Grating tiling which is considered as a feasible technology can be used to obtain multiple-meter-sized large-aperture gratings, and stability control is one of key problems. So, it is important to find available and fit proportion integration differentiation (PID) closed loop control in required high accuracies condition (submicroradian orientations and nanometer positions). In this paper, by means of the principle of back propagation (BP) nerve net modified PID controller, the first value that a system in zero initial conditions using step response produces is replaced with sign function to achieve single neuron adaptive PID control, and based on that, in order to improve the risetime of the system at the beginning of the response, a changeable neuron proportionality coefficient replaces the constant K value. Simulation result indicates that the proposed control algorithm will not bring greater overshoot with quicker risetime than the traditional PID controller; and the experimental result indicates that the proposed control algorithm can guarantee the precision of the grating-tiling.