orientation　update　mechanism　e,　and　action　selection　strategy　G.　Here,　the　intrinsic　motivation　is　measured　not　only　by　the　novelty　of　the　sates,　but　also　by　the　correlation　between　the　states　and　the　target　in　order　to　trade　off　exploration　and　exploitation　in　navigation.　Simulation　experiments　have　been　carried　out　to　testify　the　validation　of　TOCIM,　and　some　other　similar　models　have　been　compared.　The　experiment　results　show　that　our　model　has　advantage　in　training　time　of　navigation.　©　2022　ACM.