From　the　aspect　of　potential　energy,　human-like　motion　planning　of　robotic　arm　is　studied.　However,　the　potential　energy　is　not　involved　in　the　potential　criterions,　because　it　is　difficult　to　determine　the　parameters　in　the　spring　models.　The　minimum　potential-energy　criterion　is　improved　from　the　angle　of　the　influencing　factors　of　potential　coefficient.　Based　on　a　large　number　of　experiments,　the　influencing　factors　of　the　spring　coefficient　are　obtained.　The　factors　are　divided　into　the　objective　factors　and　the　subjective　factors.　The　objective　factors　are　eliminated　through　standard　processing　of　the　experimental　data　and　the　subjective　factors　are　classified　by　the　clustering　methodology.　The　exact　mathematical　expression　of　the　stiffness　is　determined　through　multiple　regression　analysis　on　the　basis　of　the　influencing　factors.　The　arms　postures　predicted　by　this　new　criterion　and　other　criteria　are　compared　with　the　captured　real　arm　postures.　It　turns　out　that　the　arm　postures　predicted　by　the　new　criterion　are　more　close　to　the　real　arm　postures.　This　new　criterion　is　also　applied　into　the　human-like　motion　planning　of　a　7-DOF　Robai　robotic　arm　system.　These　results　all　verity　the　validity　of　this　new　criterion.　©　2015　Journal　of　Mechanical　Engineering.