Fuzzy　logic　system　(FLS)　promises　an　efficient　way　for　obstacle　avoidance.　However,　it　is　difficult　to　maintain　the　correctness,　consistency,　and　completeness　of　a　fuzzy　rule　base　tuned　by　a　human　expert.　In　this　paper,　a　novel　approach　termed　probabilistic　fuzzy　controller　with　operant　learning　(PFCOL)　for　robot　navigation　is　presented.　Operant　learning　(OL)　is　a　form　animal　learning　way.　The　key　feature　of　this　approach　is　that　it　combines　a　probabilistic　stage　and　a　stochastic　perturbation　generator　module　into　FLS　to　handle　problems.　At　last,　the　ultimate　output　is　determined　by　these　two　uncertain　stages.　This　imitates　animal　learning　method　of　generating　stochastic　behavior　in　the　complex　and　uncertain　environment.　The　simulation　results　show　that　the　proposed　PFCOL　method　can　automatically　generate　approximate　actor　to　adapt　complex　circumstances.　Through　studies　on　obstacle　avoidance　and　goal　seeking　tasks　by　a　mobile　robot　verify　the　approach　is　superior　in　generating　efficient　fuzzy　inference　systems.