Author(s)

Shaoshan Guo, Yunqian Huang, Jiaxin Zhang

Corresponding Author

Shaoshan Guo

Abstract

This study presents the design and implementation of an intelligent lake cleaning system based on the architecture of the Artificial Intelligence of Things (AIOT). The system is structured into four layers: sensory, executive, network, and application. The sensory layer collects environmental data, while the executive layer handles the physical operations of the cleaning boat. The network layer ensures data transmission and control, and the application layer processes information to facilitate intelligent operations such as automatic cleaning and path planning. The hardware design includes modules for satellite positioning, image capture, lighting, and water level detection, with a focus on energy efficiency through the use of ARM architecture and solar power. The system employs a modified YOLOv8 model for object detection, enhanced with CBAM attention mechanisms for improved accuracy. Path planning algorithms are optimized for obstacle scenarios, enhancing the cleaning efficiency. Comparative testing of the DC-YOLOv8 model demonstrates significant improvements in accuracy and frame rate over existing models. Functional tests confirm the system's effectiveness in real-world lake cleaning operations.

Keywords

AIOT, Lake cleaning boat, YOLOv8, Real-time detection, Path planning