Design and testing of a crawler robot based on mid-axis row guidance in a shade house

In response to the poor signal quality of satellite navigation in the highly shaded environment of the Panax notoginseng shade house and the navigation issues that may arise when tracks deviate significantly from their expected path in complex environments such as soft and sticky soil. This paper uses an enhanced YOLOv5s object detection algorithm and the least squares method to identify the mid-axis of the shade house. A fuzzy adaptive PID path tracking algorithm and a computer vision navigation control system based on the mid-axis were designed for the tracked robot. A tracked robot platform was constructed and a precision ginseng seeder with hole-pressing and seeding functions was designed. The path-tracking algorithm was developed in MATLAB/Simulink and simulation experiments were conducted. The simulation tests results showed that, even under conditions of significant environmental changes and disturbances, the proposed fuzzy adaptive PID algorithm outperformed conventional PID and fuzzy logic control. Field trials were conducted in a simulated environment. Experimental tests show that Im-YOLOv5 achieved an average accuracy of 94.9 % for the root base of ginseng, with maximum and average path deviations of 13.7 mm and 3.8 mm respectively for the tracked robot guided by the mid-axis. The robot exhibited no loss of control and was capable of navigating between rows within the shade house. Seedling experiments were conducted at various speeds. Field seeding pass rates exceeded 90 % in all cases and re-seeding and missed seeding rates were both controlled below 5 %, meeting operational standards. These results provide technical references for seeding and harvesting operations using shade house planting robots.