Reliable in situ quantification of trace metals in the deep ocean is critically important for understanding marine biogeochemical cycles, yet remains analytically challenging due to extreme hydrostatic pressure, large temperature gradients, low analyte concentrations, and severe matrix interferences. Here, we report a closed-loop square-wave anodic stripping voltammetric (SWASV) analyzer that integrates microelectrode-array sensing, environmental regulation, and signal-stability feedback to enable quantitative trace-metal measurements under deep-sea conditions. The system combines MEMS-fabricated iridium microelectrode arrays with an in situ renewable mercury film, active dissolved-oxygen removal, temperature-normalized signal correction, and background-subtracted voltammetry, forming a closed-loop analytical workflow that continuously monitors and compensates for environmental and instrumental perturbations. Under simulated deep-sea conditions, temperature-induced signal deviations exceeding 50% were reduced to within ±10-15%, while baseline noise and drift were effectively suppressed. The analyzer was subsequently deployed during two deep-sea cruises in the western Pacific, where in situ concentrations of Zn(II), Pb(II), and Cu(II) were quantified at depths up to 1800 m. Independent validation against inductively coupled plasma-mass spectrometry (ICP-MS) showed good agreement, with deviations generally within ±15%. These results demonstrate that closed-loop electrochemical regulation provides a robust pathway toward long-term, unattended, and quantitative electrochemical observations in extreme marine environments.

To address the challenge of efficient autonomous navigation for six-degree-of-freedom Unmanned Underwater Vehicles (UUVs) in complex, time-varying flow environments, this paper proposes FINDER (Flow-aware Intelligent Navigation through Distilled Experience and Reinforcement Learning), a novel algorithm based on deep reinforcement learning (DRL). This framework mitigates the reliance of existing methods on prior knowledge or direct perception of flow fields through a two-stage training paradigm. First, in the Flow-aware Policy Learning (FAPL) stage, a teacher policy is trained in a high-fidelity simulation environment where UUV perceives local flow field information. Then in the Motion History Distillation (MHD) stage, knowledge distillation is utilized to transfer the flow field perception and decision-making capabilities acquired by the teacher policy to a lightweight student policy. The MHD policy relies solely on the UUV's own motion history data for decision-making, eliminating the need for external flow field sensors and significantly reducing deployment complexity and cost. This approach achieves an inference time of approximately 3 ms, demonstrating a 113-fold increase in computational efficiency over existing state-of-the-art methods based on flow field prediction. Such performance satisfies the requirements for high-frequency decision-making in equipped processors.

The underwater wireless charging system can improve the endurance of autonomous underwater vehicles (AUVs) and extend their detection range. However, the magnetic couplers for the AUV wireless charging system are currently not very compatible with AUVs, have high electromagnetic interference, and have low power levels. The new asymmetric arc-shaped magnetic coupler proposed in this paper can fit the surface of an AUV without changing its hydrodynamic model, making it suitable for use in the AUV wireless charging system. The radial coupling scheme can effectively prevent electromagnetic interference of the internal components by limiting the magnetic flux outside the AUV. By building the magnetic circuit model of the magnetic coupler and optimizing its structural parameters, the magnetic coupler has good misalignment tolerance performance. To test the effectiveness of the magnetic coupler, a practical 3 kW prototype has been developed. Experiments show that the system can achieve effective power transmission under AUV rotational misalignment +/- 20 degrees and axial misalignment +/- 20 mm.

The deep-sea human occupied vehicles (HOV) typically have the characteristics such as large dimensions, complex structures, and significant nonlinearities of their dynamic models, which makes it difficult for the operator to precisely control the vehicle for completing the tasks such as target tracking, grabbing, and specific area exploration. Establishing an accurate hydrodynamic model and obtaining an appropriate control algorithm are the key to solving the aforementioned problems. Therefore, this paper takes the 7000-meter deep-sea human occupied vehicle “Jiaolong” as the research object. The accurate hydrodynamic coefficients of the vehicle are obtained based on the maneuvering test data of the vehicle, and a complete six-degree-of-freedom model of the vehicle is established. The precision of the dynamic model is verified by the test data of Pacific deep submerging. The simulation studies of typical maneuvering conditions are conducted to demonstrate that the vehicle possesses good maneuverability. By utilizing the dynamic model, the control strategy for the HOV is derived using the control algorithm of adaptive integral sliding mode (AISMC). The simulations confirm that this control method can significantly enhance the control accuracy of the vehicle, and the challenges posed by model uncertainties of the model and external disturbances of the environment are effectively addressed.

A nonlinear large deformation cable equation system based on arc length is employed to determine the configuration of the cable under current and buoyancy loads. Different from a traditional initial method and a shooting method, we have solved this nonlinear problem as a boundary value problem with nonlinear and global boundary conditions directly. A finite difference scheme is proposed to solve the large deformation cable equation, and the Newton–Raphson iteration is used to search for numerical approximate solution. We demonstrate that this system degenerates into the catenary equation in the case of vanishing bending stiffness for the first time. The solution of the catenary equation serves as the initial guess for the three-section cable problem. This method overcomes the disadvantages of the initial value method and step method, avoiding the need to adjust the boundary location. The spatial shapes of the large deformation cable and the effects of the length and position of the buoyancy section are discussed. The impact of ocean currents is also analyzed using Morison's formula.

In order to solve the possible track deviation problem of autonomous underwater vehicle (AUV) in the course of linear track tracking, we developed a velocity three closed loop motion control algorithm combined with distance information, which is specially used for track tracking. This algorithm is based on line-of-sight guidance strategy and introduces a time-varying forward-looking distance to enhance the flexibility and response speed of the AUV. In the design of the control algorithm, we integrate the advantages of PID control, fuzzy PID control and parameter adaptive adjustment, so as to improve the adaptability and robustness of the algorithm. In particular, we use a range three closed loop speed control strategy, which means that the AUV dynamically adjusts its desired speed during sailing based on the real-time distance deviation from the end of the desired track segment. With this design, the AUV is able to slow down ahead of time and make low-speed turns, when necessary, effectively avoiding excessive course Angle adjustment. In addition, the parameter adjustment mechanism in the algorithm enables it to flexibly cope with various navigation conditions. In the case of large Angle turns, the AUV can still accurately travel along the intended track, significantly reducing the phenomenon of track deviation.

Effective docking technology is crucial for ensuring the sustained underwater autonomous operation of Autonomous Underwater Vehicle (AUV). However, the presence of ocean current disturbances and nonlinear dynamic properties of AUV pose significant challenges, making many existing docking control algorithm less effective. In order to improve the AUV docking capability in the complex environment, a novel model-free Deep Reinforcement Learning (DRL) based docking algorithm, named Adaptive Reward Shape Proximal Policy Optimization (ARSPPO), is proposed. A new Marcov Decision Process (MDP) modeling based on the novel docking quality evaluation metric of docking task is also introduced, Additionally, a large-scale parallel simulation framework is developed specifically for the rapid training of DRL for underwater vehicles. Simulation results indicate that the proposed algorithm has better sample efficiency and robustness against ocean current with improved docking performance. The real-world experiments further validate its adaptability and potential applicability in complex ocean scenarios.

This study introduces MarineGym, a high-performance reinforcement learning platform tailored for underwater robotics. It aims to address the limitations of existing underwater simulation environments in terms of reinforcement learning compatibility, training efficiency, and standardized benchmarking. MarineGym integrates a proposed GPU-accelerated hydrodynamic plugin based on Isaac Sim, achieving a rollout speed of 250,000 frames per second on a single NVIDIA RTX 3060 GPU. It also provides five models of unmanned underwater vehicles, multiple propulsion systems, and a set of predefined tasks covering core underwater control challenges. Additionally, the domain randomization toolkit allows flexible adjustments of the simulation and task parameters during training to improve the Sim2Real transfer. Further benchmark experiments demonstrate that MarineGym improves training efficiency over existing platforms and supports robust policy adaptation under various perturbations in the marine environment. We expect this platform to drive further advancements in RL research for underwater robotics. For more details about MarineGym and its applications, please visit our project page: https://marine-gym.com/.

Trajectory tracking for underactuated autonomous underwater vehicles (AUVs) is challenging due to coupling dynamics, modeling inaccuracies, and unknown disturbances. To tackle this, we propose a decoupling gaze-assisted prescribed performance controller (GAPPC). We first use an error transformation approach to achieve the prescribed performance, incorporating the line-of-sight (LOS) algorithm and an event-triggering mechanism to handle the kinematic characteristics of underactuated AUVs. Next, we develop a control strategy for the transformed error that does not require knowledge of the model parameters, including fast dynamic compensation to reduce steady-state errors. Finally, we analyze the controller’s stability and present simulation results. Simulations, which account for modeling inaccuracies and unknown ocean currents, show that the GAPPC improves stability errors by 67.3% compared to the adaptive robust controller.

IntroductionAutonomous Underwater Vehicles (AUVs) are capable of independently performing underwater navigation tasks, with side-scan sonar being a primary tool for underwater detection. The integration of these two technologies enables autonomous monitoring of the marine environment.MethodsTo address the limitations of existing seabed detection methods, such as insufficient robustness and high complexity, this study proposes a comprehensive seabed detection method based on a sliding window technique. Additionally, this study introduces a sonar image stitching method that accounts for variations in image intensity and addresses challenges arising from multi-frame overlaps and gaps. Furthermore, an autonomous target perception framework based on shadow region segmentation is proposed, which not only identifies targets in side-scan sonar images but also provides target height measurements.ResultsComprehensive seabed detection method improves accuracy by 31.2% compared to the peak detection method. In experiments, the height measurement error for this method was found to be 9%.DiscussionTo validate the effectiveness of the proposed seabed detection method, sonar image stitching method, and target perception framework, comprehensive experiments were conducted in the Qingjiang area of Hubei Province. The results obtained from the lake environment demonstrated the effectiveness of the proposed methods.

In deep-sea navigation environments, underwater vehicles are often subject to interference from ocean currents. This article takes a certain underwater vehicle as the research object. Firstly, the six degree of freedom motion equation under the ocean current is given. Then, simulations are carried out on the straight sailing motion, turning motion, and Z-shaped motion under cross flow to study the performance of underwater vehicles under transverse ocean currents. The research results indicate that the motion of underwater vehicles has strong coupling and nonlinear characteristics, and the navigation is stable in the absence of ocean currents; Under the interference of lateral ocean currents, the navigation trajectory deviates, and the longitudinal speed and longitudinal inclination angle have little effect. The overall navigation is stable, and the presence of lateral ocean currents will not damage the stability of straight, turning, and Z-shaped movements, but will cause significant lateral displacement. Due to the asymmetric shape of the underwater vehicle, it will also cause small vertical displacement; This study can provide reference for the operation and control of underwater vehicles under ocean currents.

Environmental monitoring plays a crucial role in the development of marine ranches and the surveillance of underwater aquaculture organisms. To capitalize on the real-time, long-term, and static observation capabilities of seabed networks, as well as the dynamic and large-scale monitoring potential of underwater vehicles, a novel mobile platform for ocean ranches has been proposed. This platform comprises a floating platform, a docking station, and an autonomous underwater vehicle (AUV). The floating platform utilized is a versatile ocean testing platform that can be securely anchored in close proximity to the designated observation area. To enable static monitoring alongside the floating platform, a lightweight connection station, constructed using polyvinyl chloride pipes, is designed to accompany the platform. The AUV is employed for dynamic monitoring and is seamlessly linked to the aforementioned components using docking technology. Consequently, this integrated system achieves dynamic and static observations centered around a movable floating platform. Field experiments conducted in lakes and seas have validated the efficacy of this system in multiple scenarios, both on the surface and underwater. These experiments have demonstrated the system's ability to autonomously dock, transmit wireless signals and power, facilitate long-term static observations of fixed nodes, and conduct autonomous cruising for dynamic monitoring purposes.

Autonomous Underwater Vehicle (AUV) has become an essential tool for ocean environmental observation due to their cost-effective integration of various sensors. Deep Reinforcement Learning (DRL) offers a promising solution for complex underwater problems, yet encounters limitations in sample efficiency. This paper presents a distributed parallel framework utilizing multi-threading and parallel simulations within Gazebo to overcome the constraint, leading to a significant acceleration in the training process. The framework can be used as a general training simulation platform for training policy to control underwater vehicle, helping improving autonomous ability of AUV facing complex underwater environment through state-of-the-art DRL algorithm. To verify its capability, a challenging docking control policy for AUV is trained using the Proximal Policy Optimization (PPO) algorithm. The total time consumption in training process is less than forty minutes, achieving remarkable speed improvements compared to existing methods. This framework holds the potential to expedite the development of AUV control algorithms and enhance the autonomy of AUV in complex environments.

The Autonomous Underwater Vehicle (AUV) is capable of autonomously conducting underwater cruising tasks. When combined with docking operations, the AUV can replenish its electric power after long-distance travel, enabling it to achieve long-range autonomous monitoring. This paper proposes a positioning method for the cruising and docking stages of AUVs. Firstly, a vision guidance algorithm based on monocular vision and threshold segmentation is studied to address the issue of regional noise that commonly occurs during underwater docking. A solution for regional noise based on threshold segmentation and proportional circle selection is proposed. Secondly, in order to enhance the positioning accuracy during the cruising stage, a fusion positioning algorithm based on particle filtering is presented, incorporating the Doppler Velocity Log (DVL) and GPS carried by the AUV. In simulation, this algorithm improves positioning accuracy by over 56.0% compared to using individual sensors alone. Finally, experiments for cruising and docking were conducted in Qingjiang, Hubei, China. The effectiveness of both methods is demonstrated, with successful docking achieved in four out of five attempts.

Reinforcement Learning (RL) is a promising solution, allowing Unmanned Underwater Vehicles (UUVs) to learn optimal behaviors through trial and error. However, existing simulators lack efficient integration with RL methods, limiting training scalability and performance. This paper introduces MarineGym, a novel simulation framework designed to enhance RL training efficiency for UUVs by utilizing GPU acceleration. MarineGym offers a 10,000-fold performance improvement over real-time simulation on a single GPU, enabling rapid training of RL algorithms across multiple underwater tasks. Key features include realistic dynamic modeling of UUVs, parallel environment execution, and compatibility with popular RL frameworks like PyTorch and TorchRL. The framework is validated through four distinct tasks: station-keeping, circle tracking, helical tracking, and lemniscate tracking. This framework sets the stage for advancing RL in underwater robotics and facilitating efficient training in complex, dynamic environments.

PurposeDocking technology plays a crucial role in enabling long-duration operations of autonomous underwater vehicles (AUVs). Visual positioning solutions alone are susceptible to abnormal drift values due to the challenging underwater optical imaging environment. When an AUV approaches the docking station, the absolute positioning method fails if the AUV captures an insufficient number of tracers. This study aims to to provide a more stable absolute position visual positioning method for underwater terminal visual docking.Design/methodology/approachThis paper presents a six-degree-of-freedom positioning method for AUV terminal visual docking, which uses lights and triangle codes. The authors use an extended Kalman filter to fuse the visual calculation results with inertial measurement unit data. Moreover, this paper proposes a triangle code recognition and positioning algorithm.FindingsThe authors conducted a simulation experiment to compare the underwater positioning performance of triangle codes, AprilTag and Aruco. The results demonstrate that the implemented triangular code reduces the running time by over 70% compared to the other two codes, and also exhibits a longer recognition distance in turbid environments. Subsequent experiments were carried out in Qingjiang Lake, Hubei Province, China, which further confirmed the effectiveness of the proposed positioning algorithm.Originality/valueThis fusion approach effectively mitigates abnormal drift errors stemming from visual positioning and cumulative errors resulting from inertial navigation. The authors also propose a triangle code recognition and positioning algorithm as a supplementary approach to overcome the limitations of tracer light positioning beacons.

With the growing demand for ocean exploration, research on docking technology is gaining popularity, resulting in the development of various types of docking systems. In contrast to traditional cable-based systems, this paper introduces a self-powered docking system designed for AUV recovery and recharging. To minimize energy loss in the self-powered system, an analysis of eddy current loss produced by the coils in the seawater is conducted, and a calculation method for determining the equivalent eddy current impedance is proposed. This allows for the transformation of energy loss analysis from the magnetic field to circuit analysis. Then the efficiency of the inductively coupled power transmission system in seawater is analyzed, and a frequency optimization method is proposed to minimize dissipated energy during the AUV battery charging process. Finally, the effectiveness of the frequency optimization method is validated through laboratory experiments. Moreover, the results of the sea trial confirm the feasibility of the optimal wireless charging system operating in a seawater environment.

Multi-AUV collaboration is significant in marine resource exploration, submarine scientific observation and underwater rescue. However, limited communication bandwidth and acoustic range challenge trajectory estimation of A UVs. In this paper, we design a distributed multi-AUV system and implement an optimizer based on the Alternating Direction Method of Multipliers (ADMM). In our system, each AUV is required to meet and exchange information with others locally. This optimizer treats measurements between AUVs as constraints and transforms the global optimization problem into multiple sub-problems, which allows each AUV to optimize its trajectory independently. Compared with a representative algorithm on the same dataset, this approach effectively reduces the optimization time and improves the convergence efficiency while ensuring convergence accuracy.

Spiral coaxial coils are widely used in wireless charging systems for autonomous underwater vehicles (AUVs). However, these coils can generate axial electromagnetic interference that may adversely affect the electronic components contained within the AUV. In order to overcome this issue, this paper introduced a pair of radially coupled coils which implement distributed ferrite cores. The mathematical model of the curly coils was derived, and its geometry parameters were optimized through the use of genetic algorithms. ANSYS Maxwell was used to analyze and optimize the layout of the ferrite cores. A prototype of the AUV wireless charging system was presented, demonstrating a maximum efficiency of 94% at 2.2 kW in salt water. The rotation adaptivity of the system was also tested, revealing stable output performance within the possible roll-angle variations of the AUV.

Autonomous underwater vehicles (AUVs) play an increasingly important role in underwater missions. However, traditional visual localization methods struggle to meet the accuracy requirements for AUV terminal docking in strong-illumination or turbid water conditions. This work proposes a universal navigation system using electromagnetic (EM) and light beacons for nontorpedo AUVs to be recovered using a planar docking station. The major challenge in designing this system is to deal with the incomplete sensing of the light and EM beacons. To solve this problem, a two-step fusion method of EM localization information is given first. The kinematic model of the AUV and visual information are also utilized to improve the robustness of the navigation system. Based on the localization result, a comprehensive docking strategy is proposed to address the time-varying sensing conditions. Localization and docking experiments were conducted in a pool. The test results show that the maximum EM localization error is less than 20 cm in both static and dynamic tests. The AUV completed all six docking attempts.