Abstract
This paper introduces an advanced acoustic imaging system leveraging multibeam water column data at various frequencies to detect and classify marine litter. This study encompasses (i) the acquisition of test tank data for diverse types of marine litter at multiple acoustic frequencies; (ii) the creation of a comprehensive acoustic image dataset with meticulous labelling and formatting; (iii) the implementation of sophisticated classification algorithms, namely support vector machine (SVM) and convolutional neural network (CNN), alongside cutting-edge detection algorithms based on transfer learning, including single-shot multibox detector (SSD) and You Only Look once (YOLO), specifically YOLOv8. The findings reveal discrimination between different classes of marine litter across the implemented algorithms for both detection and classification. Furthermore, cross-frequency studies were conducted to assess model generalisation, evaluating the performance of models trained on one acoustic frequency when tested with acoustic images based on different frequencies. This approach underscores the potential of multibeam data in the detection and classification of marine litter in the water column, paving the way for developing novel research methods in real-life environments.

Abstract
Exploiting stronger winds at offshore farms leads to a cyclical need for maintenance due to the harsh maritime conditions. While autonomous vehicles are the prone solution for O&M procedures, sub-sea phenomena induce severe data degradation that hinders the vessel's 3D perception. This article demonstrates a hybrid underwater imaging system that is capable of retrieving tri-dimensional information: dense and textured Photogrammetric Stereo (PS) point clouds and multiple accurate sets of points through Light Stripe Ranging (LSR), that are combined into a single dense and accurate representation. Two novel fusion algorithms are introduced in this manuscript. A Joint Masked Regression (JMR) methodology propagates sparse LSR information towards the PS point cloud, exploiting homogeneous regions around each beam projection. Regression curves then correlate depth readings from both inputs to correct the stereo-based information. On the other hand, the learning-based solution (RHEA) follows an early-fusion approach where features are conjointly learned from a coupled representation of both 3D inputs. A synthetic-to-real training scheme is employed to bypass domain-adaptation stages, enabling direct deployment in underwater contexts. Evaluation is conducted through extensive trials in simulation, controlled underwater environments, and within a real application at the ATLANTIS Coastal Testbed. Both methods estimate improved output point clouds, with RHEA achieving an average RMSE of 0.0097 m -a 52.45% improvement when compared to the PS input. Performance with real underwater information proves that RHEA is robust in dealing with degraded input information; JMR is more affected by missing information, excelling when the LSR data provides a complete representation of the scenario, and struggling otherwise.

Abstract
Autonomous Surface Vehicles (ASVs) are bound to play a fundamental role in the maintenance of offshore wind farms. Robust navigation for inspection vehicles should take into account the operation of docking within a harbouring structure, which is a critical and still unexplored maneuver. This work proposes an end-to-end docking approach for ASVs, based on Reinforcement Learning (RL), which teaches an agent to tackle collision- free navigation towards a target pose that allows the berthing of the vessel. The developed research presents a methodology that introduces the concept of illegal actions to facilitate the vessel's exploration during the learning process. This method improves the adopted Actor-Critic (AC) framework by accelerating the agent's optimization by approximately 38.02%. A set of comprehensive experiments demonstrate the accuracy and robustness of the presented method in scenarios with simulated environmental constraints (Beaufort Scale and Douglas Sea Scale), and a diversity of docking structures. Validation with two different real ASVs in both controlled and real environments demonstrates the ability of this method to enable safe docking maneuvers without prior knowledge of the scenario.

Abstract
The increasing adoption of robotic solutions for inspection tasks in challenging environments is becoming increasingly prevalent, particularly in the offshore wind energy industry. This trend is driven by the critical need to safeguard the integrity and operational efficiency of offshore infrastructure. Consequently, the design of inspection vehicles must comply with rigorous requirements established by the offshore Operation and Maintenance (O&M) industry. This work presents the design of an autonomous surface vehicle (ASV), named Nautilus, specifically tailored to withstand the demanding conditions of offshore O&M scenarios. The design encompasses both hardware and software architectures, ensuring Nautilus's robustness and adaptability to the harsh maritime environment. It presents a compact hull capable of operating in moderate sea states (wave height up to 2.5 m), with a modular hardware and software architecture that is easily adapted to the mission requirements. It has a perception payload and communication system for edge and real-time computing, communicates with a Shore Control Center and allows beyond visual line-of-sight operations. The Nautilus software architecture aims to provide the necessary flexibility for different mission requirements to offer a unified software architecture for O&M operations. Nautilus's capabilities were validated through the professional testing process of the ATLANTIS Test Center, involving operations in both near-real and real-world environments. This validation process culminated in Nautilus's reaching a Technology Readiness Level 8 and became the first ASV to execute autonomous tasks at a floating offshore wind farm located in the Atlantic.

Abstract
Underwater scenarios pose additional challenges to perception systems, as the collected imagery from sensors often suffers from limitations that hinder its practical usability. One crucial domain that relies on accurate underwater visibility assessment is underwater pipeline inspection. Manual assessment is impractical and time-consuming, emphasizing the need for automated algorithms. In this study, we focus on developing learning-based approaches to evaluate visibility in underwater environments. We explore various neural network architectures and evaluate them on data collected within real subsea scenarios. Notably, the ResNet18 model outperforms others, achieving a testing accuracy of 93.5% in visibility evaluation. In terms of inference time, the fastest model is MobileNetV3 Small, estimating a prediction within 42.45 ms. These findings represent significant progress in enabling unmanned marine operations and contribute to the advancement of autonomous underwater surveillance systems.

Abstract
Marine growth impacts the stability and integrity of offshore structures, while simultaneously preventing inspection procedures. In consequence, companies need to employ specialists that manually assess each impacted part of the structure. Due to harsh sub-sea environments, acquiring large quantities of quality underwater data becomes difficult. To mitigate these challenges a new data augmentation algorithm is proposed that generates new images by performing localized crops on regions of interest from the original data, expanding the total size of the dataset approximately 6 times. This research also proposes a learning-based algorithm capable of automatically delineating marine growth in underwater images, achieving up to 0.389 IoU and 0.508 Dice Loss. Advances in this area contribute for reducing the manual labour necessary to schedule maintenance operations in man-made submerged structures, while increasing the reliability and automation of the process.