Abstract:Unmanned Surface Vehicles (USVs) represent an advanced technology with a wide range of applications in maritime contexts, and they offer a promising alternative to manned systems, particularly for the provision of services such as riverbed mapping. Furthermore, the use of USVs in combination with other robotic systems, such as flying drones, remotely operated underwater vehicles (ROVs) and autonomous underwater vehicles (AUVs), opens up innovative ways to inspect and monitor maritime infrastructure. However, the use of these systems in dynamic environments, such as ports or shipping lanes, requires a high degree of adaptability and precision in navigation to manoeuvre safely between other traffic participants and obstacles to prevent collisions. To address this problem, USVs require sophisticated perception systems for the precise detection and classification of nearby objects. Leveraging Light Detection and Ranging (LiDAR) technology, which is renowned for its efficacy in obstacle avoidance in robotics, this paper outlines the adaptation of a LiDAR sensor for USVs. The proposed pipeline enables the detection of other moving vessels and the accurate determination of their positions and trajectories up to 150 m from the USV. This paper introduces an obstacle detection system and validates it within a port setting. A methodical approach to LiDAR data preprocessing, clustering, and point cloud extraction will be outlined. Furthermore, it explores the application of an Unscented Kalman Filter (UKF) for the projection of motion paths of identified entities. This advanced form of data analysis provides the ability to predict the future position and path of potential obstacles, optimize decision-making for autonomous navigation algorithms and significantly improve the safety of USV operations. The pipeline is tested experimentally in sea trials by enacting five different scenarios that USV may encounter in a port setting. The results of the pipeline's tracking capabilities are compared to the ground truth data collected by the tracked vessel and plotted to visualize the error.

Reliable LiDAR-based ship detection and tracking for Autonomous Surface Vehicles in busy maritime environments

A Method with Improved Accuracy and Robustness for Object Detection in Wharf Scenarios

A Real-Time Vessel Detection and Tracking System Based on LiDAR

Multimodal Perception Integrating Point Cloud and Light Field for Ship Autonomous Driving

LiDAR-based simultaneous multi-object tracking and static mapping in nearshore scenario

Video-Based Detection Infrastructure Enhancement for Automated Ship Recognition and Behavior Analysis

Multi-Scale Ship Detection from SAR and Optical Imagery Via A More Accurate YOLOv3

EfficientShip: A Hybrid Deep Learning Framework for Ship Detection in the River

Real-time Tracking and Berthing Aid System with Occlusion Handling Based on LiDAR

Real-time Tracking and Dynamic Berthing Information Extraction System with 2D LiDAR Data

Multi-modal Information Fusion for LiDAR-based 3D Object Detection Framework

A LiDAR Point Clouds Dataset of Ships in a Maritime Environment

Lidar-Based Obstacle Detection and Path Prediction for Unmanned Surface Vehicles

Research on Surface Environment Perception Via Camera-LiDAR Sensor Fusion

A Joint Ship Detection and Waterway Segmentation Method for Environment-Aware of USVs in Canal Waterways

Visual Detection Algorithm for Enhanced Environmental Perception of Unmanned Surface Vehicles in Complex Marine Environments

River Boundary Detection and Autonomous Cruise for Unmanned Surface Vehicles.

Ship Detection Based on LiDAR and Visual Information Fusion

Ship detection from coastal surveillance videos via an ensemble Canny-Gaussian-morphology framework

Estimation of Berthing State of Maritime Autonomous Surface Ships Based on 3D LiDAR

Estimation of Ship Berthing Parameters Based on Multi-LiDAR and MMW Radar Data Fusion