Marine$\mathcal{X}$: Design and Implementation of Unmanned Surface Vessel for Vision Guided Navigation

Introduction

The utilization of marine robots, especially Unmanned Surface Vessels (USVs), has seen a significant upswing due to their capabilities to perform various maritime tasks with enhanced efficiency and reduced risk to human life. These tasks can range from environmental monitoring, search and rescue operations to maritime security.

Hardware Design

MarineX, a newly developed USV, embodies a robust catamaran design renowned for its stability and buoyancy. It's been engineered to optimize its performance with a carefully selected array of hardware components, including:

• Powerful T200 thrusters capable of producing substantial thrust for precise navigation.
• An advanced electronic system consisting of Navio2 and a Raspberry Pi4 board that manages actuator control for autonomous movement.
• NVIDIA Jetson TX2 computing board coupled with a ZED Mini camera facilitating real-time advanced vision processing necessary for navigation and object tracking.

Furthermore, various sensors and actuators work in harmony for the vessel’s adept maneuverability. The catamaran design is brought to life utilizing 3D printing technology, ensuring lightweight construction with sufficient payload capacity.

Software Architecture

At the core of this USV’s operational capability lies a sophisticated software architecture. It is built on the ROS Noetic framework, facilitating intercomponent communication for sensor and actuator management. A C++ based firmware ensures seamless interface with the hardware elements. Adding to this, the NVIDIA GPU on the TX2 handles vision-related tasks efficiently, making use of state-of-the-art algorithms for object detection.

For testing and optimization, Gazebo simulation software plays a vital role. It allows for the simulation of different environmental conditions and testing of navigation algorithms without any physical risk.

Vision Guided Navigation

The USV features an autonomous, vision-controlled navigation system powered by the YOLOv5 deep learning model. The model enables the detection and tracking of objects with high accuracy and rapid processing. Its implementation on marineX allows the USV to autonomously pursue and maintain a lock on specific targets on water surfaces. The control system utilizes PID (Proportional Integral Derivative) control, making precise adjustments to vessel steering and throttle based on visual feedback.

Extensive testing of marineX in various environments, from controlled marine pools to the unpredictable open sea, has successfully showcased its ability to perform real-time object tracking reliably, confirming its suitability for a wide spectrum of maritime operations.

Conclusion

MarineX represents a leap forward in marine robotics, marrying robust hardware with advanced vision-guided navigation. The extensive testing and successful operation in varied environments underscore this USV’s capability to undertake complex maritime tasks autonomously. The successful fusion of these technological advancements brings to light the promise that such autonomous systems hold for the future of maritime exploration and security.