Acrobot, which is also called gymnastic robot, is a single input-multi output (SIMO) system which imitates motion of a barbell athlete. However, two links of this model are just equivalent to body and hand of athlete. To imitate better motion of athlete, three-linked acrobot- a complicated multi input-multi output (MIMO) under-actuated system – is developed. Parts of barbell athlete including leg, body and hand, are modelled well by this new system. In this paper, we present a method to apply an 81-rule fuzzy algorithm, which is popularly used for invert pendulum systems, to balance this model at upward equilibrium point. Instead of using a complicated 729-rule fuzzy controller which is set by knowledge of experts, we use more simple fuzzy structure. Through genetic algorithm (GA), the pre-processing and post-processing coefficients of fuzzy controller are chosen. With these suitable coefficients, our fuzzy controller is proved to work well through Matlab/Simulink simulation.