Based on state of robot, an improved tracking control method which takes the tracking error as an input and the drive voltage of the motor as an output is proposed. A theoretical domain deflation factor is added to automatically adjust the fuzzy controller, and the range of parameters was determined preliminarily by experimental measurement. Then the sparrow population was initialized by adding a reverse learning strategy, and a weight factor was designed to control the proportion of followers close to the finder to prevent precocity. The results show that the improved sparrow algorithm performs better than the original algorithm in terms of seeking accuracy and convergence speed under the test function. The method in this paper has a rapid reduction of the tracking error to zero, a smoother tracking process, and a faster arrival speed to the desired path. The method improves the working condition of the omnidirectional robot.