Abstract: This paper addresses the problem of autonomous real-time navigation for agricultural robot in greenhouse environment.A novel path planning and tracking method based on fuzzy predictive control was proposed.The navigation preview point was acquired by image segmentation using trained neural network method,then the angle error and the lateral error can be calculated.When the angle error was less than 25 degrees,the robot driving wheels were differential controlled using the planning path which was a fitting circular curve based on predictive control.When the angle error was greater than 25 degrees,the driving wheels' speed was controlled by fuzzy controller.The experiments showed that the curve of path tracking was smooth using this method.The most tracking error was less than 5cm at the speed of 0.3 m/s.