Abstract: In order to realize the autonomy excavation of excavator robot, the behavior control system architecture was built. With the excavation behavior as a baseline, the state-flow model was used to implement the excavation object, the excavation task and the excavation behavior stepwise. The object image, rake angle of mechanical arm and pressure signal of hydraulic cylinder were collected, as the event and condition for trigger and transition between behavior states of state-flow. According to the three kinds of mining environment during the excavation, which are sand, surface block material and block material buried in sand, the visual position, pressure and rake angle information were gathered and mining action state-flow model was triggered after the fuzzy clustering discrimination, then the different circumstances encountered in mining were processed distinguishingly and independently. Finally the autonomy excavation object was achieved by controlling excavation action.