水下无人系统学报

2025, 33(2): 193-193.

[Abstract](16) [FullText HTML] (15)

Abstract:

A Review of Research on Path Planning of Unmanned Surface Vessel Swarm: Deep Reinforcement Learning

HOU Yuli, WANG Ning, QIU Chidong, WENG Yongpeng

2025, 33(2): 194-203. doi: 10.11993/j.issn.2096-3920.2025-0034

[Abstract](255) [FullText HTML] (70) [PDF 941KB](89)

Abstract:
An unmanned surface vessel(USV) swarm has shown significant advantages in complex marine missions, but its path planning faces high-dimensional, dynamic, and multi-constraint challenges. Traditional path planning algorithms are difficult to meet increasingly complex needs due to weak coordination mechanisms and insufficient adaptability, while the development of deep reinforcement learning(DRL) technology provides a new research direction for the path planning of USV swarms. This paper systematically reviewed the technical framework and typical algorithms for collaborative path planning of USV swarms based on DRL. Firstly, the technical evolution context and multi-dimensional constraints of path planning of USV swarms were sorted out, and the applicable scenarios and limitations of centralized and distributed decision frameworks were analyzed. Secondly, the principle, application scenarios, and improvement directions of various typical DRL algorithms were discussed, and their advantages and disadvantages were analyzed. Finally, the main challenges and development directions in this field were summarized. This paper aims to provide a reference for the research on DRL-based collaborative path planning of USV swarms.

Underwater Object Detection Method with Enhanced Wavelet Transform Features

WEI Nan, YANG Wankou, ZHOU Weijie, JIANG Longyu

2025, 33(2): 204-211. doi: 10.11993/j.issn.2096-3920.2025-0003

[Abstract](139) [FullText HTML] (58) [PDF 816KB](45)

Abstract:
The complex and unique underwater environment results in low-quality underwater images, characterized by low contrast, blurriness, and underwater degradation, which significantly affects the capabilities of underwater object detection. To address this issue, this paper proposed an underwater object detection method with enhanced wavelet transform features. The paper introduced discrete wavelet transform(DWT) to decompose the multi-level features extracted by the deep learning framework into high- and low-frequency components. These frequency domain feature components were then interactively enhanced using a frequency domain interaction module based on the attention mechanism designed in this work, optimizing the ability of feature expression. The enhanced features were subsequently fed into the object detection network to improve the object detection performance. Experimental results demonstrate that the proposed underwater object detection method outperforms conventional object detection methods, significantly improving the ability to detect objects in underwater environments.

Underwater Visual Object Tracking Method Based on Scene Perception

HU Qianwei, WANG Daiwei, LI Renjie, YU Xiaofan, KANG Bin, SU Ruoyu

2025, 33(2): 212-219, 290. doi: 10.11993/j.issn.2096-3920.2025-0007

[Abstract](83) [FullText HTML] (42) [PDF 3264KB](31)

Abstract:
Underwater visual object tracking is a core technology for scene understanding in autonomous undersea vehicle(AUV) systems. However, challenges such as uneven illumination, background interference, and target appearance variation in complex underwater environments severely affect the accuracy and stability of traditional visual tracking methods. Existing approaches primarily rely on the appearance modeling of the target, making them unreliable in complex environments, particularly when similar targets are present, leading to misidentification and tracking drift. This paper proposed an underwater single-object tracking method based on scene perception that utilized a regional segmentation-based graph convolution module to extract all target regions in the scene. By leveraging a graph convolutional network, the proposed method captured long-range dependencies between the target region and surrounding key regions, significantly enhancing the discrimination capability against similar targets. Additionally, a dual-view graph contrastive learning strategy was introduced, which enabled unsupervised online updates for the graph convolution module by generating randomly perturbed target feature views, ensuring strong adaptability and stability of the model in complex environments. Experiments show that the proposed method is significantly better than the classical method in terms of tracking accuracy and robustness, especially in scenes with large lighting changes, complex backgrounds, and strong interference of similar targets, and the success rate and accuracy are significantly improved. These results indicate that the proposed method effectively addresses target drift challenges in underwater object tracking caused by illumination variations and background interference, maintaining stable tracking even in the presence of similar targets, thus providing an efficient and reliable tracking solution for underwater unmanned systems.

Intelligent Trust Evaluation Method for Underwater Sensor Networks Based on Fuzzy Clustering and Dynamic Weight Allocation

WANG Zhaohui, HAN Guangjie, DU Jiaxin, LIN Chuan, WANG Lei

2025, 33(2): 220-228. doi: 10.11993/j.issn.2096-3920.2024-0176

[Abstract](100) [FullText HTML] (32) [PDF 903KB](28)

Abstract:
Underwater sensor networks (USN) play a crucial role in marine environmental monitoring and other fields while facing significant security challenges. Trust models can effectively defend against insider attacks and ensure network reliability. However, most existing trust evaluation methods rely on the linear weighting of trust features and the decision-making method based on threshold comparison. In dynamic underwater environments, factors such as water flow and temperature are constantly changing in time and space, leading to differences in the variations of node trust features and overall fluctuations of trust values. This makes it challenging to effectively determine the optimal weights and the reasonable threshold, thereby affecting the accuracy of evaluation and the reliability of decision-making. To solve this issue, this paper proposed an intelligent trust evaluation method based on fuzzy clustering and dynamic weight allocation. First, a hierarchical dynamic topology model of the USN was developed to enhance universality. On this basis, communication, energy, and data features were comprehensively calculated to fully reflect node states. Then, the unsupervised machine learning algorithm, namely fuzzy C-means clustering, was employed to enable adaptive node trust decision-making. Meanwhile, a subjective and objective combination strategy was adopted to dynamically allocate weights to features according to network and environmental conditions. Consequently, the intelligent evaluation of node trust was achieved. Simulation results demonstrate that the proposed method can effectively evaluate the trust of nodes in underwater environments, improve the reliability of trust decision-making, and enhance the security of the network.

Real-Time Transformer Detection of Underwater Objects Based on Lightweight Gated Convolutional Network

LI Yuhui, CUI Huixia, LI Yaomin, JIA Senping

2025, 33(2): 229-237. doi: 10.11993/j.issn.2096-3920.2024-0182

[Abstract](92) [FullText HTML] (60) [PDF 2859KB](30)

Abstract:
To address the challenges in underwater object detection algorithms, including difficult image feature processing, redundant model architectures, and excessive parameter numbers, this paper proposed a real-time Transformer detection method for underwater objects based on a lightweight gated convolutional network. This method first constructed a convolutional gated linear unit based on the gating mechanism to dynamically modulate feature transmission. Furthermore, on this basis, a gated channel interaction module was proposed to fully decouple the token mixer from the channel mixer. Additionally, for the token mixer, the structural reparameterization technique was introduced to significantly reduce the computational cost of the model during inference. The hybrid encoder conducted the intra-scale information exchange and multi-scale feature fusion of the three features extracted by the gated backbone network, thus realizing the high fusion of shallow high-frequency information and deep semantic spatial information. The proposed model carried out a large number of experiments on different modal datasets. The results show that the model’s mAP@0.5 reaches 0.849, the overall number of parameters is 23.3×10⁶, and the FPS detection frame rate is 136.8. While maintaining excellent detection accuracy, this model achieves a smaller number of model parameters and higher detection speed, with better overall performance than other models. The results reveal that compared to a series of excellent object detection models, the proposed model features sound detection performance and efficient real-time detection.

A Sonar Image Target Detection Method with Low False Alarm Rate Based on Self-Trained YOLO11 Model

HAN Jingqi, NAN Mingxing, ZHANG Peng, CHEN Jiajie, HU Zhengliang

2025, 33(2): 238-248. doi: 10.11993/j.issn.2096-3920.2024-0165

[Abstract](108) [FullText HTML] (68) [PDF 3342KB](28)

Abstract:
Autonomous detection of sonar image targets is a key technology for unmanned undersea systems, but it faces the challenge of high false alarm rates in practical applications, which limits the quality and efficiency of mission execution by unmanned underwater systems. In this paper, an underwater target detection method based on the YOLO11 model was designed, and a false alarm rate detection method by self-training a deep learning detector on sonar images was proposed to reduce the false alarm rate. This method automatically generated proxy classification tasks based on the sonar image target detection dataset and improved the deep learning detector’s learning of target and background features through pre-training, enhancing the detector’s ability to distinguish between targets and backgrounds and thereby reducing the false alarm rate. Experimental results demonstrate that when the detector’s confidence is set to the value corresponding to the maximum F₁-score, the YOLO11 detector trained using the proposed method can reduce the false alarm rate by 11.60% compared to traditional transfer learning methods while achieving a higher recall rate. This method improves the generalization of the deep learning detector without using external datasets, providing an efficient self-training approach for underwater target detection scenarios with small sample sizes.

Underwater Visual Multi-Target Tracking Algorithm Integrating Re-parameterization and Attention Mechanism

LI Junyi, HE Mingle, LIU Chang, XU Yong

2025, 33(2): 249-260. doi: 10.11993/j.issn.2096-3920.2025-0012

[Abstract](94) [FullText HTML] (67) [PDF 1807KB](21)

Abstract:
The complex underwater environment can severely impact the stability of imaging devices and the quality of captured images, posing significant challenges for visual multi-target tracking in underwater unmanned autonomous systems. To address the difficulties arising from underwater camera jitter and image degradation, this paper proposed an underwater visual multi-target tracking algorithm that integrated re-parameterization and attention mechanisms, specifically tailored for underwater unmanned autonomous systems. First, to tackle the diversity of underwater targets and image degradation, an improved YOLOv8 algorithm based on re-parameterization and attention mechanism(RA-YOLOv8) was proposed. This algorithm effectively enhanced the network’s multi-scale feature extraction capability and improved the detection accuracy of the model by integrating a structurally re-parameterized multi-scale feature extraction convolutional structure(DBB-RFAConv) and the AMSCE-attention mechanism. Then, to address the challenges of real-time target tracking caused by underwater camera jitter, an Inner-PIoUv2-enhanced ByteTrack algorithm(IP2-ByteTrack) was proposed. Inner-PIoUv2 was used as the similarity measure in the matching process of the tracking algorithm, which enhanced the model’s performance in underwater detection and tracking tasks, improving the accuracy of tracking trajectory matching. Finally, based on the RA-YOLOv8 and IP2-ByteTrack algorithms, an underwater visual multi-target tracking algorithm that integrated re-parameterization and attention mechanisms for underwater autonomous systems was proposed. Experimental results show that the proposed algorithm exhibits excellent performance in complex underwater environments and can effectively address the shortcomings of existing methods in underwater multi-target tracking.

MARL-TS Method for Underwater Acoustic Networks in Time-Varying Channels

GAO Yu, XIAO Qiao, WANG Chaofeng

2025, 33(2): 261-271. doi: 10.11993/j.issn.2096-3920.2025-0015

[Abstract](80) [FullText HTML] (48) [PDF 1083KB](26)

Abstract:
Underwater acoustic communication faces numerous challenges in transmission scheduling and decision-making due to its high propagation delay, time-varying channel characteristics, and limited bandwidth. To enhance communication efficiency in complex underwater acoustic environments, this paper proposed a multi-agent reinforcement learning(MARL)-based cross-layer transmission scheduling(TS) method for underwater acoustic networks, termed MARL-TS. This method addressed the high propagation delay and dynamic channel environments by leveraging transmission node buffer states and channel conditions as the foundation while optimizing transmission efficiency and transmission delay of the communication network. It adaptively performs cross-layer optimization to jointly optimize power allocation and timeslot resource scheduling. To learn the optimal transmission strategy, this paper constructed a learnable policy network and a value network, integrating multi-agent cooperative learning to improve strategy optimization efficiency and adaptive decision-making capabilities. Simulation results demonstrate that compared with existing reinforcement learning-based multiple access control(MAC) protocols, MARL-TS significantly enhances transmission efficiency and reduces transmission delay. Notably, it exhibits superior adaptability and stability in multi-node and high-load scenarios, offering a novel approach for optimizing complex underwater communication systems.

A Secure Communication Method for Unmanned Undersea Systems Oriented to Federated Learning

WU Jiajia, XU Ming

2025, 33(2): 272-279. doi: 10.11993/j.issn.2096-3920.2025-0010

[Abstract](75) [FullText HTML] (48) [PDF 849KB](21)

Abstract:
To address the information leakage issue in unmanned undersea systems, a secure communication method oriented to federated learning was proposed in this paper. By considering the complexities and bandwidth limitations of acoustic channels, an unbiased gradient compression method using Kashin compression was introduced. This method reduced the dimensionality of transmission gradients through orthogonal projection and quantization, thereby decreasing communication costs while preserving information integrity. To mitigate information leakage, a privacy-preserving method utilizing a feedback channel was designed, and the normalized acoustic channel transmission matrix and random sequences were employed to generate secret keys, so as to ensure that eavesdroppers cannot decrypt the model parameters. By adopting multi-objective optimization techniques, the Pareto optimal solution was found to strike a balance between model accuracy and secure throughput. Simulation results show that compared with the existing methods, the proposed method in this paper can effectively improve training accuracy and secure throughput while maintaining low latency.

Design of Intelligent Interpretation Network for Underwater Acoustic Communication Receiver with Multiple Signal Modulations

LIU Lanjun, CHENG Zining, CHEN Jialin, LI Ming, LIU Honghao

2025, 33(2): 280-290. doi: 10.11993/j.issn.2096-3920.2024-0174

[Abstract](83) [FullText HTML] (19) [PDF 2407KB](15)

Abstract:
An intelligent interpretation network design scheme for underwater acoustic communication receivers with multiple signal modulations was proposed to meet the requirements of channel adaptive underwater acoustic high-quality communication in complex application scenarios. It supported four signal modulation methods including orthogonal frequency division multiplexing(OFDM), single carrier modulation(SCM), multi carrier frequency domain spread spectrum(MC-FDSS), and single carrier with time domain spread spectrum(SC-TDSS). Intelligent interpretation modules based on fully-connected deep neural network(FC-DNN) and long short-term memory(LSTM) networks were used to replace traditional channel estimation and channel equalization modules. A deep learning network structure that facilitated parallel expansion was designed for non-spread spectrum and spread spectrum signal modulation methods. Network training and testing were conducted based on five typical time-varying channel models. The test results show that the two designed intelligent interpretation networks have significantly improved system performance compared to traditional least squares(LS) estimation + zero-forcing(ZF) equalization and LS estimation + minimum mean squared error(MMSE) channel estimation equalization methods. Under a signal-to-noise ratio of 5 dB, the system error rates of OFDM and SCM non-spread spectrum signal modulation methods are reduced by about 10 times and 100 times, respectively. Under a signal-to-noise ratio of −5 dB, the system error rates of MC-FDSS and SC-TDSS spread spectrum signal modulation methods are reduced by about 100 times and 1 000 times, respectively. The system performance of the two designed intelligent interpretation networks is comparable, and they both have good generalization performance. The computational complexity of the intelligent interpretation network based on FC-DNN is relatively low.

Prediction of SNR Based on SVR and Adaptive Transmission Power Method for Underwater Acoustic Communication

ZHENG Jixing, YUAN Yufan, ZHUO Xiaoxiao, LU Xuesong, QU Fengzhong, WEI Yan

2025, 33(2): 291-298. doi: 10.11993/j.issn.2096-3920.2024-0180

[Abstract](64) [FullText HTML] (24) [PDF 873KB](12)

Abstract:
Marine environmental noise is influenced by many factors such as ocean waves, wind, rain, marine organisms, ships, and industrial activities. Its power is highly random. However, the continuous effect of factors such as sea surface temperature and tidal height can also make the power have certain periodic characteristics. Underwater environmental noise can directly affect the communication packet error rate during underwater acoustic communication. Although increasing the transmission power can raise the received signal-to-noise ratios and decrease the packet error rate, it also enhances the average energy consumption of communication. Therefore, in order to reduce the packet error rate and average energy consumption of underwater acoustic communication, this paper analyzed and predicted the signal-to-noise ratios time series based on the support vector regression(SVR) algorithm and proposed an adaptive transmission power method for underwater acoustic communication based on signal-to-noise ratio prediction. The simulation results show that compared with the exponential smoothing and autoregressive integrated moving average model(ARIMA) methods, the SVR algorithm based on the linear kernel function has the best performance in predicting signal-to-noise ratios and the smallest prediction error on test data. Under different modulation methods, the proposed adaptive transmission power method for underwater acoustic communication can improve the success rate of data packet transmission while reducing energy consumption per kilobyte.

Underwater Acoustic Rapidly Time-Varying Channel Equalization Technique Integrating Deep Learning and Domain Knowledge

DENG Ke, WANG Shuaijun, YU Hua, ZHANG Jian, CHEN Junfan, WU Zhouping

2025, 33(2): 299-306. doi: 10.11993/j.issn.2096-3920.2024-0163

[Abstract](60) [FullText HTML] (27) [PDF 1498KB](14)

Abstract:
Multicarrier communication schemes, such as orthogonal frequency division multiplexing(OFDM), are the mainstream solutions for achieving high spectral efficiency in underwater acoustic transmissions. These schemes effectively address frequency-selective fading caused by multipath acoustic propagation in underwater environments. However, in rapidly time-varying scenarios, inter-carrier interference(ICI) can severely compromise transmission reliability. To mitigate the time-frequency doubly-selective fading in such underwater acoustic rapidly time-varying channels and reduce the bit error rate(BER) of OFDM systems, this paper proposed an underwater acoustic rapidly time-varying channel equalization method that combined deep learning with domain knowledge. Different from regarding the outcomes of traditional channel estimation and equalization detection as preprocessing results or supplementary information sources for deep neural networks(DNNs), this paper employed the structured information from classical frequency-domain equalization models to assist in the training and inference of DNN models, so as to counteract the adverse effects of ICI and adapt to scenarios where there is a mismatch between the actual deployment channel environment and the training channel environment. Numerical simulation and sea trial results confirm that the proposed approach can effectively reduce the BER of receivers and achieve faster model convergence, and the potential to achieve stronger generalization performance under unknown channel conditions.

Trust Model for Underwater Wireless Sensor Networks Based on Variational Autoencoders

ZHANG Jiahao, WEI Sizhou, XIA Na

2025, 33(2): 307-316. doi: 10.11993/j.issn.2096-3920.2024-0181

[Abstract](75) [FullText HTML] (38) [PDF 1612KB](16)

Abstract:
In underwater wireless sensor networks(UWSNs), the complex underwater acoustic communication environment and the limited resources of nodes make malicious node attacks more covert and threatening. Therefore, researching effective malicious node detection methods is crucial for maintaining network stability and data security. This paper proposed a trust model for UWSNs based on variational autoencoders(VAEs), which evaluated node behavior credibility to identify malicious nodes. First, the model aggregated the behavioral feature data from the underwater node transmission process, extracting various indicators such as node location, packet delivery ratio, and delay, thereby forming a trust dataset. The dataset was then encoded and trained, and variational inference was employed to map the data to a latent space and obtain the probability distribution of this space. Finally, based on the probability distribution, the model decoded and reconstructed the data to derive node behavior credibility, thus completing the trust evaluation of nodes. Comparative experimental results show that compared to methods such as the intrusion detection-based trust management system, the proposed model improves trust evaluation accuracy by at least 10.5% and demonstrates significant advantages in operational stability.

Direction of Arrival Estimation Algorithm for Underwater Distributed Sources Based on Deep Neural Network

LIANG Yinian, LI Jie, CHEN Fangjiong, JI Fei, YU Hua

2025, 33(2): 317-324. doi: 10.11993/j.issn.2096-3920.2024-0178

[Abstract](80) [FullText HTML] (32) [PDF 1015KB](17)

Abstract:
In view of the limitation of traditional subspace-based direction of arrival(DOA) estimation algorithms that rely on prior coherence information for localizing distributed sources with varying coherences, a DOA estimation algorithm for underwater distributed sources based on deep neural network(DNN) was proposed in this paper. By leveraging the separability of temporal and angular coherence components in the partially coherent distributed source signal model and simplifying the model by segmented mean normalization, a DNN model was constructed and trained with multiple samples of different coherence coefficients, thus achieving robust DOA estimation for distributed sources with different coherence levels. Simulation results indicate that the proposed method can effectively estimate the distributed source parameters with different coherent coefficients without relying on coherence prior knowledge. The proposed method is compared with four traditional subspace-based methods and one deep convolutional neural network algorithm, and the results show that the root mean square error of the proposed method under the coherently distributed source case is 0.42° lower than that of other methods under different signal-to-noise ratios(SNRs) and snapshots; under the incoherently distributed source case, the RMSE of the proposed method is 0.04° lower than that of other methods with SNR greater than 0 dB and snapshots greater than 600. In the range of full coherence coefficients, the proposed methods all show better estimation performance, which proves their applicability in complex underwater environments.

Multi-Underwater Target Interception Strategy Based on Deep Reinforcement Learning

GAN Wenhao, PENG Yunfei, QIAO Lei

2025, 33(2): 325-332. doi: 10.11993/j.issn.2096-3920.2025-0004

[Abstract](114) [FullText HTML] (24) [PDF 5783KB](23)

Abstract:
In the context of multiple autonomous undersea vehicles(AUVs) executing underwater target interception missions, AUVs are required to make precise decisions based on both enemy and partner information, navigating the dual challenges of competition and cooperation. Most existing research typically focuses on single-target interception in simple environments and lacks a detailed exploration of collaborative mechanisms for multi-target interception mechanisms in complex environments. Therefore, this paper proposed a multi-agent deep reinforcement learning framework for AUVs to learn interception strategies in environments with complex obstacles and time-vary ocean currents, with a focus on cooperation in many-to-many game scenarios. First, a hierarchical maneuvering framework was introduced to improve the decision-making ability of AUVs through a three-layer loop structure. Next, the multi-agent proximal policy optimization algorithm was used to construct a scalable state and action space and design a compound reward function, enhancing interception efficiency and cooperation of AUVs. Finally, a population expansion–curriculum learning approach was incorporated within a centralized training and distributed execution architecture to help AUVs master generalizable cooperation strategies. Training results show rapid convergence and high success rates of the proposed interception strategies. The simulation experiments show that the trained AUVs can use the same set of models in multiple population configurations to effectively intercept multiple intruding targets through cooperation while avoiding obstacles.

Adaptive Optimization Control of Unmanned Surface Vessel with Thruster Fault

GAO Ying, LU Mingchun, ZHANG Rubo, WANG Ning

2025, 33(2): 333-340, 388. doi: 10.11993/j.issn.2096-3920.2025-0013

[Abstract](57) [FullText HTML] (20) [PDF 2506KB](21)

Abstract:
In view of both unknown dynamics and thruster fault, an adaptive model predictive-based fault-tolerant control scheme was proposed for an unmanned surface vessel(USV). Firstly, the dynamics model of the USV with faults was established, and the unknown nonlinear dynamics and external disturbance in the dynamics model were formed into a lumped nonlinear function. The unknown part in the dynamics was approximated by the neural network. In order to achieve high performance and accurate tracking of the desired trajectory, an adaptive autonomous fault-tolerant control strategy was designed by combining model predictive control and backstepping control, with the index function of control input and state error as variables. Then, based on Lyapunov stability theory, it was proven that all signals in the closed-loop system were bounded. The control strategy constructed under this framework could not only compensate for the influence of actuator faults and unknown nonlinear dynamics on the system but also ensure that the tracking error of the system converges to the ideal precision. Simulation results verify the effectiveness and rationality of the proposed method.

Extreme Learning-Based Robust Adaptive Path Tracking Control of Underactuated Unmanned Surface Vessel

HE Xinyu, WANG Ning, WU Haojun

2025, 33(2): 341-349. doi: 10.11993/j.issn.2096-3920.2024-0170

[Abstract](92) [FullText HTML] (29) [PDF 1260KB](15)

Abstract:
An extreme learning-based robust adaptive path tracking control scheme was proposed for underactuated unmanned surface vessels(USVs) with unknown dynamics, parameter uncertainties, and disturbances. Firstly, the surge-guided line-of-sight guidance law was used to guide the surge speed and heading angle at the same time, so as to avoid the singularity of the guidance process and make the USV converge to the desired path quickly. Secondly, the unknown dynamics including system uncertainties and external disturbances were encapsulated into a lumped unknown term, and hidden layer nodes were randomly generated by the single-hidden layer feedforward network(SLFN) of the extreme learning machine to identify the unknown term and avoid relying on USV prior knowledge and dimension explosion problem. Then, by designing an adaptive compensator for the approximation residual, the output weight and the approximation residual of the SLFN were updated online at the same time to form a dual-channel learning mechanism, which can not only enhance the approximation ability but also improve the tracking accuracy. Finally, an adaptive path tracking controller was designed so that the surge velocity and heading angle guidance errors of USVs can gradually converge to a small neighborhood near the origin. Simulation studies verify the effectiveness and superiority of the proposed scheme.

Prediction of Lightweight AIS-Based Ship Trajectories with Spline Interpolation

HU Jiantao, LI Tianjiao, LIU Hui, LI Shuxin, CHENG Xu

2025, 33(2): 350-358. doi: 10.11993/j.issn.2096-3920.2024-0164

[Abstract](187) [FullText HTML] (54) [PDF 1776KB](18)

Abstract:
Automatic identification system(AIS) of ships provides a large amount of real-time ship navigation data, which has become an indispensable key data source in many fields such as maritime traffic management, search and rescue operations, and risk assessment. Among them, ship trajectory prediction has received wide attention. However, realizing accurate long-time trajectory prediction faces two major problems: One is the integrity of AIS data itself, and the other is the efficiency of prediction models. Therefore, how to effectively deal with the missing AIS data and how to construct a lightweight and efficient prediction model have become the key problems to be solved. In this paper, a lightweight AIS-based ship trajectory prediction method with spline interpolation was proposed. Spline interpolation was used to fill in the missing AIS data, and a lightweight linear layer structure was introduced to reduce the complexity of the deep learning model. The experimental results show that the method can effectively interpolate the missing AIS data, significantly reduce the number of parameters and computation of the deep learning model, and then improve the prediction accuracy of ship trajectories.

Detection of Shipborne Charging Components and Docking of Shore-Based Manipulator Based on MobileNetV2

XIA Tenghui, WANG Yueying, WU Nailong, LIU Futeng

2025, 33(2): 359-366. doi: 10.11993/j.issn.2096-3920.2024-0167

[Abstract](87) [FullText HTML] (38) [PDF 1972KB](13)

Abstract:
To realize the autonomous charging of unmanned surface vessels(USVs), a method based on MobileNetV2 was proposed for the detection of shipborne charging components and the docking of shore-based manipulators. Firstly, binocular camera D435i was employed to collect RGB and depth maps as inputs and MobileNetV2-based detection network was utilized to estimate the position and attitude of charging components. Then, the position and attitude of the charging components in the coordinate system of the manipulator base were calculated by coordinate transformation. Subsequently, the charging plug at the end of the manipulator was driven close to the charging component to achieve a preliminary docking. Finally, the docking strategy was adopted to search for the internal holes for final docking. In this study, an experimental platform for docking charging components was built in a real environment to verify the effectiveness of the proposed method. The charging components of USVs could be accurately identified by this method. Besides, the manipulator completed the docking of the charging plug and the charging components under the control of the proportional-differential torque control strategy based on gravity compensation, providing a new idea for the autonomous charging of USVs.

Cooperative Control and Intelligent Optimization for Air-Sea Heterogeneous Unmanned Systems

KE Can, CHEN Huifang, XIE Lei

2025, 33(2): 367-379. doi: 10.11993/j.issn.2096-3920.2024-0173

[Abstract](87) [FullText HTML] (32) [PDF 2758KB](28)

Abstract:
In order to cope with increasingly complex ocean missions, an air-sea heterogeneous unmanned system composed of unmanned aerial vehicles(UAVs), unmanned surface vessels(USVs), and unmanned undersea vehicles(UUVs) was constructed to study the cooperative control problem. For the information exchange problem of a heterogeneous unmanned system, each domain consisted of a leader and multiple followers, where cross-domain communication was required between the leaders of each domain. Meanwhile, for the trajectory planning issue of the leader in each domain, a cooperative trajectory planning algorithm based on the artificial potential field method was proposed for the leader in each domain to reach the target location while avoiding obstacles. For the limited communication resources, an impulsive hierarchical formation control protocol with intermittent communication was designed for followers in each domain, which reduces communication overhead while achieving formation control under obstacle avoidance. Besides, for the multi-objective optimization problem of convergence time and communication energy consumption in cooperative control algorithm, an improved multi-objective quantum-behavior particle swarm optimization algorithm was proposed by designing contraction-expansion coefficient and dynamic dense distance strategy, which was used to intelligently select the impulsive interval for each domain, achieving a good compromise between the convergence time and communication energy consumption of cooperative control algorithm. Simulation results demonstrate that the air-sea heterogeneous unmanned system can achieve formation control while avoiding obstacles and reducing communication overhead, and the proposed algorithm has better convergence and global search ability than the traditional multi-objective quantum-behavior particle swarm optimization algorithm.

Unmanned Surface Vessel Cluster Path Planning Based on Deep Reinforcement Learning

YU Manjiang, HE Jiawei, XING Bowen

2025, 33(2): 380-388. doi: 10.11993/j.issn.2096-3920.2024-0179

[Abstract](227) [FullText HTML] (71) [PDF 2887KB](32)

Abstract:
With the wide application of unmanned surface vessels(USVs) in the field of maritime search, the traditional path planning algorithms fail to meet the complex rescue scenarios, which can lead to local optimum, low task completion rate, and slow convergence speed. For this reason, a path planning method for USV cluster cooperative search and rescue was proposed. Firstly, a long and short-term memory module was introduced based on the multi-agent deep deterministic policy gradient algorithm to enhance the ability of the USVs to utilize the temporal information in path planning; secondly, a multi-level representational experience pool was designed to improve the training efficiency and data utilization and reduce the interference between different experiences; finally, stochastic network distillation was used as a curiosity mechanism to provide intrinsic rewards for the USVs to explore new regions and solve the convergence due to the sparse rewards. The simulation experiment results show that the improved algorithm improves the convergence speed by about 38.46% compared with the original algorithm, and the path length has been shortened by 27.02%. In addition, the obstacle avoidance ability has been significantly improved.

Current Issue
2025, Volume 33, Issue 2

Journal News> >

Download> >

Current Issue 2025, Volume 33, Issue 2

Journal News> >

Download> >

Current Issue
2025, Volume 33, Issue 2