This　paper　presents　an　operant　conditioning　automata　model　(hereinafter　referred　to　＂OCM＂),　and　designs　a　bionic　autonomous　learning　method　which　can　be　used　to　describe　and　simulate　a　bionic　autonomous　learning　process　The　model　can　be　considered　as　an　active　learning　permitting　to　select　a　better　action　according　to　psychology　behavior　propensity,　and　the　aim　is　to　learn　to　find　the　optimal　action　finally　During　the　learning　process,　the　system　selects　an　action　randomly　according　to　the　probability　distribution　of　action　selection,　which　is　updated　by　the　behavior　propensity　from　the　environment　We　apply　our　model　on　skinner-pigeon　experiment　In　simulation,　we　confirmed　that　this　model　could　successfully　simulate　operant　conditioning