Design and Development of an Autonomous Guided Mobile Manipulator for Precision Tomato Picking and Harvesting

In many countries of the world, labor shortages and youth disinterest in agriculture have compounded the challenge in tomato harvesting, resulting in food scarcity, and inefficiencies. This project aims to design and develop a prototype of an Autonomous Guided Mobile Manipulator for Precision Tomato Harvesting to address these challenges. The system integrates four major subsystems in closed-loop architecture: sensing, mobility, vision, and manipulation. A low-cost ESP32-CAM module was used for real-time tomato detection using a deep learning model trained on annotated dataset and deployed in a light weight TensorFlow Lite format. An Arduino Mega served as the main controller, and a 5 DOF robotic arm of total length 45.3cm equipped with five servo motors each having a torque of 15kg.cm. Mobility was enabled through a differential drive platform having standard DC motors of torque 5kg.cm with speed of 600RPM, and ultrasonic sensors for obstacle avoidance. Experimental results showed that the vision system achieved an average detection accuracy of 90% in normal daylight, while the integrated system achieved a harvest success rate of 68–75% in semi-field conditions. Overall, this project demonstrates the feasibility of affordable robotic harvesters in addressing labor shortages and a great contribution in precision agriculture.