Abstract: In a double pendulum robot, there are four equilibrium states, namely one natural stable position (down-down) and three unnatural stable positions (down-up, up-down, up-up). With transfers between these states, 20 acrobatic actions can be formed (12 states transfer actions and 8 circumgyration actions). Using the human simulated intelligent control (HSIC) theory based on sensor-motor intelligent schema, an intelligent control system for the double pendulum robot which has the structure of multi controllers and multi control modes is designed. A quasi-equivalent modeling method and an improved genetic algorithm are adopted for the accurate parameters identification of the double pendulum model and the optimization of numerous characteristic and control parameters in controller. By this way, not only the design and parameter optimization of complex HSIC controller are changed easily, but also a very difficult problem which is to transfer quickly from computer simulation of model to real-time control of physical system is solved successfully. Finally, we take the state transfer (swinging up) control from down-up to up-down as an example to explain the arbitrary transfer control between the four equilibrium states of the double pendulum, and show how to design controller with HSIC theory. The successful simulation and real-time control testified the validation of proposed theory and method.