The application of immersive virtual maintenance technology can find the problems of the products in the process of design, guaranteeing the qualities and reducing the life-cycle cost. As an indispensable part in immersive virtual maintenance, motion control for virtual human is a critical factor for promoting simulation efficiency. But it still remains on basis of motion editing or images, causing it a time-consuming task to simulate the maintenance process. Therefore, we propose a real time motion control algorithm based on optical motion capture, making the virtual maintenance both immersive and efficient. To ensure the algorithm fast enough, an editable human model is constructed based on simplified human skeleton. In order to make the virtual human work in an unlimited range while the simulation worker moves in a limited space, a walking gesture is defined and a prototype of virtual human’s action database is built. To obtain continuous visual effects, the view direction of the virtual human has been smoothed by adopting a gyroscope. Finally, the algorithm and its effectiveness have been proven by experiments.