Tutorials
Organizers: Jennifer Paduan and Eve Lundsten
Half day – AM
Abstract: The nuts and bolts of using the MB-System software package for processing multibeam sonar data collected on Autonomous Underwater Vehicles (AUVs). MB-System (https://www.mbari.org/technology/mb-system/) is open-source software for processing and display of seafloor mapping data. Surveys from submerged platforms such as AUVs and ROVs involve challenges beyond conventional mapping, including designing complicated surveys in complex terrain, addressing interference from other acoustic sources, assessing attitude sensor biases, correcting for inertial navigation drift, and merging multiple surveys collected over time on terrain that has experienced changes, such as submarine canyons and active volcanoes. This tutorial will introduce MB-System to the participants and then walk them through how to: 1. Design AUV survey missions 2. Set up MB-System processing (the pre-processing stage) for 3 different AUV platforms 3. Edit multibeam bathymetry data in both waterfall and 3-D point cloud views 4. Estimate and correct for attitude biases 5. Adjust vehicle navigation to match features in overlapping swathes 6. Merge surveys 7. Generate common data products, including GIS objects These steps will be shown as sequences of screen grabs and videos in PowerPoint and by actively working on example datasets. Templates of scripts we work with will be provided, and our work-flow and methods will be discussed, including of newer features of MB-System. This tutorial will focus on 1-meter-scale multibeam bathymetry and backscatter collected using MBARIs Dorado Mapping AUVs, WHOIs AUV Sentry, and SOIs Hugin Superior AUV. Since MB-System supports a large number of data formats from most commonly used mapping sonars, these examples should be broadly relevant to users of other AUV or ship-mounted sonars. Participants who have experience with multibeam sonars, particularly on submerged platforms, will find this tutorial most beneficial. Novice users will come away with an introduction to MB-System and an appreciation for the complexities of working with AUV mapping data. This tutorial will NOT include installation or troubleshooting of MB-System working on participants computers.
Organizers: Kevin Hardy, Rob Morris, and Brock Rosenthal
Full Day (limited to 25 attendees)
Abstract: This course will provide an understanding of free vehicle and ocean lander design, applications, advantages, limitations, and seaborne operations. Using video presentations, printed materials, and hands-on demonstrations, attendees will be actively involved in the process of learning how to acquire or build their own ocean landers, and the immediate steps they can take to gain the confidence needed to pull the quick release. Freed from winch and wire, ocean landers can spend months, even years, at any depth. The surface vessel is likewise free from the ocean lander, and able to move off to pursue other cruise objectives. Multiple ocean landers can survey wide areas from a nonspecialized surface vessel. Ocean landers can carry traps, samplers, sensors, and cameras, and make surface-to-seafloor profiles. A number of commercial components are readily available to create a purpose-built ocean lander. A general-purpose ocean lander with interchangeable payload modules could support several science departments, providing a common motor pool for deep ocean access. The discussion will be reinforced with real-world examples of applications and experiences, good and bad, from the presenters’ personal experience and those of other operators. Free vehicles can be small enough to be lifted with one hand from the ocean, yet strong enough to journey to the bottom of any ocean trench. Small free vehicles with no HazMat have been flown overnight on passenger aircraft to remote destinations, assembled dockside, loaded and taken to sea. Sitting stationary on the seafloor, ocean landers are absolutely quiet, allowing recording of ambient sound fields. Nothing beats an ocean lander for persistence, silently monitoring for multiple years. Ocean engineers gain valuable experience, confidence, and competence by training in the design and deployment of free vehicles and ocean landers. This tutorial will take attendees through an interactive process of defining and designing a free vehicle based on a real cruise profile. Commercial Off-The-Shelf (COTS) sources of components will be discussed, permitting end-users the opportunity to intelligently consider the Make-Buy decision, based on factors of time, cost, and quality.
Organizers: Ankur Verma, Ayush Goyal, and Aaron Micallef
Half day – PM
Abstract: Multi-modal ocean sensing has entered a new regime of data scale. Deep water measurements like DAS can produce Terabytes of data per day. Combing DAS data with other complementary data modalities like ADCP, weather data, current data, among others is often required for comprehensive scientific interpretation. Consequently, the data workflow to go from TBs of raw data to a sound scientific understanding is becoming more complex. In the MBARI Geo-Sense study and in comparable deployments Lightscline has worked with, we see a common thread of problems consuming a lot of SME time and effort: 1. Needle-in-a-haystack discovery 2. Plot multiplicity 3. Multi-modal pulling and alignment 4. Context switching 5. Folder-structure navigation This tutorial addresses how solving these workflow problems compresses months of SME effort into hours, using the Geo-Sense DAS deployment as a hands-on exercise. Using the Lightscline dashboard for TB-scale multi-modal data, participants will reproduce Geo-Sense papers central figures from raw data in a single guided day and extend the analysis into a frequency band the published pre-print does not cover. The goal is to leave participants with a working intuition for how oceanographic signals, cableseafloor coupling, tidal forcing, and earthquake signals appear in raw DAS, and a workflow they can transfer to their own multi-modal ocean datasets within hours instead of months.
Organizer: Rayette Toles-Abdullah, Austin Park and Muni Annachi
Half-Day – PM
Abstract: Marine ecosystems face unprecedented pressures from climate change, overfishing, and habitat degradation. NOAA Fisheries and the global ocean science community increasingly rely on AI and computer vision to process decades of fisheries survey imagery, electronic monitoring video, and autonomous underwater vehicle footage transforming manual review processes that once took months into near-real-time species detection and abundance estimation. This half-day hands-on tutorial provides ocean scientists and engineers with practical skills to build, train, and deploy marine species detection models. Participants will work through a complete machine learning pipeline: from ingesting labeled fisheries imagery, to training object detection models, to deploying inference endpoints capable of processing underwater video at scale. Learning Objectives: – Understand the end-to-end ML lifecycle for marine species detection, from data preparation through model deployment – Gain hands-on experience with Amazon SageMaker for training computer vision models on fisheries survey data – Learn techniques for handling common challenges in underwater imagery: variable lighting, occlusion, class imbalance across species, and domain shift between survey regions – Build and deploy a scalable inference pipeline suitable for electronic monitoring catch review and fisheries stock assessment support The tutorial is structured around a realistic use case drawn from fisheries science: automating species identification in bottom-trawl survey imagery to support stock assessments. Participants will use pre-labeled datasets representative of NOAA Fisheries survey protocols, train detection models using SageMaker built-in algorithms and custom training scripts, evaluate model performance against scientific accuracy requirements, and deploy models as real-time inference endpoints. No prior machine learning experience is required participants should have basic Python familiarity. All compute resources are provided. Participants will leave with a working pipeline template, reusable notebooks, and best practices for applying AI to their own data.
More Information: Marine Species Identification Using Amazon SageMaker
Organizer: Zhengnan Li and Milica Stojanovic
Half-day – AM
Abstract: Reproducible evaluation of underwater acoustic receivers requires the ability to drive any candidate signal through a realistic channel and a realistic noise field, with full control over the experimental conditions. Most researchers solve this problem in private, with code and channel files that never leave the originating laboratory. The Underwater Acoustic Channel Library, available at github.com/uwa-channels, removes this barrier. It provides eight measured channels from sites including the North Atlantic, Singapore, Hawaii, Norway, Japan, the Mariana Trench, and the Pacific, spanning transmission ranges from tens of meters to thousands of kilometers and center frequencies from 75~Hz to 25~kHz. The companion MATLAB and Python toolboxes implement three core operations: replay (pass any signal through a measured channel), noisegen (generate realistic ocean noise with pink Gaussian, spatially correlated Gaussian, or impulsive alpha-stable statistics), and unpack (reconstruct the full time-varying impulse response from the compressed storage format). Despite the documentation and tested example scripts, our experience is that first-time users still take days to weeks to integrate the toolbox into their own evaluation pipelines, especially when they want to go beyond the canned examples. This tutorial is designed to lead the attendees through the process in a time-efficient and comprehensive manner. Attendees will leave with a working pipeline on their own laptop, replaying their own signals through several measured channels, with realistic noise added, and with a clear mental model of the underlying representations and assumptions.
Organizers: Shahriar Negahdaripour and Guilherme Oliveira
Half-day – AM
Abstract: We propose a full-day tutorial covering fundamental computer vision principles and techniques based on the processing of 2-D multi-beam forward-looking sonar imagery, as well as key advantages in integrating the encoded information in 2-D sonar image intensities induced by acoustic backscatter with the visual cues from overlapping optical images. Due to the exponential growing interest in the topical areas, we expect strong participation from both researchers and graduate students is marine robotics and computer vision as well as deep learning engineers who seek to apply modern AI techniques to complex, low-visibility sub-sea environments. Attendees will leave with an understanding of how to bridge the gap between traditional optical computer vision pipelines and advanced sonar and acoustic-optical sensor fusion models.
Workshops
Organizer: Organizers: Ms. Sidney Dills, Dr. Ruoying (Roy He), and Dr. James (Jim) Doyle
Full-day
Abstract: Artificial Intelligence and Machine Learning are rapidly transforming meteorology and oceanography (METOC), creating new opportunities to deliver decision-making advantages to naval operations. The ability to process complex environmental data and run advanced forecast models at the tactical edge is critical for real-time awareness, particularly in communications-denied maritime environments. This full-day workshop provides a comprehensive introduction to an emerging operational paradigm for AI-enabled forecasting, both afloat and ashore.
The session focuses on deploying next-generation METOC capabilities by bridging cutting-edge research from defense labs and academia with operational engineering from industry partners. Participants will explore the latest advancements in accelerating data assimilation with machine learning, running AI on low-power maritime hardware through edge computing, and ensuring operator trust. The workshop emphasizes the operational rigor, accuracy, and reliability required to safely transition these capabilities to the fleet.
Through a series of keynotes, technical briefs, and scenario discussions, attendees will progress from understanding strategic requirements to engaging with real-world scenarios, such as deploying relocatable models in contested environments. The workshop is designed for physical oceanographers, atmospheric and data scientists, defense professionals, and software engineers from government, academia, and industry who are focused on developing and deploying the next generation of tactical environmental forecasting.
Organizers: Samira Daneshgar Asl, Robert Waterman and Keith Van Graafeiland
Half-day – PM
Abstract: Marine oil spills from shipping activity, offshore infrastructure, and extreme weather events continue to pose significant risks to ocean ecosystems, coastal communities, and maritime operations. Rapid detection and mapping are critical for guiding response actions, assessing environmental impacts, and supporting transparency among agencies, industry, and the public. Synthetic Aperture Radar (SAR), particularly Sentinel‑1, has become a cornerstone for operational oil spill monitoring due to its all‑weather, day‑night imaging capability and sensitivity to surface roughness changes caused by floating oil. This workshop provides a hands‑on introduction to an ArcGIS Onlinebased workflow for detecting, extracting, and visualizing marine oil spills using Sentinel‑1 Radiometric Terrain Corrected (RTC) SAR imagery, cloud‑native geospatial analysis, and pretrained GeoAI deep learning models. Participants will learn how to access analysis‑ready Sentinel‑1 imagery from a global archive, apply on‑the‑fly preprocessing within a web mapping environment to optimize SAR data for oil slick detection, and automate pixel‑based classification using a pretrained deep learning model designed for marine conditions. The workflow emphasizes efficiency and scalability, minimizing manual interpretation while enabling rapid response across large spatial and temporal extents directly in the cloud. Through a real‑world case study, attendees will progress from imagery selection and preprocessing to oil slick extraction, vectorization, and web‑based visualization for sharing results with stakeholders. The workshop will also address practical considerations such as false positives, interpretation limitations, credit and compute management, and best practices for integrating SAR‑derived oil spill products into broader operational response and environmental monitoring frameworks. This workshop is designed for ocean scientists, maritime safety professionals, environmental agencies, and geospatial practitioners seeking to operationalize SAR and GeoAI within ArcGIS Online for marine monitoring, emergency response, and decision support.
Organizer: Sophie Scopazzi
Half-day – AM
Abstract: Build an open-source CTD (and other systems) using consumer-off-the-shelf components for less than $500. Examples of system adaptability into a temperature logger, bottom pressure recorder, surface float, and an off-grid weather station will be in-person to experiment with during the workshop.
Organizer: Rachel Spratt
Half-day – AM
Abstract: The big data revolution in oceanography often carries a high pricetag for proprietary software and computing power. This workshop promotes equity in research by exclusively showcasing work built on free-tier, open-source ecosystems. Our objective is to highlight students who achieve high-level science using free tools like GEE [1], NASA AppEEARS [2], and the Copernicus Database Ecosystem JupyterHub [3]. We will have a targeted focus, addressing critical gaps in Coastal Arctic and Coastal Gulf of America ecosystems. We will use the afternoon to deconstruct the how-to of the most successful morning posters.
Organizers: Brian Bingham, Christine Buzzell, Josh Mangelson, Evan Palmer, Ivan Stenius, and Mabel M. Zhang
Half-day – PM
Abstract: For ground and aerial autonomy, simulation is part of the daily development cycle. Demonstrating that a navigation or perception stack is correct can require tens to hundreds of millions of simulated miles, which on-vehicle testing cannot cover, and tools such as CARLA, AirSim, and Isaac Sim are now common ground across research groups, vendors, and program offices~[1,2]. Ocean robotics has the same need. Surface, underwater, and maritime aerial systems must be tested across distances, durations, and weather and sea states that on-water trials cannot reach, and the same simulators are increasingly asked to produce synthetic data for machine-learning perception pipelines~[3,4]. The current state of practice in ocean simulation is fragmented. Our working literature review~[3] identifies several open-source projects that each cover part of the problem: Gazebo-based stacks such as VRX~[5], DAVE~[6], and the MBARI LRAUV simulation; the game-engine-based HoloOcean~[4] on Unreal and OceanSim~[7] on NVIDIA Isaac Sim; the standalone Stonefish~[8]; and the hydrodynamics specialist WEC-Sim. Each has its own strengths in fidelity, multi-domain coverage, headless CI use, and suitability for vectorized ML training. No single project today plays the role that CARLA or AirSim play for the driving and aerial communities, and federated one-off funding for individual simulators has slowed convergence. OCEANS 2026 in Monterey gathers the academic, naval, industrial, and ocean-instrumentation groups whose combined needs define what an ocean simulator has to be, which makes it the right venue for this conversation. We scope the workshop to multi-domain simulators that meet at least two of three working criteria: (i) open source under a usable license; (ii) documentation good enough that a new user can stand the tool up without reading the source; (iii) a large active user base, ideally beyond just maritime applications. This scope excludes one-off academic prototypes. It includes projects built on closed engines (Unreal, Unity, Isaac Sim) when the marine-specific layer is itself open and meets the criteria.
Organizer: Jean-Paul Dube
Half-day – PM
Abstract: This half day workshop establishes an interactive, data-generating simulation environment to evaluate the structural and operational resilience of commercial maritime assets facing accelerated, mandatory regional and international Underwater Radiated Noise (URN) legislation. Moving beyond passive presentations, this session utilizes a structured human in the loop simulation to stress-test engineering compliance pathways against capital allocation constraints. Participants are organized into cross-functional stakeholder cohorts (naval architects, marine engineers, fleet operators, and asset valuation underwriters), to resolve a localized legislative disruption scenario. The simulation forces cohorts to balance hydrodynamic realities (cavitation inception speeds, machinery attenuation profiles) against commercial asset lifecycle valuation and portfolio risk metrics. Crucially, the simulation models the principal-agent information asymmetries that manifest between vessel owners, charterers, and capital underwriters regarding a hull’s true acoustic profile under regulatory stress. Participants will analyze the utility of independent, third-party technical verification as an operational mechanism to standardize baseline performance metrics, protect asset lifecycles, and mitigate portfolio impairment during regulatory transitions. The aggregated decision-making data generated during the session will be synthesized into an open-access technical roadmap, providing an empirical framework for underwriting technology transition risks across the maritime capital stack.
Organizers: Stefan Domino and Vadim Pavlov
Half-day – AM
Abstract: Over the last two decades, advances in numerical methodologies and their deployment on high-performance computing platforms have transformed numerous engineering disciplines, facilitating both deeper physical insight and improved design for multiphysics applications. Such modeling and simulation tools have reshaped the engineering analysis landscape, driving advancements in renewable energy systems design, strengthening fire safety mitigation strategies, and advancing environmental fluids characterization. This workshop explores how recent high-fidelity computational advances can be leveraged to transform understanding of biological systems via morphological designs in maneuvering and cruising specialists (e.g., drag quantification and energy budgets) to advancing the characterization of marine ethology (e.g., thermoregulation of ocean sunfish, functional design quantification of male pufferfish nest structures, etc.). The workshop objectives are to capture current exemplars for computational approaches that are focused on solving marine-based use cases, while promoting the open sharing of knowledge, experience and current challenges/opportunities. Our workshop will provide examples that illustrate a science-based workflow, including essential attributes such as 1) robust solid geometry definition and creation; 2) meshing approaches (e.g., focused on quality and suitability to resolve critical features of the application); 3) the computational fluid dynamics (CFD) engine that includes exercising an appropriate set of numerical schemes, and finally; 4) analysis and discovery of multiphysics processes that are commonly found in marine-based use cases. Pitfalls and lessons learned for each aspect of this workflow will be identified, as will best practices. We will foster and encourage new partnerships/research paths via a multi-institution, cross-disciplinary collaborative approach. This workshop is timely given the long-standing United States investment in computational tools that reside in the open-source ecosphere. Many tools have not been fully leveraged to the OCEANS 26 community, allowing for a unique opportunity to partner computational scientists with oceanic researchers.
