This paper presents a system modeling based control scheme of an ultra precision positioning system for a grating ruling machine. Since the positioning system having a long stroke with ultra precision, the positioning system consists of a coarse positioning stage driven by a servo motor and a fine positioning stage driven by a piezoelectric ceramic. In order to improve positional accuracy and remove the noise components of motion, a hybrid control scheme based on the system modeling is implemented. Considering position-dependent and time-dependent behaviors of the stages, a model based LQ controller is utilized to the coarse stage and a PID feedback controller based on neural network is utilized to the fine stage. Experiment results reveal the efficient and robust of the control scheme and show that the positional accuracy has been readily achieved within 8.6 nm.