studies-rov

Studies of ROV concepts 2026

1 - Inertial Sensor Self-Calibration Module Using Attitude Heading and Reference System for Autonomous Underwater Vehicle Navigation

Authors: Abdul Aziz Abd Kadir, Herdawatie Abdul Kadir This research enhances navigation accuracy in Autonomous Underwater Vehicles (AUVs), which rely on Inertial Measurement Units (IMUs) for orientation and motion tracking but are prone to noise-induced inaccuracies. To mitigate this, an Inertial Sensor Self- Calibration Module was developed to dynamically adjust calibration parameters in real time. A comparative analysis of calibration and filtering techniques, including Magnetic Calibration, EKF variants, and AHRS, revealed the AHRS filter as the most precise, reducing angular error by up to 97.81%. These results underscore the AHRS filter's effectiveness in improving AUV navigation in challenging underwater conditions, offering broader implications for ocean exploration and environmental monitoring.

2 - Development of a Cost-Effective UUV Localisation System Integrable with Aquaculture Infrastructure

Authors: Thein Than Tun, Loulin Huang, and Mark Anthony Preece Unmanned Underwater Vehicles (UUVs) are being used in aquaculture farms for difficult tasks like inspecting fish net-pens. They are mostly remote-operated, but using them autonomously faces challenges, mainly high costs and sensor reliability. A new system using April-Tags for localization offers a budget-friendly solution for testing UUV autonomy. It can achieve precise accuracy with minimal infrastructure changes. This could help demonstrate the benefits of autonomous UUVs to stakeholders for future adoption.

Click on the octopus to return to the top of the page

3 - Underwater Object Recovery Using a Hybrid-Controlled ROV with Deep Learning-Based Perception

Authors: Inés Pérez-Edo, Salvador López-Barajas, Raúl Marín-Prades, and Pedro J. Sanz Large ROVs/AUVs require costly support infrastructure, prompting interest in lightweight, modular alternatives. This article proposes a hybrid-controlled ROV architecture for known-object recovery, integrating autonomous perception, natural language interaction via a ROS-based agent, and low-level control for vehicle dynamics and autonomous behaviors like approach and grasping. Evaluated in simulation, a 12×8×5 m test tank, and harbor scenarios, the system successfully recovered a black box using a BlueROV2—demonstrating reduced operational risk, complexity, and cost for underwater interventions.

4 - Design and Development of a Sensor-Enhanced Remotely Operated Underwater Vehicle (ROUV) Platform for Environmental Monitoring

Authors: Dimitrios Tziourtzioumis, George Minos, Triantafyllia Anagnostaki, Eleftherios Kenanidis, and Theodoros Kosmanis This study introduces a modular, sensor-enhanced remotely operated underwater vehicle (ROUV) designed for environmental monitoring and aquaculture. Equipped with sensors for temperature, pH, dissolved oxygen (DO), and electrical conductivity (EC), along with real-time data acquisition, the system prioritizes modularity, reproducibility, and robustness. Field tests conducted at an aquaculture site in the North Aegean Sea during 2025 demonstrated depth-resolved measurements with high accuracy. The platform achieved stable and repeatable performance in real-world conditions.

5 - Underwater Antenna Technologies with Emphasis on Submarine and Autonomous Underwater Vehicles (AUVs)

Authors: Dimitrios G. Arnaoutoglou, Tzichat M. Empliouk, Dimitrios-Naoum Papamoschou, Yiannis Kyriacou, Andreas Papanastasiou, Theodoros N. F. Kaifas, and George A. Kyriacou This article provides a comprehensive review of antenna types for future underwater communication and sensing systems, evaluating their performance for Autonomous Underwater Vehicles. It examines magnetic induction coils, electrically short dipoles, wideband traveling-wave antennas, printed planar antennas, and magnetoelectric resonators, comparing footprint, operating frequency, bandwidth, and realized gain. The analysis highlights trade-offs between miniaturization and radiation efficiency and underscores the need for innovative designs to overcome underwater propagation limits.

8 - Reinforcement-Learning-Based Adaptive PID Depth Control for Underwater Vehicles Against Buoyancy Variations

Authors: Jian Wang, Shuxue Yan, Honghao Bao, Cong Chen, Deyong Yu, Jixu Li, Xi Chen, Rui Dou, Yuangui Tang, and Shuo Li This paper proposes a hybrid control framework combining Proximal Policy Optimization (PPO) with adaptive Proportional-Integral-Derivative (PID) tuning to address buoyancy-induced instability in sampling AUVs. PPO dynamically adjusts PID parameters online, while safety mechanisms protect actuators. A dual-error state and actuator command buffering mitigate system lag and inertia. Simulations and experiments (pool and field tests) show superior performance over conventional PID and pure PPO: faster convergence, lower steady-state error, and smoother control signals, demonstrating robustness under real hydrodynamic disturbances.

9 - Resistance Analysis of Remotely Operated Vehicle (ROV)

Indonesia's extensive underwater resources and infrastructure require advanced tools, such as Remotely Operated Vehicles (ROVs), for safe, efficient exploration and maintenance. This study investigates the hydrodynamic drag characteristics of the locally developed SN-01 ROV (Segara Nauta) using Computational Fluid Dynamics (CFD). The results highlight the critical impact of motion direction and speed on resistance, informing engine selection and service speed optimization. Ultimately, the findings support a more efficient, tailored ROV design for sustainable maritime operations in Indonesia.

6 - Recent Advances in Underwater Energy Systems and Wireless Power Transfer for Autonomous Underwater Vehicle Charging

Authors: Ramamurthi Sekar, Narayanamoorthi Rajamanickam, Hassan Z. Al Garni, Jamal Aldahmashi, and Ahmed Emara This article reviews recent advances in underwater wireless power transfer (UWPT) for autonomous underwater vehicles (AUVs), driven by growing demand for long-endurance offshore infrastructure maintenance and oceanographic surveillance. Conventional recharging methods hinder true autonomy, making UWPT a promising alternative. The paper surveys key UWPT components: marine power sources, power converter and compensation topologies, and magnetic coupler designs. It also addresses critical engineering challenges unique to underwater environments.

7 - ORCA: AI-powered Autonomous Underwater Vehicle for Subaquatic Exploration

Authors: Oskar Natan, Wiwit Suryanto, Andi Dharmawan, Rifda Hakimasari, Zaidan Hakim ORCA is an Autonomous Underwater Vehicle designed to perform missions independently, addressing navigation, exploration, and interaction with the underwater environment. The project emphasizes a complete development cycle from design to manufacturing, using Inventor for CAD and fabrication via Computer Numerical Control (CNC) machining, laser cutting, and 3D printing, while advancing control systems, Proportional Integral Derivative (PID) controllers, and vision. It also expands to environmental perception, object detection, deep learning, and path-planning, enabling autonomous navigation, obstacle avoidance, object detection, and payload manipulation with a gripper.

10 - A Control-Oriented Thruster Management Framework for Fault-Tolerant Propulsion of Remotely Operated Vehicles

Authors: Lu Wang, Yi Wu, Chao Fang, Jie Gao, Yonggang Gu, Chao Zhai, and Zhen Zhang This paper addresses limitations of conventional static thrust allocation in ROV propulsion, which struggles with real-world actuator imperfections like cavitation-induced thrust loss, dead zones, inter-thruster coupling, and partial failures. Instead of treating propulsion management as a static force distribution problem, the authors propose a control-oriented framework that embeds actuator dynamics and thrust execution uncertainty directly into the feedback loop. Experiments on an ROV platform show superior thrust executability, reduced coupling disturbances, and robust performance during thruster failure compared to traditional methods.

11 - Volumetric Path Planning and Visualization for ROV-Based Forward-Looking Sonar Scanning of 3D Water Areas

Authors: Yu-Cheng Chou, and Wei-Shan Chang This study introduces a deterministic, geometry-driven framework for volumetric path planning of ROVs equipped with multibeam forward-looking sonar in predefined circular and rectangular underwater volumes. It generates layered planar scan trajectories by integrating sonar parameters (detection range, horizontal/vertical beamwidths) and volume geometry, analytically estimates mission duration using path length and vehicle kinematics, and proposes a distance-based target position certainty metric, combined with the active sonar equation, to assess scan effectiveness and estimate likely target locations.

12 - Performance analysis of a micro underwater Remotely Operated Vehicle (ROV)

Authors: Fauzal Naim Zohedi, Chan Yeow Chuan, Mohd Shahrieel Mohd Aras, Alias Khamis, Zairi Ismael Rizman This paper presents an underwater Remote Operated Vehicle (ROV) designed to replace human divers in hazardous offshore tasks by mitigating risks and costs associated with underwater operations. This ROV incorporates a robust frame, waterproof endoscopic camera, and pressure/depth sensor to withstand submersion. Testing in a laboratory pool confirmed stable performance: 90% negative buoyancy enabled reliable descent and ascent, while consistent velocity and acceleration were achieved in forward, backward, and vertical motion. Angling the horizontal thrusters at 45° improved maneuverability and turning speed.

13 - A Multidisciplinary Optimization Method for the Wing of Autonomous Underwater Vehicle

Author: Qi Chen This paper presents a multidisciplinary design optimization method that integrates aerodynamic and structural analyses to enhance design collaboration. A parametric model of the AUV wing was developed using CATIA, a software suite created by Dassault Systèmes that is widely used for computer-aided design (CAD). Following this, grid generation was carried out in ICEM Surf, a powerful tool for advanced surface modeling. A computational fluid dynamics simulation was conducted to evaluate the hydrodynamic performance. The hydrodynamic loads were then analyzed and optimized, resulting in a reduction of the wing structure's mass while maintaining safety. This approach provides a reliable pathway for improving AUV performance and design efficiency.

14 - Current Status and Prospects of Metrological Calibration Technologies for Underwater Robot Environmental Perception Systems

Authors: Yunyue Chen, Yuan Huang, Jiadongmei Yang, Jingzhou Cao The metrological calibration of environmental perception systems for underwater robots is essential for accurate marine data. Current challenges include interference in measurements and a lack of standard calibration procedures. Future technology aims to develop an intelligent underwater metrology system using Digital Twin, AI for sensor error prediction, and Mobile Calibration Platforms for in-situ value transfer.

18 - 3D Trajectory Tracking Based on Super-Twisting Observer and Non-Singular Terminal Sliding Mode Control for Underactuated Autonomous Underwater Vehicle

Authors: Zehui Yuan, Long He, Ya Zhang, Shizhong Li, Chenrui Bai and Zhuoyan Qi This paper discusses a control method for three- dimensional trajectory tracking in underactuated autonomous underwater vehicles that accounts for external disturbances and model uncertainties. It integrates backstepping dynamic surface control, a non- singular terminal sliding mode, and virtual velocity commands to stabilize positional errors while addressing the "differential explosion" issue using first-order low-pass filters to smooth derivatives. A non-singular terminal sliding surface enhances the convergence rate and robustness, allowing finite-time convergence of velocity- tracking errors. A disturbance observer algorithm provides high-precision disturbance estimation.

15 - PIRATE - Precision Imaging Real-Time Autonomous Tracker & Explorer

Authors: Dan Zlotnikov, and Ohad Ben-Shahar PIRATE (Precision Imaging Real-time Autonomous Tracker and Explorer) is an unmanned surface vehicle that autonomously collects data and tracks mobile underwater targets. It integrates acoustic localization, visual perception, and closed-loop navigation. Using a mobile acoustic receiver, PIRATE estimates target positions through time- difference-of-arrival (TDoA) measurements, guiding its adaptive behavior. The system synchronizes navigation, acoustic monitoring, and visual processing dynamically based on mission context.

16 - Numerical Investigation on Hydrodynamic Characteristics of Variable Flexible Tube Underwater Object Suction Robot

Authors: Yida Zhu, Fenglei Han, Qing Chang, Wangyuan Zhao, Shuxuan Liang, and Jiaqi Yu Remotely operated underwater vehicles (ROVs) are crucial in underwater robotics, particularly in deep-sea aquaculture. Conventional suction-tube robots face challenges in accurately collecting target organisms in complex reef environments. This paper introduces an underwater object suction robot with a variable flexible tube. It highlights the importance of stable vehicle motion for vision-based tasks, as hydrodynamic disturbances can impair visual accuracy. A systematic numerical investigation using computational fluid dynamics (CFD) simulations assesses the ROV's hydrodynamic characteristics and provides insights for motion control strategies to enhance visual robustness.

17 - Study and Feasibility of Underwater Acoustic Data Transmission

Authors: Bessie A. Ribeiro, Fabio C. Xavier, Viviane R. Barroso, Viviane F. da Silva, Theodoro A. Netto, and Caroline Ferraz The increasing demand for deep-water offshore oil and gas production has driven the need for continuous and cost-effective monitoring of subsea equipment. Traditional methods using ROVs and AUVs incur delays and high operational costs. This study explores the feasibility of underwater acoustic communication systems for real-time data transmission, utilizing the BELLHOP ray-tracing model to simulate transmission loss, multipath effects, and signal- to-noise ratio (SNR) under various conditions.

19 - UMI-Underwater: Learning Underwater Manipulation without Underwater Teleoperation

Authors: Hao Li, Long Yin Chung, Jack Goler, Ryan Zhang, Xiaochi Xie, Huy Ha, Shuran Song, Mark Cutkosky Underwater robotic grasping is difficult due to degraded imagery and costly data collection. This system autonomously gathers underwater grasp demonstrations via self-supervision and transfers on-land knowledge through depth-based affordances that bridge domain gaps and resist lighting shifts. An affordance model deploys zero-shot underwater via geometric alignment, then trains an affordance-conditioned diffusion policy. Pool experiments show gains in grasping performance, background robustness, and generalization to land-only objects, outperforming RGB baselines.

20 - Flexible hot-film sensors for underwater wall shear stress measurement

Authors: Ziqiang Xiang, Qingshan Wu, Guanjun Liu This study introduces a flexible hot-film sensor for underwater wall shear stress (WSS) measurement, addressing the limitations of conventional techniques, including intrusiveness, low spatial resolution, poor adaptability to curved surfaces, and high cost. Fabricated via a low-cost screen-printing process using conductive carbon paste on a polyimide substrate with waterproof packaging, the sensor was integrated into a Constant Temperature Anemometer (CTA) circuit. Calibration in a turbulent channel flow, using the pressure drop method and Darcy-Weisbach equation, demonstrated high performance: the carbon paste showed a stable TCR of 3347 ppm/C degrees, a calibration curve with R² = 0.971, a resolution of 0.05 Pa, a response time under 0.35 s, and excellent cyclic stability.

21 - Neural Network-Augmented Actuation Control System Designed for Path Tracking of Autonomous Underwater- Transportation Systems Under Sensor and Process Noise

Authors: Faheem Ur Rehman, Syed Muhammad Tayyab, Hammad Khan, Aijun Li, and Paolo Pennacchi Underwater-transportation systems hold significant potential for military and commercial use. This study applies Neural Network-Augmented Control (NNAC) to three autonomous underwater systems: Rigid-Connection (RCTS), Flexible-Connection (FCTS), and Leader–Follower- Formation (LFFCTS) transportation systems. Using Extended Kalman Filter (EKF) for state estimation under noise, NNAC demonstrates robust, adaptive control with superior trajectory tracking compared to PID controllers. Among configurations, RCTS achieves the highest tracking accuracy with lowest power consumption, while FCTS consumes the most power with comparatively lower accuracy.

24 - Environment-Aware Optimal Placement and Dynamic Reconfiguration of Underwater Robotic Sonar Networks Using Deep Reinforcement Learning

Authors: Qiming Sang, Yu Tian, Jin Zhang, Yuyang Xiao, Zhiduo Tan, Jiancheng Yu, and Fumin Zhang This paper introduces a deep reinforcement learning (DRL) framework for adaptive sonar placement and dynamic reconfiguration of distributed AUV networks in underwater dynamic target detection, classification, localization, and tracking (DCLT).

22 - Collision tests and data acquisition analysis of unmanned underwater vehicles based on multi-sensor monitoring

Authors: Guanlun Guo, Hao Liu, Min Jiao, Li Min, Wei Zhou, Hai Tan, and Hui Cao This study examines UUV large-angle collision dynamics underwater via experiment-data-simulation methods with multi-sensor monitoring and fluid-structure models. Results show axial acceleration drops sharply and load paths restructure; at 60°, underwater strain falls to one-third of air values while duration exceeds 30 ms. Fluid synergy (lubrication, inertia, pressure) optimizes housing response. The validated scheme supports anti-shock design and structural health monitoring for marine equipment.

23 - Numerical investigation of thermal wake evolution behind an underwater vehicle in stratified flow

Authors: Bo Huang, Zhongyan Liu, Xu Liu, Hongyu Fu, Qi Zhang, Mengchun Pan, Jiafei Hu A three-dimensional numerical model was developed to study the thermal wake of an underwater vehicle in stratified flow under the Boussinesq approximation. The study found that background stratification suppresses vertical wake diffusion, causing the thermal wake to elongate primarily downstream. Propeller rotation and wake entrainment create a distinct temperature-anomaly structure that intensifies in the near wake, peaks downstream, and then decays due to diffusion. The findings aid in predicting thermal signatures and identifying wakes of underwater vehicles in complex ocean environments.

25 - Unified Stochastic Differential Equation Modeling and Fuzzy-RL Control for Turbulent UWOC

Authors: Bowen Si, Jiaoyi Hou, Dayong Ning, Yongjun Gong, Ming Yi and Fengrui Zhang Underwater wireless optical communication (UWOC) for AUVs suffers from the coupled effects of ocean turbulence and platform motion, which static models fail to capture over time, thereby impairing beam tracking. The paper proposes a dynamic differential equation (SDE) model derived from the Radiative Transfer Equation via a diffusion approximation, validated by inverse Fokker–Planck methods, and a hybrid Fuzzy-Q-learning controller with a supervisor for deep fades. Simulations and HIL experiments yield a 3.64 mm median alignment error and >80% reduction in transient error compared with PID, enhancing link stability.

26 - MCHS-SLAM: A Multi-Constraint Hybrid Strategy SLAM Framework for AUV-Based Seafloor Terrain Mapping

Authors: Jianan Qiao, Bin Liu, Yan Huang, Jiancheng Yu, Xiaolong Ju, and Hao Feng Autonomous Underwater Vehicles (AUVs) face inevitable positioning error accumulation during seafloor terrain mapping due to the lack of underwater satellite navigation. Additionally, the complex seafloor topography with alternating flat and undulating regions degrades the performance of traditional single-constraint SLAM methods. To overcome these issues, a Multi-Constraint Hybrid Strategy (MCHS) Simultaneous Localization and Mapping (SLAM) framework is proposed, which uses a hierarchical constraint architecture to optimize pose estimation, local geometric consistency, and global loop closure detection.

27 - Near-Bottom ROV-Borne Self-Potential Exploration of Seafloor Massive Sulfi de Deposits on the Southwest Indian Ridge

Authors: Zuofu Nie, Chunhui Tao, Zhongmin Zhu, and Jianping Zhou This study used a remotely operated vehicle (ROV) to conduct a high-resolution self-potential (SP) survey at the Yuhuang hydrothermal field on the Southwest Indian Ridge, flying at a near-bottom altitude of ~10 m to detect seafloor massive sulfide (SMS) deposits. An adaptive octree mesh enabled efficient 3D inversion, revealing a main anomaly zone correlated with known mineralization and an additional zone suggesting undiscovered deposits. Results aligned well with independent near-bottom observations and drilling data.

28 - Research on Real-Time Trajectory Planning and Tracking Control for Multi-ROV Shipwreck Search

Authors: Wenyang Gan, Haozhe Liang and Caixia Cai Multi-ROV cooperative systems face two key challenges in sunken ship search missions: overlapping operational regions cause redundant coverage and blind zones, while hydrodynamic disturbances and actuator limits degrade tracking accuracy. To address these, a dynamic window approach (DWA) with search distance penalties is introduced for trajectory planning, paired with a cascaded control architecture, comprising a model predictive kinematic controller and an adaptive sliding mode dynamic controller with an extended state observer, for robust disturbance rejection.

29 - Deep Reinforcement Learning for Autonomous Underwater Navigation: A Comparative Study with DWA and Digital Twin validation

Authors: Zamirddine Mari, Mohamad Motasem Nawaf, and Pierre Drap Underwater autonomous navigation faces challenges like GPS absence, poor visibility, and obstacles. Using the BlueROV2 platform, this study applies deep reinforcement learning (PPO) with an 84-dimensional observation space combining target navigation, occupancy grids, and raycasting for situational awareness. Compared to the traditional Dynamic Window Approach (DWA), the PPO policy excels in cluttered environments due to superior local adaptation and fewer collisions. Validation in both simulation and real-world tests (with a 3D digital twin for safety) confirms the policy’s transferability from simulation to reality, highlighting deep RL’s potential for underwater robotics.

30 - Multi-Objective Robotics Optimization Using Improved MO-BxR Algorithms

Authors: Ravipudi Venkata Rao, Harishankar Morazha Variam, and Joao Paulo Davim This paper addresses multi-objective robotics optimization, critical for enhancing robotic performance, efficiency, and reliability, by enhancing parameter-free BxR metaheuristics with three novel variants: archive-based, opposition-based, and self-adaptive multi-population (SAMP). Evaluated on five real-world robotic problems, including AUV shape design and manipulator inverse kinematics, the methods are rigorously benchmarked using GD, IGD, SPC, and HV metrics, with statistical validation via Friedman tests, Conover post hoc analysis (Holm-corrected), and Vargha–Delaney A12 effect sizes over 30 runs.

31 - A Bayesian Optimization-Based AUV Swarm Model in a Double- Gyre Flow Field

Authors: Tengfei Yang, Ziwen Zhang, Guoqiang Tang, Yan Yang, Qiang Zhao, Hao Wang, Minyi Xu, and Shuai Li Conventional multi-AUV control relies on quasi-steady assumptions and empirical tuning, causing fragmentation and low polarization under flow disturbances. The proposed DGF-OAS model uses heading-aware graph attention for adaptive weighting, Lennard-Jones potentials for spacing, embedded double-gyre flows, and Bayesian optimization of alignment/repulsion terms. Simulations yield 96% polarization, 15.84% faster task completion, and 97% success rate, outperforming baselines.

32 - Slope Terrain Gait Planning and Admittance Control Method for Underwater Quadruped Robots Based on Righting Moment compensation

Authors: Kang Zhang, Hao Zhang, Hong Chen, Guanqiao Chen, Zongxia Jiao, Yuang Zhang, Wei Chen, Xinliang Wang, and Junjie Liu Benthic AUVs, underwater quadruped robots, combine the efficiency of submersibles with the stability of legged robots for deep-sea exploration. However, their design creates a righting moment that resists hull alignment on steep slopes, leading to propulsion loss when exceeding hydrostatic pitch limits. To address this, the paper introduces a control framework featuring optimal force distribution, adaptive trajectory probing, and admittance control.

33 - Adaptive Learning from Quantized Signals for AUV Formation Tracking Control

Authors: Chao Wang, Xiaolei Li, Pengfei Yang, Jiange Wang, and Yuzhong Wang This paper addresses AUV formation tracking under quantized communication/actuation via an adaptive learning framework extracting policies from quantized data. Integrated into backstepping with quantized consensus and hysteresis compensation, Lyapunov analysis proves bounded signals and convergent errors. Simulations confirm accurate tracking and adaptability.

34 - A Survey of Machine Learning Algorithms for Autonomous Vehicles

Authors: Agnieszka Lazarowska, Monika Rybczak, Mirosław Lacki, Krystian Kozakiewicz, Jozef Lisowski, and Andrzej Stateczny This review examines machine learning algorithms applied to autonomous vehicles (UUVs, USVs, UAVs, and ground robots) from 2020–2026. It covers four functional areas: environment perception, Simultaneous Localization and Mapping (SLAM), collision avoidance/path planning, and motion control. The analysis includes supervised, unsupervised, semi-supervised, reinforcement, and deep reinforcement learning methods, evaluating their performance and robustness across underwater, surface, aerial, and land environments. The paper identifies key challenges and proposes future directions for enhancing autonomous vehicle safety, autonomy, and reliability.

35 - Hovering Control of ROV Based on Event-Triggered Image Moments Nonlinear Model Predictive Control

Authors: Ziyang Shen, Yunxiu Zhang, Shengguo Cui, Peng Zhang, and Siyue Wang This paper proposes an event-triggered image-moment nonlinear model predictive control (ETIM-NMPC) method for vision-based hovering control of remotely operated vehicles (ROVs). To overcome limitations in constraint handling and excessive online optimization in conventional approaches, ETIM-NMPC uses image moments as visual features, operates in a reduced control- relevant subspace, and updates the control sequence only when an event-triggering condition is met. Evaluated in four simulation scenarios (Ideal, Disturbed, Dynamic- target, Complex), ETIM-NMPC achieved robust hovering with translational RMSE < 0.15 m in all directions—meeting the required accuracy.

36 - Quasi-Static In Situ Deep Learning for Forward-Looking Sonar Target Detection in Complex Underwater Environments

Authors: Yixuan Chen, Zhenqing Ding, Yu Feng, Jiale He, Ziqin Xie, Tinggang Xiong, Kai Chen, and Qi Gao Forward-looking sonar (FLS) target detection for autonomous underwater vehicles (AUVs) faces challenges from acoustic distortions, environmental volatility, and limited annotated data, limiting standard deep learning approaches. This study introduces a quasi- static in situ learning paradigm for underwater acoustic target detection (UATD), combining scene priors with a lightweight deep learning detector through echo- intensity-based probability weighting and acoustic attenuation compensation. The method enhances pixel- wise images and fuses statistical descriptors with deep learning predictions for dynamic environmental adaptation.

37 - Numerical Investigation of Hydrodynamic Performance of an AUV Moving near the Bottom Wall

Authors: Nguyen Dong and Ngo Van He This study uses CFD simulations to analyze how near- bottom operation affects the hydrodynamics of an AUV, comparing open-water conditions with a near-bottom case (clearance ratio h/LA = 1.93). Results show that proximity to the seabed alters pressure fields, wake structures, and increases total resistance and flow separation. These findings help improve AUV design, control, and operation for shallow-water or seabed missions.

38 - BDAT-Planner: Bioinspired Dynamic Adaptive Threshold Planner for Underwater Collision Avoidance of AUVs

Authors: Boyang Zhang, Zhicheng Zhang and Weixing Feng This paper introduces the BDAT-Planner, a bioinspired dynamic adaptive threshold framework for autonomous underwater vehicles (AUVs) to achieve robust, intelligent collision avoidance in complex, unknown underwater environments. Unlike conventional spiking methods that rely on fixed thresholds, limiting adaptability to dynamic interference, the proposed BDAT mimics biological neural systems by enabling per-neuron dynamic threshold adjustment in the spiking dynamic adaptive actor network (SDAAN), thereby maintaining neural homeostasis and balancing excitation/inhibition.

39 - Mathematical Modeling and Simulation of a Hybrid additive–Subtractive ROV with Experimental Validation for Reef Exploration

Authors: Miguel F. Delgado-Pamanes, Bruno Silva-Morales, Luis A. Reyes-Osorio, Octavio Garcia-Salazar, Marisol Vázquez-Tzompantzi, and Aldo Jonathan Munoz-Vazquez This paper presents a mathematical and numerical framework for designing and analyzing an ROV for shallow-water reef exploration. The open-frame vehicle uses a sealed housing and four-thruster system for omnidirectional, low-speed maneuverability. Hydrodynamics are modeled via incompressible RANS equations solved numerically, with thruster flow captured through an MRF formulation that resolves thrust directly without artificial sources.

40 - Object manipulation of the variable topology truss system

Authors: Andrew Jang-Ho Bae, Myeongjin Choi, Haorui Li, Mark Yim, TaeWon Seo This paper introduces an object manipulation strategy for the Variable Topology Truss (VTT) system, a truss robot with actuated members and passive joints. The proposed hybrid control framework regulates both position and force simultaneously without explicit decoupling. At the actuator level, a force feedback controller compensates for friction, while at the task level, a static model computes member forces to achieve desired end-effector forces. Experiments on a single module and full system validate force-tracking performance, and two configurations demonstrate object manipulation with combined position and force tracking.

41 - A Study of the Three-Dimensional Localization of an Underwater Glider Hull Using a Hierarchical Convolutional Neural Network Vision Encoder and a Variable Mixture-of-Experts Transformer

Authors: Jungwoo Lee, Ji-Hyun Park, Jeong-Hwan Hwang, Kyoungseok Noh, and Jinho Suh This paper introduces a deep learning-based perception framework for detecting underwater gliders and estimating their 3D pose (position and heading) using camera-sonar fusion. To address challenges in harsh underwater environments, such as visual degradation and motion disturbances, it combines a hierarchical CNN vision encoder (for robust, multi-level feature extraction) with a transformer architecture enhanced by a variable mixture-of-experts mechanism (to boost representational power without sacrificing efficiency). The system enables precise, collision-free ROV navigation for automated glider recovery and sensor inspection.

42 - Multi-Source Sensor Fusion Localization Method for Autonomous Underwater Vehicles Based on Deep Learning

Authors: Xin Pan, Guoli Feng, Haiyan Zeng, and Qunhong Tian This study introduces a deep-learning-based multi-sensor fusion framework to enhance Autonomous Underwater Vehicle (AUV) localization, addressing limitations of single- sensor and traditional multi-sensor methods. The framework integrates high-frequency data from INS and DVL for continuous position propagation, while USBL and Sonar provide periodic corrections. Three models (DNN, LSTM, BSsMSLNN) were evaluated with results showing the proposed framework reduces long-term error accumulation and improves localization accuracy. The BSsMSLNN-based method achieved the best performance in trajectory fitting, RMSE, and R², offering a viable solution for high-precision AUV navigation in GPS- denied environments.

43 - E2E-AUD: An End-to-End Adaptive Underwater Detection Framework Integrating Physical Priors and Frequency-Adaptive Learning.

Authors: Wenhao Zhou, Junbao Zeng, Shuo Li, and Yuexing Zhang Underwater detection is critical for Autonomous Underwater Vehicles (AUVs) but faces challenges like image degradation, target deformation, and dense small objects. Traditional methods separate image enhancement and use fixed-shape convolutions, leading to poor handling of irregular targets and fine details. The proposed E2E-AUD framework integrates a lightweight enhancement module (UnitModule) for joint optimization, employs linear deformable convolution (LDConv) for adaptive target modeling, and uses Haar wavelet downsampling (HWD) to preserve texture details.

44 - Design, Kinematic Analysis and Experimental Validation of a New Graded Guidance and Locking Mechanism for Deepwater Multi-Way Quick Connector

Authors: Haixia Gong, Wei He, Qin Si, Yusong Dai, Fuqiang Zu and Liquan Wang This study introduces a novel deep-water multi-way quick connector (MQC) prototype featuring a tiered tolerance guidance mechanism, an L-shaped helical cam locking system, and a real-time visual attitude indicator. Finite element analysis and physical testing demonstrated its capability to handle extreme misalignments (25 mm offset, 5° horizontal and 15° axial rotation) at 10 mm/s docking speed without exceeding elastic stress limits. The L-shaped locking system, analyzed using advanced contact mechanics models, eliminated impact forces and maintained plastic strain (0.00926) well below the ASME failure threshold (0.0865), even under 10,000 psi separation loads.

45 - Direct X-Rudder Path-Following Control for Underactuated AUVs via TIB-CSAC

Authors: Jiehui Tan, Yushan Sun, Liwen Zhang, Puxin Chai, and Zhan Liu This study introduces a direct X-rudder control method for underactuated autonomous underwater vehicles (AUVs) using Task-Informed Inductive-Bias Conservative Soft Actor–Critic (TIB-CSAC) to enhance path-following performance under varying path geometries and velocities. The method learns control policies directly from current and subsequent path segments, eliminating the need for complex guidance laws and controller tuning. A task-informed inductive-bias encoder improves state representation efficiency, while a multi-head conservative value evaluation mechanism addresses long-tail task difficulties during training.

46 - YOLO-CAB: An Efficient Deep Learning-Based Underwater Object Detection Method for Autonomous Underwater Vehicles

Authors: Runze Li, Changdong Yu, Shuaiyu Bao, Zijian Li, and Jinyi Yao Existing underwater object detection methods struggle with low recall and inaccurate localization due to light scattering, absorption, and low target-background contrast. To overcome these challenges, The authors propose YOLO-CAB, a YOLOv13-based detector that enhances feature extraction, boundary perception, and localization accuracy. Key innovations include the Context- Aware Large Selective Kernel (CALSK) module for adaptive spatial feature enhancement, the Spatial Boundary Attention Module (SBAM) for refining target boundaries, and the Momentum-based Category-Aware Weighted Intersection over Union (MCAWIoU) loss to improve localization and confidence for challenging samples.

47 - Underwater Robot Object Detection Algorithm Based on YOLOv11

Authors: Yongqing Shi, Wei Chen, Duo Wan and Lu Han Underwater target recognition is challenging due to scattering, absorption, and illumination variability. This study introduces a degradation-aware YOLOv11s-based framework to enhance underwater robotic detection by improving feature reconstruction, attention mechanisms, and bounding-box regression. Key innovations include: (1) Spatial Channel Reconstruction Convolutions (SCConv) to reduce redundancy and capture target deformations, (2) Shuffle Attention to refine fine-grained features, and (3) Focaler-IoU for stable regression.

48 - Target product grabbing of factory robotic arm based on digital twin technology

Authors: Lu Che, Xin Xu, Weipeng Mu, and Peng Wang In smart manufacturing, factory robotic arm target grabbing is critical for material transfer, yet current systems suffer from low gripping accuracy and weak anti- interference. To address this, this study proposes a digital twin-based robotic arm grabbing model, leveraging virtual-real collaborative optimization. The method integrates cross-platform communication (via rosSocket middleware), a high-fidelity digital twin (built with SolidWorks/Unity3D), collision detection for planning efficiency, and composite algorithms (RRT-Connect + tool calibration).

49 - Development of a Bio-robotic Swimmer Based on the Maneuverability of the California Sea Lion

Authors: Nicholas Marcouiller, Shraman Kadapa, Anthony Drago, Frank E. Fish, Magan L. Leftwich, Harry G. Kwatny, and James L. Tangorra The development of unmanned underwater vehicles (UUVs) for complex environments, like coastal regions, requires new maneuvering techniques. Researchers have developed a bio-robotic system modeled after the California sea lion, focusing on its swimming and maneuverability. The system features an articulatable head and pelvis, flexible fore flippers generating 3D forces, and adjustable hind flippers. Tests show the system can use hydrostatic and hydrodynamic forces to move in 3D space, serving as a research platform for evaluating how body articulation and flipper movements influence underwater locomotion.