Aimed at the puzzle of PID parameter tuning, the paper proposed a sort of parameter auto-tuning method based on self-learning. By means of fuzzy auto-tuning parameter on line for PID controller, the system could be automatically switched to normal work state to guarantee choiceness quality of control system. Once the change of system performance was happened, it would carry through the parameter auto-tuning process until obtaining the better control effect. The paper took the design of high precision control as an example and validated the correctness of parameter auto-tuning method. The simulation effect shows that it is better in control quality and dynamical performance indexes, higher in system robustness and control precision.