This　paper　presents　the　methods　for　position　tracking　of　the　manipulator　based　on　the　model　predictive　control　approach.　Benefiting　from　characteristic　of　predictive　control　-　a　procedure　for　a　finite　horizon　optimal　control　problem　-　this　method　count　the　minimal　control　when　control　energy　reaches　its　least　in　range.　Moreover,　the　mutative　control　use　of　a　posteriori　and　a　priori　numerical　information,　which　is　readily　available　in　the　scheme,　is　evaluated　through　the　computed　closed-loop　trajectories.　Being　different　from　the　conventional　MPC,　the　predictive　model　used　in　this　paper　is　non-linear.　The　key　attribute　of　scheme　presented　is　an　internal　model,　on　which　the　predictive　control　updates　with　each　certain　angle-range　changing.　Although　compute　needs　support　of　hardware,　the　effectiveness　of　the　proposed　scheme　can　be　concluded　if　this　approach　be　applied　to　the　eye-to-hand　RR　manipulator　system.