A method, system, device and medium for clustering in underwater acoustic sensor networks

Abstract

This application discloses a method, system, device, and medium for clustering underwater acoustic sensor networks, relating to the field of underwater acoustic communication technology. The method includes: acquiring the current network state characteristics of each node in the target underwater acoustic sensor network; inputting the current network state characteristics into an underwater acoustic sensor network clustering model to obtain the current optimal clustering scheme; based on the current optimal clustering scheme, electing cluster head nodes through a dynamic role rotation mechanism to generate an adaptive network topology; evaluating the performance of the adaptive network topology to obtain a set of performance indicators; if the performance indicators do not meet the performance threshold, or if the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence level threshold, then initiating a complete optimization learning process. This application effectively solves the technical problem that existing clustering methods struggle to online and adaptively balance multi-objective performance and continuously self-optimize to adapt to dynamic underwater environments.

Description

Technical Field

[0001] This application relates to the field of underwater acoustic communication technology, and in particular to a method, system, device and medium for clustering underwater acoustic sensor networks. Background Technology

[0002] Underwater acoustic sensor networks are critical information infrastructure supporting applications such as marine environmental monitoring, resource exploration, and underwater security. The rationality of their network topology directly determines the overall performance, energy efficiency, and lifespan of the system. To manage large-scale networks, clustering has become an effective topology control method. Its core lies in electing cluster head nodes to aggregate and forward data, thereby optimizing network energy consumption and communication quality.

[0003] However, traditional underwater acoustic network clustering methods face severe challenges in complex and dynamic underwater environments. First, most traditional protocols select cluster heads based on probability or fixed rotation periods, failing to fully consider the heterogeneity of node energy distribution and their three-dimensional spatial relationships. This easily leads to uneven energy consumption, creating energy gaps and shortening the overall network lifetime. Second, nodes in underwater environments are easily affected by ocean currents and other factors, causing significant time-varying network topology. Static or semi-static clustering structures struggle to respond quickly to such dynamic changes, resulting in frequent communication link interruptions and coverage blind spots. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, device and medium for underwater acoustic sensor network clustering, so as to solve the technical problem that existing clustering methods are difficult to balance multi-objective performance online and adaptively and continuously self-optimize to adapt to dynamic underwater environments.

[0005] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for clustering underwater acoustic sensor networks. The method includes: acquiring the current network state characteristics of each node in a target underwater acoustic sensor network, wherein the current network state characteristics include at least energy distribution characteristics, spatial distribution characteristics, and communication link characteristics; inputting the current network state characteristics into an underwater acoustic sensor network clustering model to obtain a current optimal clustering scheme; wherein the underwater acoustic sensor network clustering model is trained based on a historical optimization database; based on the current optimal clustering scheme, electing cluster head nodes through a dynamic role rotation mechanism to generate an adaptive network topology; performing performance evaluation on the adaptive network topology to obtain a set of performance indicators; if the performance indicators meet a performance threshold, confirming the effectiveness of the adaptive network topology, and setting the current network state... The current network state characteristics, the current optimal clustering scheme, and the performance indicators are stored as new sample data in the historical optimization database. If the performance indicators do not meet the performance threshold, or if the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence level threshold, a complete optimization learning process is initiated. The complete optimization learning process includes: determining a new optimal clustering scheme based on the current network state characteristics using a multi-objective optimization algorithm; generating a new adaptive network topology based on the new optimal clustering scheme and performing performance evaluation to obtain corresponding new performance indicators; storing the verified new optimal clustering scheme, the current network state characteristics, and the new performance indicators as new sample data in the historical optimization database, and updating the underwater acoustic sensor network clustering model.

[0006] Optionally, obtaining the current network state characteristics of each node in the target underwater acoustic sensor network includes: The remaining energy and energy consumption rate of the data collection nodes are used to calculate the energy distribution unevenness and the overall network energy level based on the remaining energy and energy consumption rate, thus constituting the energy distribution characteristics. The three-dimensional Euclidean distance and depth gradient between nodes are collected multiple times, and the node density difference, spatial dispersion and average communication distance are calculated based on the three-dimensional Euclidean distance and depth gradient collected in a single collection to form the spatial distribution characteristics; the topology change rate is calculated based on the changes of the three-dimensional Euclidean distance and depth gradient collected multiple times over time. The communication link features are defined by collecting at least one of the following: delivery success rate, received signal-to-noise ratio, path loss, and single-hop delay. The energy distribution non-uniformity and the spatial dispersion are used to evaluate the topological complexity of the target underwater acoustic sensor network.

[0007] Optionally, the underwater acoustic sensor network clustering model is a deep neural network model; The historical optimization database stores network state characteristics, corresponding clustering schemes, and performance metrics under different network scenarios.

[0008] Optionally, determining the new optimal clustering scheme through a multi-objective optimization algorithm specifically includes: Based on the current network state characteristics, a strategy direction is initially selected from multiple candidate clustering strategies according to a preset rule set; the multiple candidate clustering strategies include load balancing priority strategy, latency optimization priority strategy, energy consumption balancing priority strategy, spatial clustering-based clustering strategy, and hybrid optimization clustering strategy; The non-dominated sorting genetic algorithm NSGA-II is used to optimize the total network energy consumption, maximum communication latency and load balance. The algorithm optimizes and selects candidate clustering schemes based on the initial selection strategy to determine the new optimal clustering scheme. The step of initially selecting a strategy direction from multiple candidate clustering strategies based on the current network state characteristics and a preset rule set includes: If the energy distribution unevenness is higher than the first threshold and the node density difference is higher than the second threshold, then the load balancing priority strategy is initially selected. If the spatial dispersion is higher than the third threshold and the communication distance is higher than the fourth threshold, then the latency optimization priority strategy is initially selected. If the overall energy level of the network is lower than the fifth threshold, the energy consumption balancing priority strategy is initially selected. If the spatial clustering algorithm determines that the node distribution presents a clustering pattern, then the aforementioned clustering strategy based on spatial clustering is initially selected. If the network topology change rate is higher than the sixth threshold, then the hybrid optimization clustering strategy is initially selected.

[0009] Optionally, the dynamic role rotation mechanism includes: Based on the node's remaining energy, location centrality, and historical role frequency, a cluster head election weight is calculated using a multi-criteria evaluation function, and an optimal cluster head set is generated according to the weight. Specifically, when the network topology change rate is detected to exceed a set threshold, or when the remaining energy of a node is lower than the dynamic energy threshold, cluster head reselection or cluster reconstruction is triggered to regenerate the optimal cluster head set.

[0010] Optionally, the performance evaluation metrics include at least network coverage, inter-cluster energy consumption balance, packet delivery success rate, and estimated network lifetime. The performance thresholds are design values ​​set for each indicator.

[0011] Optionally, updating the underwater acoustic sensor network cluster model means: using the sample data accumulated in the historical optimization database to periodically or trigger retrain the underwater acoustic sensor network cluster model in order to optimize the model parameters of the underwater acoustic sensor network cluster model.

[0012] In a second aspect, this application provides an underwater acoustic sensor network clustering system, implemented using the underwater acoustic sensor network clustering method described in any one of the first aspects above, the underwater acoustic sensor network clustering system comprising: The feature acquisition module is used to acquire the current network state features of each node in the target underwater acoustic sensor network. The current network state features include at least energy distribution features, spatial distribution features, and communication link features. The scheme generation module is used to input the current network state features into the underwater acoustic sensor network clustering model to obtain the current optimal clustering scheme; wherein, the underwater acoustic sensor network clustering model is obtained by training based on a historical optimization database; The role switching module is used to elect cluster head nodes and generate an adaptive network topology based on the current optimal clustering scheme through a dynamic role rotation mechanism. The performance evaluation module is used to evaluate the performance of the adaptive network topology and obtain performance indicators. If the performance indicators meet the performance threshold, the adaptive network topology is confirmed to be effective, and the current network state characteristics, the current optimal clustering scheme, and the performance indicators are stored as new sample data in the historical optimization database. If the performance indicators do not meet the performance threshold, or the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence threshold, the complete optimization learning process is initiated. The optimization learning module is used to determine a new optimal clustering scheme based on the current network state characteristics using a multi-objective optimization algorithm; generate a new adaptive network topology based on the new optimal clustering scheme and perform performance evaluation to obtain the corresponding new performance indicators; store the verified new optimal clustering scheme, the current network state characteristics, and the new performance indicators as new sample data in the historical optimization database, and update the underwater acoustic sensor network clustering model.

[0013] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the underwater acoustic sensor network clustering method described in any one of the above.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the underwater acoustic sensor network clustering method described above.

[0015] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, system, device, and medium for clustering underwater acoustic sensor networks. By inputting real-time acquired network state features into a pre-trained underwater acoustic sensor network clustering model, it can directly and quickly output the optimal clustering scheme adapted to the current state. This changes the traditional method's reliance on fixed rules or complex offline computation, significantly improving the real-time performance and environmental adaptability of decision-making. Secondly, based on the current optimal clustering scheme, a dynamic role rotation mechanism is used to elect cluster head nodes, generating an adaptive network topology. For example, the dynamic role rotation mechanism can be simply based on a time period; every fixed time interval, all nodes re-elect cluster head nodes, with the election criteria being random selection or based solely on remaining energy. In this way, the nodes in the network are organized into several clusters, each with a cluster head node responsible for data aggregation and forwarding, forming a topology that can respond to environmental changes. Finally, the adaptive network topology is evaluated to obtain a set of performance metrics. For example, the performance evaluation can focus on only a few easily measurable metrics, such as network connectivity (whether there are isolated nodes) and average energy consumption. These metrics are obtained through simulation or actual deployment testing to quantify the effectiveness of the current clustering scheme. If the performance metrics meet the performance threshold, the adaptive network topology is confirmed to be effective, and the current network state characteristics, the current optimal clustering scheme, and the performance metrics are stored as new sample data in the historical optimization database. If the performance metrics do not meet the performance threshold, or the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence level threshold, a complete optimization learning process is initiated. This complete optimization learning process includes: determining a new optimal clustering scheme based on the current network state characteristics using a multi-objective optimization algorithm; generating a new adaptive network topology based on the new optimal clustering scheme and evaluating its performance to obtain the corresponding new performance metrics; storing the validated new optimal clustering scheme, the current network state characteristics, and the new performance metrics as new sample data in the historical optimization database, and updating the underwater acoustic sensor network clustering model. This application, by acquiring and analyzing the real-time state characteristics of the underwater acoustic sensor network and combining them with a clustering model trained based on historical data, can intelligently generate and adjust the network topology. When the performance of the existing scheme is insufficient or the model confidence level is low, a complete optimization learning process can be initiated to actively explore and generate better clustering schemes and feed them back to the model for continuous improvement. Therefore, this application effectively solves the problems of uneven node energy consumption, shortened network lifetime, and communication interruption and coverage blind spots caused by topology time-varying in underwater environments. It also solves the technical problem that existing clustering methods are unable to online and adaptively balance multi-objective performance and continuously self-optimize to adapt to dynamic underwater environments. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a method for clustering underwater acoustic sensor networks according to an embodiment of this application; Figure 2 A flowchart illustrating the complete optimization learning process provided in one embodiment of this application; Figure 3 A schematic diagram of the functional modules of an underwater acoustic sensor network clustering system provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] Traditional underwater acoustic sensor network clustering methods fail to adequately consider the heterogeneity of node energy distribution and three-dimensional spatial relationships in complex and dynamic underwater environments. This easily leads to uneven energy consumption, energy voids, and a shortened overall network lifetime. Furthermore, static or semi-static clustering structures struggle to respond quickly to node movement and time-varying network topology in underwater environments, resulting in frequent communication link interruptions and coverage blind spots. To address these technical problems, this application proposes an underwater acoustic sensor network clustering method. This method utilizes intelligent learning and optimization mechanisms to improve the adaptability, energy efficiency, and overall performance of underwater acoustic sensor networks.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] This application provides a method for clustering underwater acoustic sensor networks. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server.

[0022] Please see Figure 1 In an exemplary embodiment, the underwater acoustic sensor network clustering method includes: S110. Obtain the current network state characteristics of each node in the target underwater acoustic sensor network.

[0023] The current network state characteristics include at least energy distribution characteristics, spatial distribution characteristics, and communication link characteristics. Energy distribution characteristics reflect the balance of energy consumption by nodes in the network and the overall energy level; spatial distribution characteristics describe the positional relationships, density, and movement trends of nodes in three-dimensional space; and communication link characteristics characterize the reliability and efficiency of communication quality between nodes.

[0024] In an exemplary embodiment, obtaining the current network state characteristics of each node in the target underwater acoustic sensor network may include: The remaining energy and energy consumption rate of the collected nodes are used to calculate the energy distribution non-uniformity and the overall network energy level, thus constituting the energy distribution characteristics. The three-dimensional Euclidean distance and depth gradient between nodes are collected multiple times, and the node density difference, spatial dispersion, and average communication distance are calculated based on the three-dimensional Euclidean distance and depth gradient from a single collection, thus constituting the spatial distribution characteristics. The topology change rate is calculated based on the changes in the three-dimensional Euclidean distance and depth gradient over time from multiple collections. At least one of the following communication link characteristics is collected: delivery success rate, received signal-to-noise ratio, path loss, and single-hop delay, thus constituting the communication link characteristics.

[0025] Among them, energy distribution non-uniformity and spatial dispersion are used to evaluate the topological complexity of the target underwater acoustic sensor network.

[0026] Specifically, the remaining energy of a node refers to the battery energy currently available to each node in the underwater acoustic sensor network, typically expressed in joules or as a percentage. The energy consumption rate refers to the rate at which a node consumes energy per unit of time, such as milliwatts per second or joules per second.

[0027] By collecting the remaining energy and energy consumption rate of nodes, the energy distribution unevenness can be calculated. This indicator reflects the balance of energy consumption among nodes in the network. High unevenness may lead to premature failure of some nodes, forming network holes. The overall network energy level represents the total energy reserve of the entire network and is an important basis for measuring the overall network lifetime. These calculation results together constitute the energy distribution characteristics, providing important energy dimension information for subsequent clustering decisions.

[0028] Meanwhile, the three-dimensional Euclidean distance between nodes refers to the straight-line distance between any two nodes in the underwater acoustic sensor network in three-dimensional space, which can be obtained through acoustic ranging, GPS, or inertial navigation systems. The depth gradient refers to the rate of change of a node's position in the vertical direction or the relative depth difference. The dynamic nature of the underwater acoustic environment causes frequent changes in node positions, thus requiring multiple data acquisitions. Based on the three-dimensional Euclidean distance and depth gradient from a single acquisition, node density differences can be calculated, reflecting the density of node distribution in a local area; spatial dispersion measures the degree of dispersion of nodes throughout the monitoring area; and average communication distance is an important parameter for evaluating network connectivity and communication overhead. These indicators collectively constitute the spatial distribution characteristics. Furthermore, by comparing the three-dimensional Euclidean distance and depth gradient acquired multiple times at different time points, the topology change rate can be calculated. The topology change rate quantifies the dynamics of the network topology and is crucial for triggering and adjusting dynamic clustering strategies.

[0029] Furthermore, communication link characteristics are crucial for evaluating the quality and reliability of underwater acoustic communication. Delivery success rate refers to the proportion of data packets successfully transmitted from the source node to the destination node, reflecting the link's reliability. Received signal-to-noise ratio (SNR) is the ratio of received signal power to noise power; a high SNR generally indicates better communication quality. Path loss refers to the energy attenuation of the signal during propagation due to absorption, scattering, and diffusion, directly affecting communication distance and reliability. Single-hop delay refers to the time required for a data packet to travel between two directly communicating nodes, and is an important indicator of network real-time performance. Collecting at least one of these parameters can comprehensively or partially reflect the quality of the current communication link, providing a basis for clustering algorithms to select appropriate cluster heads and members and optimize data routing.

[0030] It is important to note that energy distribution unevenness reflects the degree of energy consumption balance in the network; high unevenness indicates a complex network energy state and potential bottlenecks. Spatial dispersion describes the distribution of nodes in physical space; high dispersion may lead to poor network connectivity or the formation of isolated clusters. Combining these two metrics allows for a comprehensive evaluation of the topological complexity of underwater acoustic sensor networks. Topological complexity is a key factor influencing clustering strategy selection and performance; for example, high complexity may require more frequent reconfiguration or more robust clustering algorithms.

[0031] The embodiments of this application can obtain more comprehensive, accurate, and multi-dimensional current network state characteristics. Specifically, the acquisition of energy distribution characteristics enables the clustering algorithm to fully consider the energy status of nodes, avoiding premature node failure due to energy depletion, thereby extending the overall network lifetime. The collection of spatial distribution characteristics, especially the calculation of topology change rate, enables the system to perceive the dynamics of the underwater acoustic environment and adjust the clustering structure in a timely manner to adapt to node movement and environmental changes, ensuring network connectivity and coverage. The acquisition of communication link characteristics directly reflects the quality and efficiency of data transmission, helping the clustering algorithm to select high-quality communication paths and reduce data packet loss rate and transmission latency. Energy distribution non-uniformity and spatial dispersion, as evaluation indicators of topology complexity, provide the clustering model with a deeper understanding of network structure, enabling it to select more appropriate clustering strategies for network environments of different complexities. These detailed feature information significantly improve the ability of the underwater acoustic sensor network clustering model to understand the actual network situation, thereby generating more adaptive, robust and efficient clustering schemes. This effectively solves the problem of poor clustering results caused by insufficient or inaccurate feature information, and thus optimizes the overall performance and operating efficiency of the network.

[0032] S120. Input the current network state features into the underwater acoustic sensor network clustering model to obtain the current optimal clustering scheme. The underwater acoustic sensor network clustering model is trained based on a historical optimization database.

[0033] In an exemplary embodiment, the underwater acoustic sensor network clustering model is a deep neural network model. The historical optimization database stores network state characteristics, corresponding clustering schemes, and performance metrics for different network scenarios.

[0034] Specifically, in the clustering scenario of underwater acoustic sensor networks, a deep neural network model is introduced to enhance the clustering model's ability to learn and model the complex dynamic characteristics of underwater acoustic sensor networks. Compared to traditional machine learning models, deep neural networks can handle high-dimensional, nonlinear network state feature data and uncover deeper correlations, thereby improving the accuracy and adaptability of the clustering scheme. Specifically, this deep neural network model can employ a multilayer perceptron (MLP) architecture, learning the mapping relationship between network state features and the clustering scheme through multiple fully connected layers; or it can use a graph neural network (GNN) to better capture the topological structure and interactions between nodes in the underwater acoustic sensor network. The number of layers, the number of neurons in each layer, the activation function, the optimizer, the loss function, and other parameters of the model can all be finely tuned according to the actual network data and performance requirements.

[0035] Meanwhile, the historical optimization database stores network state characteristics, corresponding clustering schemes, and performance metrics under different network scenarios. This historical optimization database serves as the training dataset for the deep neural network model, providing the model with rich and diverse learning samples. By storing the "network state characteristics-clustering scheme-performance metrics" triple, the deep neural network model can learn which clustering scheme yields optimal performance under specific network states. This allows the model to learn from historical experience and generalize to new, unencountered network scenarios. The data in this database can originate from simulation environments, operational data from actually deployed underwater acoustic sensor networks, or be generated through expert experience and manual optimization. The database should possess effective data insertion, query, and update mechanisms to support continuous learning and iterative optimization of the model, ensuring that it always contains the latest and validated optimization data.

[0036] By implementing the underwater acoustic sensor network clustering model as a deep neural network model, this application effectively enhances the model's ability to learn and model the complex dynamic characteristics of underwater acoustic sensor networks. Deep neural networks, with their multi-layer nonlinear processing capabilities, can automatically extract deep abstract features from high-dimensional, nonlinear network state characteristics, thereby more accurately capturing the complex mapping relationship between network states and optimal clustering schemes under different network scenarios. Simultaneously, the rich and diverse network state features, corresponding clustering schemes, and performance indicators stored in the historical optimization database provide high-quality training data for the deep neural network model. This allows the model to fully learn from historical experience, improving its generalization ability and decision accuracy, ensuring that it can output more intelligent and adaptable optimal clustering schemes when facing varied underwater acoustic environments, thereby effectively improving the overall operating efficiency and stability of the underwater acoustic sensor network.

[0037] For example, the underwater acoustic sensor network clustering model is a deep neural network model (such as CNN, DNN, or Transformer model), and the training process of the underwater acoustic sensor network clustering model may include: Samples are extracted from the historical optimization database. The quantitative indicators of "energy distribution characteristics, spatial distribution characteristics, and communication link characteristics" are used as input features (independent variable X), and the corresponding optimal clustering scheme is used as the output category (dependent variable Y). The dataset is divided into a training set (70%), a validation set (20%), and a test set (10%).

[0038] Configure the number of layers in the deep neural network (e.g., input layer, 3 hidden layers, output layer), the number of neurons in each layer, the activation function (e.g., ReLU function), the loss function (e.g., cross-entropy loss function), and the optimizer (e.g., Adam optimizer).

[0039] The training set is input into the model, the predicted output is calculated through forward propagation, the error is calculated based on the loss function, the model parameters are adjusted through backpropagation, and the training is iterated until the accuracy on the validation set converges (e.g., accuracy ≥ 92%).

[0040] Use the test set to verify the model's generalization ability and ensure that the model can output accurate clustering schemes for unseen network state features.

[0041] S130. Based on the current optimal clustering scheme, a cluster head node is elected through a dynamic role rotation mechanism to generate an adaptive network topology.

[0042] In an exemplary embodiment, the dynamic role rotation mechanism includes: Based on the node's remaining energy, location centrality, and historical role frequency, cluster head election weights are calculated using a multi-criteria evaluation function, and an optimal cluster head set is generated according to these weights.

[0043] Specifically, when the network topology change rate is detected to exceed a set threshold, or when the remaining energy of a node is lower than the dynamic energy threshold, cluster head reselection or cluster reconstruction is triggered to regenerate the optimal cluster head set.

[0044] Specifically, the dynamic role rotation mechanism is a strategy used in underwater acoustic sensor networks to periodically or event-drivenly select and update cluster head nodes, aiming to achieve balanced distribution of network load, extend the overall network lifespan, and improve network performance. Its core lies in ensuring that cluster head nodes always possess optimal communication and management capabilities through an intelligent election process, and preventing specific nodes from prematurely failing due to prolonged cluster head responsibilities.

[0045] Through the above technical solution, this application can comprehensively consider the remaining energy of nodes, location centrality, and historical role frequency, and calculate a more reasonable cluster head election weight through a multi-criteria evaluation function. This ensures that the selected cluster head nodes not only have sufficient energy and superior locations, but also achieve fair rotation of cluster head roles, effectively preventing specific nodes from prematurely exhausting their energy due to long-term cluster head roles, thereby significantly extending the overall network lifetime and improving load balancing. Furthermore, when the network topology change rate is detected to exceed a set threshold or the remaining energy of a node is detected to be below a dynamic energy threshold, cluster head re-election or cluster reconfiguration can be triggered in a timely manner, ensuring that the network topology can quickly adapt to dynamic changes in the underwater acoustic environment, avoiding communication interruptions or performance degradation due to outdated topology, thereby improving the network's robustness and adaptability.

[0046] S140. The performance of the adaptive network topology is evaluated to obtain a set of performance indicators.

[0047] In an exemplary embodiment, the performance evaluation metrics include at least network coverage, inter-cluster energy balance, packet delivery success rate, and estimated network lifetime.

[0048] Among them, the performance threshold is the design value set for each indicator.

[0049] The performance thresholds are design values ​​set for each individual metric and can be adjusted according to the actual application scenario. An example is shown below: Network coverage ≥ 90%; inter-cluster energy consumption balance ≥ 0.7; data packet delivery success rate ≥ 95%; estimated network lifetime ≥ 30 days; maximum communication latency ≤ 1s; when all indicators meet the corresponding thresholds, the performance is deemed to meet the requirements; if any indicator fails to meet the requirements, the performance is deemed to fail to meet the requirements.

[0050] Network coverage refers to the proportion of the entire target monitoring area that all nodes in an underwater acoustic sensor network can effectively monitor or communicate with. In underwater acoustic sensor networks, the actual coverage range of nodes may be affected by the complexity of the underwater environment (such as changes in sound speed, multipath effects, and noise interference). Therefore, evaluating network coverage requires comprehensive consideration of the node's sensing range, communication range, and underwater channel characteristics. For example, it can be obtained by calculating the ratio of the union of the sensing ranges of all nodes to the total monitoring area, or by using simulation to evaluate the actual coverage effect under different underwater environmental parameters.

[0051] Inter-cluster energy consumption balance measures the uniformity of energy consumption among different clusters in an underwater acoustic sensor network. In clustered networks, cluster head nodes typically undertake more communication and data fusion tasks, resulting in higher energy consumption. If energy consumption is uneven among clusters, some cluster head nodes may run out of energy prematurely, causing these clusters to fail and affecting the overall network lifetime and functionality. Inter-cluster energy consumption balance can be assessed by calculating indicators such as the standard deviation, variance, or maximum-to-minimum energy consumption ratio of each cluster. The goal is to minimize these indicators to extend the overall network lifetime.

[0052] In an underwater acoustic sensor network, packet delivery success rate represents the proportion of data packets sent from the source node that successfully reach the destination node. Underwater communication links are characterized by high latency, high error rates, and limited bandwidth, all of which can lead to packet loss. Therefore, packet delivery success rate is a key indicator of network communication reliability. It is typically calculated as the ratio of successfully received packets to the total number of packets sent. Improving this rate may require techniques such as retransmission mechanisms, error correction coding, or routing optimization.

[0053] Network lifetime estimate refers to the estimated time an underwater acoustic sensor network can continue operating while maintaining its basic functions (e.g., maintaining a certain level of coverage, connectivity, or data transmission capability). In energy-constrained underwater environments, network lifetime is a crucial metric for evaluating the quality of network design. Network lifetime estimate is typically calculated based on the initial energy of nodes, average energy consumption rate, network topology, and communication patterns. For example, it can be defined as the time it takes for the first node to run out of energy, or the time it takes for network coverage to drop below a certain threshold.

[0054] Performance thresholds are pre-defined, acceptable minimum standards or target values ​​for the aforementioned performance metrics. These thresholds are determined based on specific application requirements, system design goals, and environmental constraints. For example, network coverage might be set at no less than 90%; packet delivery success rate might be set at no less than 80%; inter-cluster energy balance might require a standard deviation less than a specific value; and estimated network lifetime might require at least several weeks or months. By setting explicit performance thresholds, it is possible to objectively determine whether the currently generated adaptive network topology meets actual operational requirements.

[0055] Through the above technical solution, this application, when evaluating the performance of adaptive network topologies, no longer relies solely on fuzzy or singular evaluation criteria, but introduces multi-dimensional and specific performance indicators, including network coverage, inter-cluster energy consumption balance, packet delivery success rate, and estimated network lifetime. These indicators comprehensively reflect the key performance characteristics of underwater acoustic sensor networks in complex underwater environments. Simultaneously, by setting clear performance thresholds for each indicator, the performance evaluation process becomes more objective and quantifiable. When any indicator fails to reach the preset threshold, the system can promptly identify the deficiencies of the current topology and trigger a complete optimization learning process, thereby avoiding the deployment of inefficient or unstable network configurations. This ensures that the generated adaptive network topology is not only theoretically optimal but also meets the requirements of high reliability, long lifetime, and efficient data transmission in actual operation, significantly improving the overall performance and robustness of the underwater acoustic sensor network.

[0056] S150. If the performance indicators meet the performance threshold, the adaptive network topology is confirmed to be effective, and the current network state characteristics, the current optimal clustering scheme, and the performance indicators are stored as new sample data in the historical optimization database.

[0057] S160. If the performance indicators do not meet the performance threshold, or if the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence level threshold, then the complete optimization learning process will be initiated.

[0058] In an exemplary embodiment, such as Figure 2 As shown, the complete optimization learning process includes: S161. Based on the current network state characteristics, a new optimal clustering scheme is determined through a multi-objective optimization algorithm.

[0059] In an exemplary embodiment, determining a new optimal clustering scheme using a multi-objective optimization algorithm may specifically include: Based on the current network status characteristics, a preliminary strategy direction is selected from multiple candidate clustering strategies according to a preset rule set. The multiple candidate clustering strategies include load balancing priority strategy, latency optimization priority strategy, energy consumption balancing priority strategy, spatial clustering-based clustering strategy, and hybrid optimization clustering strategy.

[0060] The load balancing priority strategy aims to evenly distribute the workload of each node in the network, such as data forwarding volume and cluster head responsibilities, to avoid local node overload leading to rapid energy depletion or communication congestion, thereby extending the overall stable operation time of the network. The latency optimization priority strategy focuses on minimizing the transmission latency of data from the source node to the destination node, which is crucial for underwater acoustic sensing applications with high real-time requirements. This strategy may be achieved by building shorter communication paths, selecting high-bandwidth links, or optimizing data forwarding mechanisms. The energy consumption balancing priority strategy aims to make the energy consumption of all nodes in the network as even as possible, preventing some nodes from failing prematurely due to energy depletion, thereby maximizing the overall lifespan of the network. The spatial clustering-based clustering strategy uses the physical proximity of nodes to group them into geographically compact clusters. This strategy can effectively reduce intra-cluster communication distance and energy consumption, and simplify the data aggregation task of the cluster head. The hybrid optimization clustering strategy combines the advantages of the above strategies, aiming to simultaneously optimize multiple conflicting network performance indicators, and is suitable for complex and variable underwater acoustic sensing network environments or environments with high requirements for multiple performance indicators.

[0061] The non-dominated sorting genetic algorithm NSGA-II is used to optimize the total network energy consumption, maximum communication latency and load balancing. It optimizes and selects candidate clustering schemes based on the initial selection strategy to determine the new optimal clustering scheme.

[0062] NSGA-II is a highly efficient multi-objective optimization algorithm capable of simultaneously handling multiple conflicting optimization objectives and generating a set of Pareto optimal solutions. The "improvement" here can be reflected in adjustments to the algorithm's encoding method, crossover and mutation operators, congestion distance calculation, or elite retention strategy to better adapt to the dynamics and uncertainties of underwater acoustic communication, thereby improving the algorithm's convergence speed and solution quality. Through this algorithm, an optimal trade-off can be found between total network energy consumption, maximum communication latency, and load balancing, thus obtaining the optimal clustering scheme that meets the current network requirements.

[0063] In an exemplary embodiment, the initial selection of a strategy direction from multiple candidate clustering strategies based on the current network state characteristics and a preset rule set may include: If the unevenness of energy distribution is higher than the first threshold and the difference in node density is higher than the second threshold, then a load balancing priority strategy is initially selected; if the spatial dispersion is higher than the third threshold and the communication distance is higher than the fourth threshold, then a latency optimization priority strategy is initially selected; if the overall network energy level is lower than the fifth threshold, then an energy consumption balancing priority strategy is initially selected; if the node distribution is determined to be clustered by a spatial clustering algorithm, then a clustering strategy based on spatial clustering is initially selected; if the network topology change rate is higher than the sixth threshold, then a hybrid optimization clustering strategy is initially selected.

[0064] Specifically, based on the current network state characteristics, a preliminary strategy direction is selected from multiple candidate clustering strategies according to a preset rule set, including the following scenarios: If the energy distribution unevenness is higher than the first threshold and the node density difference is higher than the second threshold, then the load balancing priority strategy is initially selected. This indicates that there are areas of uneven energy distribution and sparse / dense node distribution in the network. In this case, prioritizing load balancing can effectively avoid local hotspots and extend the network lifetime. If the spatial dispersion is higher than the third threshold and the communication distance is higher than the fourth threshold, then the latency optimization priority strategy is initially selected. When the node distribution is relatively scattered and the communication link is long, the communication latency will increase significantly, so prioritizing latency optimization is crucial. If the overall network energy level is lower than the fifth threshold, then the energy consumption balancing priority strategy is initially selected. When the network energy is generally low, balancing energy consumption is the most effective means to extend the overall network lifetime. If the node distribution is determined to have a clustered pattern through a spatial clustering algorithm, then a clustering strategy based on spatial clustering is initially selected. When nodes naturally form physical clusters, using a spatial clustering strategy can organize the network more efficiently and reduce communication overhead. If the network topology change rate exceeds the sixth threshold, a hybrid optimization clustering strategy is initially selected. In dynamic environments with frequent topological changes, a single strategy may be difficult to adapt, while a hybrid optimization strategy can provide stronger robustness and adaptability.

[0065] By introducing a strategy initialization mechanism based on the characteristics of the current network state, this application can intelligently select the most suitable optimization direction according to the complex environment of the underwater acoustic sensor network in real time. This avoids the computational overhead and convergence efficiency problems caused by blindly applying general optimization algorithms. Combined with the non-dominated sorting genetic algorithm NSGA-II, this application can efficiently find the Pareto optimal solution set among total network energy consumption, maximum communication latency, and load balancing, thereby generating an optimal clustering scheme that is more adapted to the current network conditions. This hierarchical optimization method significantly improves the intelligence and adaptability of the clustering process, ensuring that a high-performance adaptive network topology can be formed quickly and effectively in different network scenarios, thereby extending network lifetime, reducing communication latency, and improving the reliability of data transmission.

[0066] S162. Generate a new adaptive network topology based on the new optimal clustering scheme and evaluate its performance to obtain the corresponding new performance indicators.

[0067] S163. The verified new optimal clustering scheme, current network state characteristics, and new performance indicators are stored as new sample data in the historical optimization database, and the underwater acoustic sensor network clustering model is updated.

[0068] In an exemplary embodiment, updating the underwater acoustic sensor network cluster model refers to periodically or trigger-based retraining of the underwater acoustic sensor network cluster model using sample data accumulated in the historical optimization database, in order to optimize the model parameters of the underwater acoustic sensor network cluster model.

[0069] The use of accumulated sample data in the historical optimization database refers to a dynamically growing dataset that stores network state characteristics under different network scenarios, corresponding optimal clustering schemes, and performance metrics of these schemes. This data represents the experience continuously learned and accumulated by the system during actual operation, signifying network configurations and performance validated under various conditions. The accumulated sample data continuously enriches itself over time and with network operation, providing a solid foundation for the model's continuous learning.

[0070] Specifically, the underwater acoustic sensor network clustering model is retrained periodically or triggered, including two methods: periodic retraining and triggered retraining. Periodic retraining refers to retraining the underwater acoustic sensor network clustering model at preset time intervals (e.g., hourly, daily, weekly, or monthly). This method ensures that the model can regularly absorb the latest network operation data and environmental change information, thereby gradually adapting to the long-term evolution trend of the network. Triggered retraining refers to immediately initiating retraining of the underwater acoustic sensor network clustering model when specific conditions are met. These conditions may include, but are not limited to: significant changes in network topology, drastic fluctuations in node energy distribution, continuous decline in communication link quality, accumulation of new sample data reaching a preset threshold, or the confidence level of the clustering scheme output by the model continuously falling below the confidence level threshold. Triggered retraining enables the model to quickly respond to temporary events or key state changes, and promptly adjust its internal parameters to cope with new challenges. Regardless of whether periodic or triggered retraining is used, the retraining process involves using all or part of the accumulated sample data in the historical optimization database to readjust the model's internal parameters. This is typically accomplished through optimization algorithms (such as gradient descent and its variants) designed to minimize the error between the model’s predictions and the actual optimal clustering scheme, and to improve the model’s ability to generalize on unseen data.

[0071] Optimizing the model parameters of an underwater acoustic sensor network clustering model refers to the variables used within the clustering model for decision-making or prediction. For example, if the model is a deep neural network, the model parameters include the weights and biases of neurons in each layer. The goal of optimizing the model parameters is to enable the model to more accurately output the optimal clustering scheme given an input (the current network state characteristics) and to achieve better performance metrics in real-world networks. Through continuous optimization, the model can better capture the complex dynamic characteristics of the underwater acoustic sensor network, improving the accuracy and efficiency of clustering decisions.

[0072] Through the above technical solution, the underwater acoustic sensor network clustering model can systematically utilize the continuously accumulated sample data in the historical optimization database for periodic or triggered retraining. This mechanism ensures that the model parameters can be continuously optimized and adjusted, enabling it to better adapt to the dynamic changes in the underwater acoustic environment and the evolution of network topology. Therefore, the model can maintain high prediction accuracy and generalization ability, effectively avoiding performance degradation caused by environmental changes or data drift, thus continuously providing high-quality optimal clustering schemes and significantly improving the long-term operating efficiency, stability, and adaptability of the underwater acoustic sensor network.

[0073] Based on the same inventive concept, this application also provides an underwater acoustic sensor network clustering system for implementing the underwater acoustic sensor network clustering method described above. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more underwater acoustic sensor network clustering system embodiments provided below can be found in the limitations of the underwater acoustic sensor network clustering method described above, and will not be repeated here.

[0074] Please see Figure 3 , Figure 3 This application provides a functional module diagram of an underwater acoustic sensor network clustering system according to an embodiment of the present application. Specifically, the system may include: The feature acquisition module 301 is used to acquire the current network state features of each node in the target underwater acoustic sensor network. The current network state features include at least energy distribution features, spatial distribution features, and communication link features.

[0075] The scheme generation module 302 is used to input the current network state features into the underwater acoustic sensor network clustering model to obtain the current optimal clustering scheme; wherein, the underwater acoustic sensor network clustering model is obtained by training based on the historical optimization database.

[0076] The role switching module 303 is used to elect cluster head nodes and generate an adaptive network topology based on the current optimal clustering scheme through a dynamic role rotation mechanism.

[0077] The performance evaluation module 304 is used to evaluate the performance of the adaptive network topology and obtain performance indicators. If the performance indicators meet the performance threshold, the adaptive network topology is confirmed to be effective, and the current network state characteristics, the current optimal clustering scheme, and the performance indicators are stored as new sample data in the historical optimization database. If the performance indicators do not meet the performance threshold, or the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence level threshold, the complete optimization learning process is initiated.

[0078] The optimization learning module 305 is used to determine a new optimal clustering scheme based on the current network state characteristics through a multi-objective optimization algorithm; generate a new adaptive network topology based on the new optimal clustering scheme and perform performance evaluation to obtain the corresponding new performance indicators; store the verified new optimal clustering scheme, current network state characteristics and new performance indicators as new sample data in the historical optimization database, and update the underwater acoustic sensor network clustering model.

[0079] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a clustering method for underwater acoustic sensor networks.

[0080] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0081] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0082] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0083] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0084] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0085] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0086] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0087] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A clustering method for underwater acoustic sensor networks, characterized in that, The underwater acoustic sensor network clustering method includes: The process involves: acquiring the current network state characteristics of each node in the target underwater acoustic sensor network, including at least energy distribution characteristics, spatial distribution characteristics, and communication link characteristics; collecting the remaining energy and energy consumption rate of the nodes, and calculating the energy distribution non-uniformity and the overall network energy level based on the remaining energy and energy consumption rate to constitute the energy distribution characteristics; repeatedly collecting the three-dimensional Euclidean distance and depth gradient between nodes, and calculating the node density difference, spatial dispersion, and average communication distance based on the three-dimensional Euclidean distance and depth gradient collected in a single collection to constitute the spatial distribution characteristics; calculating the topology change rate based on the changes in the three-dimensional Euclidean distance and depth gradient collected over time; and collecting at least one of the communication link delivery success rate, received signal-to-noise ratio, path loss, and single-hop delay to constitute the communication link characteristics; wherein the energy distribution non-uniformity and spatial dispersion are used to evaluate the topological complexity of the target underwater acoustic sensor network. The current network state features are input into the underwater acoustic sensor network clustering model to obtain the current optimal clustering scheme; wherein, the underwater acoustic sensor network clustering model is obtained by training based on a historical optimization database; Based on the current optimal clustering scheme, a cluster head node is elected through a dynamic role rotation mechanism to generate an adaptive network topology. The performance of the adaptive network topology was evaluated to obtain a set of performance metrics; If the performance index meets the performance threshold, the adaptive network topology is confirmed to be effective, and the current network state characteristics, the current optimal clustering scheme, and the performance index are stored as new sample data in the historical optimization database. If the performance index does not meet the performance threshold, or if the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence threshold, then the complete optimization learning process is initiated. The complete optimization learning process includes: Based on the current network state characteristics, a new optimal clustering scheme is determined through a multi-objective optimization algorithm; Based on the new optimal clustering scheme, a new adaptive network topology is generated and its performance is evaluated to obtain the corresponding new performance indicators. The verified new optimal clustering scheme, the current network state characteristics, and the new performance indicators are stored as new sample data in the historical optimization database, and the underwater acoustic sensor network clustering model is updated. The updating of the underwater acoustic sensor network cluster model refers to: using the sample data accumulated in the historical optimization database to periodically or trigger retrain the underwater acoustic sensor network cluster model in order to optimize the model parameters of the underwater acoustic sensor network cluster model.

2. The clustering method of underwater acoustic sensor network according to claim 1, characterized in that, The underwater acoustic sensor network clustering model is a deep neural network model; The historical optimization database stores network state characteristics, corresponding clustering schemes, and performance metrics under different network scenarios.

3. The clustering method of underwater acoustic sensor network according to claim 1, wherein, The process of determining a new optimal clustering scheme using a multi-objective optimization algorithm specifically includes: Based on the current network state characteristics, a strategy direction is initially selected from multiple candidate clustering strategies according to a preset rule set; the multiple candidate clustering strategies include load balancing priority strategy, latency optimization priority strategy, energy consumption balancing priority strategy, spatial clustering-based clustering strategy, and hybrid optimization clustering strategy; The non-dominated sorting genetic algorithm NSGA-II is used to optimize the total network energy consumption, maximum communication latency and load balance. The algorithm optimizes and selects candidate clustering schemes based on the initial selection strategy to determine the new optimal clustering scheme. The step of initially selecting a strategy direction from multiple candidate clustering strategies based on the current network state characteristics and a preset rule set includes: If the energy distribution unevenness is higher than the first threshold and the node density difference is higher than the second threshold, then the load balancing priority strategy is initially selected. If the spatial dispersion is higher than the third threshold and the communication distance is higher than the fourth threshold, then the latency optimization priority strategy is initially selected. If the overall energy level of the network is lower than the fifth threshold, the energy consumption balancing priority strategy is initially selected. If the spatial clustering algorithm determines that the node distribution presents a clustering pattern, then the aforementioned clustering strategy based on spatial clustering is initially selected. If the network topology change rate is higher than the sixth threshold, then the hybrid optimization clustering strategy is initially selected.

4. The clustering method of underwater acoustic sensor network according to claim 1, wherein, The dynamic role rotation mechanism includes: Based on the node's remaining energy, location centrality, and historical role frequency, a cluster head election weight is calculated using a multi-criteria evaluation function, and an optimal cluster head set is generated according to the weight. Specifically, when the network topology change rate is detected to exceed a set threshold, or when the remaining energy of a node is lower than the dynamic energy threshold, cluster head reselection or cluster reconstruction is triggered to regenerate the optimal cluster head set.

5. The clustering method of underwater acoustic sensor network according to claim 1, wherein, The performance evaluation metrics include at least network coverage, inter-cluster energy balance, packet delivery success rate, and estimated network lifetime. The performance thresholds are design values ​​set for each indicator.

6. A water acoustic sensor network clustering system, characterized in that, Based on the underwater acoustic sensor network clustering method as described in any one of claims 1-5, the underwater acoustic sensor network clustering system includes: The feature acquisition module is used to acquire the current network state features of each node in the target underwater acoustic sensor network. These current network state features include at least energy distribution features, spatial distribution features, and communication link features. The module collects the remaining energy and energy consumption rate of the nodes, and calculates the energy distribution non-uniformity and the overall network energy level based on the remaining energy and energy consumption rate, constituting the energy distribution features. It also collects the three-dimensional Euclidean distance and depth gradient between nodes multiple times, and calculates the node density difference, spatial dispersion, and average communication distance based on the three-dimensional Euclidean distance and depth gradient collected in a single acquisition, constituting the spatial distribution features. Furthermore, it calculates the topology change rate based on the changes in the three-dimensional Euclidean distance and depth gradient collected over time. Finally, it collects at least one of the following communication link features: delivery success rate, received signal-to-noise ratio, path loss, and single-hop delay, constituting the communication link features. The energy distribution non-uniformity and spatial dispersion are used to evaluate the topological complexity of the target underwater acoustic sensor network. The scheme generation module is used to input the current network state features into the underwater acoustic sensor network clustering model to obtain the current optimal clustering scheme; wherein, the underwater acoustic sensor network clustering model is obtained by training based on a historical optimization database; The role switching module is used to elect cluster head nodes and generate an adaptive network topology based on the current optimal clustering scheme through a dynamic role rotation mechanism. The performance evaluation module is used to evaluate the performance of the adaptive network topology and obtain performance indicators. If the performance indicators meet the performance threshold, the adaptive network topology is confirmed to be effective, and the current network state characteristics, the current optimal clustering scheme, and the performance indicators are stored as new sample data in the historical optimization database. If the performance indicators do not meet the performance threshold, or the confidence level of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than the confidence threshold, the complete optimization learning process is initiated. The optimization learning module is used to determine a new optimal clustering scheme based on the current network state characteristics using a multi-objective optimization algorithm; generate a new adaptive network topology based on the new optimal clustering scheme and perform performance evaluation to obtain the corresponding new performance indicators; store the verified new optimal clustering scheme, the current network state characteristics, and the new performance indicators as new sample data in the historical optimization database, and update the underwater acoustic sensor network clustering model.

7. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the underwater acoustic sensor network clustering method according to any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the underwater acoustic sensor network clustering method according to any one of claims 1-5.

Images