Since　the　gradual　learning　process　like　humans　or　animals　of　two-wheeled　robot　cannot　be　realized　by　the　traditional　control　methods,　an　autonomous　operant　conditioning　automaton　(AOCA)　is　established　based　on　Skinner＇s　theory　of　operant　conditioning　for　self-balance　learning　control　of　robots.　A　bionic　learning　algorithm　based　on　AOCA　is　proposed　to　balance　the　two-wheeled　robot.　The　corresponding　simulation　experiments　for　self-balance　learning　control　of　the　two　wheeled　robot　are　given,　in　which　the　robot　effectively　realizes　autonomous　balance.　Theoretical　analysis　and　simulation　show　that　the　autonomous　operant　conditioning　automata　bionic　learning　model　applied　to　the　two-wheeled　robot　for　autonomous　balance　learning　control　makes　the　robot　progressive　formation　of　self-organization,　development　and　improvement　of　its　balance.　Copyright　©　2014　Acta　Automatica　Sinica.　All　rights　reserved.