A Study on Performance Prediction for Development of AI-Based Undersea Plant Autonomous Collecting Drone

This study designed an AI-based undersea drone capable of autonomously detecting and harvesting undersea plants. The proposed system actively compensates for changes in water depth through variable buoyancy control and utilizes a YOLOv8 based plant recognition algorithm to detect target plants in real time.The robot arm and electric gripper perform collecting movements based on AI recognition results, and the location movement is implemented using the Undersea Positioning System to enable tracking of the drone’s location even in environments where there is no undersea GPS communication network. As a result of the simulation, the design is predicted to achieve a harvest success rate of about 70~80%, a position error of less than 0.5m, and a recognition accuracy of more than 85%. The main contribution of this study is to present an integrated design framework for AI perception, buoyancy control, robotic arm control and an undersea autonomous harvesting system.