Aiming at the problem of obstacles threat for wheeled AGV tracking along a vineyard ridge road, based on distance measurement and heading angle detection, rolling optimization principle was used to propose an anti-collision obstacle avoidance navigation algorithm of wheeled AGV. Cross-arranged 8-way ultrasonic sensors were used to detect the boundaries of the ridge road and obstacles on the ridge road. The same obstacle matching, obstacle type criterion and trajectory prediction model of dynamic obstacle were put forward. The AGV lateral position deviation and heading deviation angle were used as inputs of fuzzy controller to obtain expected guidance angle of AGV front wheel, which provided a referenced heading angle for wheeled AGV to bypass obstacles. Based on the rolling optimization principle, the obstacle avoidance navigation strategies of adjusting heading deviation angle and vehicle speed were proposed, and the AGV navigation algorithm of bypassing static obstacles, avoiding dynamic obstacles by deceleration or parking of AGV were designed. The software of LabVIEW was used to develop the obstacle avoidance navigation algorithm program, and navigation data was sent to the PLC to control AGV steering in real-time. Simulation experiments showed that the navigation algorithm can make the AGV track with the mode of deviation rectification along preset path of ridge road, which can guide the AGV to avoid collision with obstacles on the ridge road, which verified the effectiveness of the navigation algorithms. It can provide a reference for anti-collision warning and obstacle avoidance technology of front-wheel guided vehicle like unmanned wheeled tractor to track along a ridge road.