Salvador López-Barajas;Alejandro Solis;Raúl Marín;Pedro J. Sanz
{"title":"Enhanced I-AUV Manipulation Capabilities Using Multicamera Perception and LLM-Based Interaction","authors":"Salvador López-Barajas;Alejandro Solis;Raúl Marín;Pedro J. Sanz","doi":"10.1109/JOE.2026.3708352","DOIUrl":"https://doi.org/10.1109/JOE.2026.3708352","url":null,"abstract":"Inspection, Maintenance, and Repair tasks on underwater structures are risky, costly, and complex, typically requiring professional divers or remotely operated vehicles. This work presents a framework for autonomous object grasping without relying on visual markers such as ArUco tags. The proposed solution uses an Intervention Autonomous Underwater Vehicle equipped with a multicamera system and a deep learning vision module. The neural network model was trained with images of target objects captured in simulation and dry environments rather than underwater conditions. Using these pretrained networks, the robot can accurately detect and grasp objects underwater despite the differences between training and operational environments. The framework also integrates a natural language processing module powered by a Large Language Model within the Robot Operating System (ROS) through ChatGPT’s API. This enables intuitive human–robot interaction, allowing operators to issue manipulation commands in natural language specifying target objects. The ROS-based intelligent agent interprets these commands and launches the autonomous behaviors required for precise grasping tasks. Experiments conducted in simulation and controlled water tank environments demonstrate the effectiveness and adaptability of the proposed system for underwater manipulation.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"2630-2640"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11618390","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391708","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"An Acoustic Tomography Experiment in a Stratified Lake at Super-Kilometer Ranges With Application of the Matched Nonreciprocity Method","authors":"John C. Wells;Ernst M. Uzhansky;Naokazu Taniguchi","doi":"10.1109/JOE.2026.3713817","DOIUrl":"https://doi.org/10.1109/JOE.2026.3713817","url":null,"abstract":"We present results of an acoustic tomography experiment conducted in a stratified lake at super-kilometer ranges. In late November 2018, five near-bottom transducers were deployed at bottom depths ranging from 9 to 68 m in the North Basin of Lake Biwa, Japan, with horizontal separations between the transducer pairs ranging from 6.6 to 15.2 km. Over a period of 13 days, the transducers performed quasi-simultaneous reciprocal transmissions every 10 min, using 11-bit maximum length quasi-random sequences modulating a 5-kHz carrier frequency. The observed arrival-time evolution, supported by ray tracing simulations, reveals two distinct propagation patterns in the recorded signals: an early-arriving group predominantly traversing the warm epilimnion, and a later-arriving group propagating through the colder hypolimnion. Analysis of travel-time nonreciprocity focused on the two stations, separated by 6.6 km, that communicated reciprocally for the entire 13 days. Following a wind-induced excitation, the early- and late-arrival groups exhibited anticorrelated travel-time nonreciprocities. For the same transect, we present the first application of the Matched Nonreciprocity Method to quantify current-induced nonreciprocity in acoustic transmission. Previously validated only using synthetic data, this method yielded plausible estimates of the depth-averaged current between the two transducers. This enables the estimation of currents in the presence of strong travel-time nonlinearities and ray losses resulting from wind-induced mixing of the water column and thermocline oscillations associated with internal gravity waves.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"2816-2827"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391717","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"AI-Powered CoralWatch-Aligned Framework for Coral Health Assessment and Bleaching Detection In Ocean Monitoring Applications","authors":"Lyes Saad Saoud;Irfan Hussain","doi":"10.1109/JOE.2026.3707995","DOIUrl":"https://doi.org/10.1109/JOE.2026.3707995","url":null,"abstract":"Monitoring coral bleaching is essential for understanding reef degradation and supporting timely conservation. However, current CoralWatch assessments largely depend on diver-based surveys and visual color-chart interpretation, making them subjective, labor-intensive, and difficult to scale. This study presents an AI-powered, CoralWatch-aligned framework for automated bleaching assessment from underwater imagery, with robotic deployment serving primarily as an enabling data-acquisition context. The system uses a remotely operated vehicle equipped with onboard cameras and a mounted CoralWatch reference card to capture standardized underwater images. A modular pipeline combines YOLOv12 for coral region detection, the Segment Anything Model for instance-level segmentation, and GPT-4o for protocol-guided bleaching interpretation using calibrated relative color references. The framework does not require species-level classification or fixed absolute color thresholds. To address underwater visibility degradation, we adopt a dual-data set evaluation strategy involving an annotated coral detection data set and two underwater enhancement data sets, one synthetic and one field-acquired. These data sets support system validation rather than constitute standalone contributions. RAUNE-Net, WaterNet, and UT-UIE are benchmarked to assess enhancement quality and downstream detection stability. Results show measurable detection improvements after image enhancement, while comparisons with expert annotations indicate consistent CoralWatch-level bleaching interpretation. Unlike manual surveys or purely offline workflows, the proposed approach supports standardized, reproducible, and scalable coral health assessment. Its platform-agnostic design allows deployment with robotic systems or diver-acquired imagery, providing a practical pathway toward objective AI-assisted reef monitoring.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"3677-3709"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391720","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"SPADE: Sparsity Adaptive Depth Estimator for Zero-Shot, Real-Time, Monocular Depth Estimation in Underwater Environments","authors":"Hongjie Zhang;Gideon Billings;Stefan B. Williams","doi":"10.1109/JOE.2026.3707696","DOIUrl":"https://doi.org/10.1109/JOE.2026.3707696","url":null,"abstract":"Underwater infrastructure requires frequent inspection and maintenance due to harsh marine conditions. Current reliance on human divers or remotely operated vehicles is limited by perceptual and operational challenges, especially around complex structures or in turbid water. Enhancing the spatial awareness of underwater vehicles is key to reducing piloting risks and enabling greater autonomy. To address these challenges, we present SPADE: sparsity adaptive depth estimator, a monocular depth estimation pipeline that combines a pretrained relative depth estimator with sparse depth priors to produce dense, metric scale depth maps. Our two-stage approach first scales the relative depth map with the sparse depth points, then refines the final metric prediction with our proposed Cascade Conv-Deformable Transformer blocks. Our approach achieves improved accuracy and generalization over state-of-the-art baselines, with zero-shot sim-to-real transfer, and runs efficiently at over 15 FPS on embedded hardware, promising to support practical underwater inspection and intervention.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"2715-2725"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391743","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Simultaneous Position and Velocity Estimation Using Doppler Shift and Time-of-Arrival in Acoustic Positioning Signals","authors":"Atsushi Tsuchiya;Tadashi Ebihara;Naoto Wakatsuki;Tohru Yoshihara","doi":"10.1109/JOE.2026.3697315","DOIUrl":"https://doi.org/10.1109/JOE.2026.3697315","url":null,"abstract":"In this article, we present a method for simultaneously estimating the position and velocity of an underwater vehicle using only an acoustic positioning system. This method measures time-of-arrival (TOA) and Doppler shift in acoustic positioning signals using a Doppler-compensated cross-correlation technique capable of handling wideband signals. An extended Kalman filter (EKF) incorporates TOA and Doppler shift as observations, enabling high-precision position and velocity estimation. We improve computational efficiency through dynamically selecting resampling points in the Doppler-compensated cross-correlation process based on the prior probability distribution from the EKF. Performance is evaluated using experimental data collected under actual sea conditions. Within a positioning area of 50-by-40 m and a maximum baseline length of 6 m, the method achieved 80th percentile errors of 1.18 m or less for position estimation and 0.19 m/s or less for velocity estimation.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"2847-2858"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Optical Ocean Recipes: Creating Realistic Data Sets to Facilitate Underwater Vision Research","authors":"Patricia Schöntag;David Nakath;Judith Fischer;Rüdiger Röttgers;Kevin Köser","doi":"10.1109/JOE.2026.3692005","DOIUrl":"https://doi.org/10.1109/JOE.2026.3692005","url":null,"abstract":"The development and evaluation of underwater machine vision is challenging, partly due to the lack of controlled, ground-truthed environments that realistically account for optical challenges such as color distortion, reduced contrast, or dynamic light patterns caused by artificial illumination. The severity of these effects varies across regions, depths, and seasons, and current data sets do not fully capture this diversity, leaving the transferability of evaluation results to new conditions uncertain. We propose an evaluation framework that allows more controlled and systematic evaluation. The introduced <italic>Optical Ocean Recipes</i> are a method to realistically recreate the optical properties of water of given reference water types in a laboratory tank environment using colorants and scattering materials. We demonstrate experimentally that with optimized combinations of the three food colorants Patent Blue VF, copper chlorophyllin, and Brown HT, and magnesium hydroxide as a scattering agent, we achieve tank water that is more similar to measurements of real ocean water than previously used optical additives, evaluated using its spectral absorption and scattering coefficients. The controlled environment enables the creation of ground-truth data for a range of vision tasks, including water parameter estimation, image restoration, image segmentation, visual SLAM, and underwater image synthesis. Our work includes the generation of a demonstration data set created in different optical conditions simulating the Jerlov water types (IB, II, 3C). On this data set, we exemplify robustness evaluations of two representative vision tasks, image matching and instance segmentation. In the case of image matching, we show how the tank environment facilitates access to ground truth necessary to evaluate established performance metrics that are otherwise more difficult to acquire. At the same time, the <italic>Optical Ocean Recipes</i> ensure water appearance comparable to open-water data. We analyze how different factors in the image formation process affect method performance, particularly the influence of water properties. Such disentanglement is only possible with ground truth for multiple quantities combined with diverse, realistic data. As we show, the <italic>Optical Ocean Recipes</i> enable this kind of comprehensive evaluation framework for underwater machine vision research.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"3661-3676"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11615017","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391682","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Investigating Location-Regularized Self-Supervised Feature Learning for Seafloor Visual Imagery","authors":"Cailei Liang;Adrian Bodenmann;Emma J Curtis;Samuel Simmons;Kazunori Nagano;Stanley Brown;Adam Riese;Blair Thornton","doi":"10.1109/JOE.2026.3702911","DOIUrl":"https://doi.org/10.1109/JOE.2026.3702911","url":null,"abstract":"High-throughput interpretation of robotically gathered seafloor visual imagery is beneficial for large-scale habitat class inference. Although recent research has suggested that location metadata can enhance self-supervised representation learning (SSL), its benefits across different SSL strategies, models and seafloor image data sets are underexplored. This study evaluates the impact of location-based regularization on six state-of-the-art SSL frameworks, which include convolutional neural network (CNN) and vision transformer (ViT) models with varying latent-space dimensionality. Experiments on multiple seafloor image data sets show that incorporating location metadata into SSL consistently enhances downstream classification performance. Location-metadata regularization is particularly effective for CNN models and compact latent representations, while ViTs demonstrate strong generalization across data sets even without explicit spatial constraints. These findings highlight the value of leveraging location information in SSL and support the use of transformer-based models for robust seafloor image analysis.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"2726-2736"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391652","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yerin Hwang;Eun-Joo Lee;Hajin Song;Byoung-Nam Kim;Young Ho Kim;Ho Kyung Ha;Yohan Choi;Jae-Hun Park
{"title":"A Novel Approach to Estimate 3-D Current Fields Using Coastal Acoustic Tomography System in Yeosu Bay, South Korea","authors":"Yerin Hwang;Eun-Joo Lee;Hajin Song;Byoung-Nam Kim;Young Ho Kim;Ho Kyung Ha;Yohan Choi;Jae-Hun Park","doi":"10.1109/JOE.2026.3724238","DOIUrl":"https://doi.org/10.1109/JOE.2026.3724238","url":null,"abstract":"Coastal acoustic tomography (CAT) is a technology used to observe variations in water temperature or current fields by analyzing two-way acoustic travel time signals between CAT stations in coastal environments. We have developed a novel method that combines inverse analysis with an artificial neural network (ANN) model to estimate high-resolution, 3-D current fields. High-resolution coastal ocean numerical model outputs, with a temporal resolution of 30 min and a spatial resolution of approximately 100 m, were used as the main data set. To evaluate the proposed method, the Bellhop ray tracing model was first used to compute reciprocal travel times between five CAT stations using 15 days of high-resolution ocean numerical model outputs. Then, a slope-fitting method was applied to each transmission path to generate an extended data set of travel time differences, while reducing computational demand. These travel time differences were used as input to the ANN model, which estimated path-averaged currents across five depth layers. Finally, an inverse analysis was conducted for each depth layer to reconstruct the full 3-D current field. Validation confirmed that the method effectively captured temporally varying, two-layer-like current structures that respond to tidal phase. The approach was further applied to in situ CAT observations, and its performance was verified through comparison with acoustic Doppler current profiler measurements. Our results indicate that the proposed novel method is suitable for operational 3-D current field monitoring in coastal regions using CAT systems.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"2828-2846"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11763477","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391659","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"CORALL: A COLREGs-Guided Risk-Aware LLM for Decision-Making in Maritime Autonomous Surface Ships","authors":"Klinsmann Agyei;Pouria Sarhadi;Wasif Naeem","doi":"10.1109/JOE.2026.3703695","DOIUrl":"https://doi.org/10.1109/JOE.2026.3703695","url":null,"abstract":"The paper introduces a novel Large Language Models (LLMs)-based solution for real-time COLREGs-based decision-making in collision encounters. The COLREGs, published by the International Maritime Organisation (IMO), have underpinned maritime collision avoidance regulations in open seas to ensure vessel safety. As COLREGs were originally defined for human seafarers, their translation into computer language for autonomous operations has received significant attention in the literature; however, a universal solution remains unavailable. A key requirement is explainable decision-making for autonomous operations or captain assistance. This study leverages the power of LLMs to develop such a decision-making algorithm. The proposed <underline>CO</u>LREGs-guided <underline>R</u>isk-<underline>A</u>ware <underline>LL</u>M (CORALL) algorithm integrates a supervisory decision-making module with an execution layer for path tracking and collision avoidance under external disturbances. The resulting framework enables COLREGs-compliant, risk-informed LLM-based navigation, with online performance demonstrated on a high-speed ship model. The tailored LLM processes navigation outputs and risk indices, identifies the COLREGs encounter type, and generates decisions with accompanying explanations. The proposed solution is tested on all the 22 Imazu's benchmark problems, evaluated using the Imazu benchmark scenarios, a widely accepted standard in maritime collision avoidance research comprising 22 progressively complex encounter situations. To complement the study, the proposed algorithm is verified using a developed Hardware-in-the-Loop (HIL) test rig, with end-to-end execution times measured to demonstrate the feasibility of real-time implementation on low-cost processors. This risk-aware AI solution, thought to be the first of its kind, opens a door to future research on advanced decision-making algorithms in safe maritime transport and similar operational environments.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"3183-3201"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fan Xiao;Zicong Zhuang;Fan Jiang;Hengyue Bi;Jingchao Cao;Junyu Dong
{"title":"LUNA: A Lightweight and Degradation-Aware Network for Underwater Image Quality Assessment","authors":"Fan Xiao;Zicong Zhuang;Fan Jiang;Hengyue Bi;Jingchao Cao;Junyu Dong","doi":"10.1109/JOE.2026.3716118","DOIUrl":"https://doi.org/10.1109/JOE.2026.3716118","url":null,"abstract":"With the rapid development of marine exploration and ecological monitoring, underwater vision systems have become increasingly important. However, underwater images frequently suffer from severe degradation, including color distortion, low contrast, and local blurring, posing significant challenges for reliable image quality assessment. Existing deep learning-based methods typically rely on complex networks with large numbers of parameters, leading to high computational costs. Moreover, these methods often do not adequately account for the unique degradation patterns characteristic of underwater images. To address these limitations, we propose a lightweight underwater network with degradation awareness for no-reference underwater image quality assessment (LUNA), specifically designed for efficient and accurate underwater image quality evaluation. LUNA incorporates three core modules: 1) a color feature attention module that effectively captures global color degradation; 2) an efficient multiscale attention module that extracts both fine-grained details and structural features, and a spatial shifting module with an adapter that enhances spatial context modeling. Extensive experiments on underwater image quality assessment (UWIQA) and underwater image quality database (UIQD) databases demonstrate that LUNA exhibits competitive or superior comprehensive performance, and achieves a favorable tradeoff between accuracy and efficiency, thus improving deployment feasibility in resource-constrained underwater scenarios.","PeriodicalId":13191,"journal":{"name":"IEEE Journal of Oceanic Engineering","volume":"51 4","pages":"3755-3765"},"PeriodicalIF":5.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149391699","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}