The　behavior-based　method　is　more　robust　and　better　in　real-time　performance　than　the　traditional　method　for　the　robot　control　in　the　unknown　environment.　A　behavior-based　intelligent　controller　is　designed　for　the　control　of　a　2-dof　manipulator　in　this　paper.　The　intelligent　controller　(IC)　is　based　on　the　critic-action　model　and　takes　the　reinforcement　learning　algorithm　as　the　learning　algorithm.　By　using　the　critic-action　model,　the　controller　can　learn　the　behavior　online　and　does　not　rely　on　the　system　model.　The　experiment　shows　that　the　controller　can　realize　the　continuous　control　for　the　2-dof　manipulator　and　the　manipulator　can　reach　the　target　position　quickly.