Weak network communication method suitable for underwater acoustic channel model

Abstract

The invention discloses a weak network communication method suitable for an underwater acoustic channel model. The problems that an underwater acoustic channel is high in time delay, low in bandwidth, high in bit error rate, time-varying in link state, limited in node energy and the like are solved. According to the application, link state information and link connection information obtained by prediction are acquired regularly through a communication node; on the basis of the link state information, the link connection information, the to-be-forwarded data information and system feedback parameters used for real-time strategy adjustment, a forwarding strategy is generated through a mapping relation; processing the to-be-forwarded data according to R, L, S and P in a forwarding strategy, and sending the to-be-forwarded data to target underwater acoustic equipment through an underwater acoustic communication module; the processing comprises the following steps of: fragmenting the data into S data fragments according to the data fragment number S; and according to the number R of repeated data transmission times, respectively copying and transmitting S data fragments for R times, and determining a final transmission link for data transmission based on the number L of selectable links and the optimal path P.

Description

Technical Field

[0001] This application relates to the field of underwater acoustic communication technology, and in particular to a communication method suitable for underwater acoustic channel models. Background Technology

[0002] With the continuous development of underwater detection, monitoring and control technologies, the types of data carried on underwater acoustic communication networks have gradually become richer, expanding from traditional environmental data acquisition to real-time control commands, critical command responses, equipment status information and various types of data (configuration, executable programs, acquired data, etc.). These services have placed higher demands on the reliability and real-time performance of data transmission.

[0003] However, as a typical weak network environment, the underwater acoustic channel presents inherent physical characteristics that pose severe challenges to reliable real-time transmission, challenges never before encountered in terrestrial communications:

[0004] First, there is the problem of protocol failure caused by extremely high propagation delays. The speed of sound in water is approximately 1500 meters per second, resulting in round-trip delays of several seconds or even tens of seconds. This makes reliable transmission protocols based on handshake, acknowledge, and retransmission (ARQ) mechanisms, widely used in terrestrial networks, extremely inefficient in this environment. Because the communication link is almost idle during the long wait for acknowledgment signals, the effective throughput is extremely poor, completely failing to meet the timeliness requirements of real-time control services. Existing underwater acoustic communication methods mostly use fragmentation and acknowledgment mechanisms to improve reliability, but channel variations often lead to the continuous retransmission of data packets, further exacerbating the low transmission efficiency.

[0005] Secondly, there is the dilemma of transmission reliability arising from the coexistence of high bit error rate and extremely low bandwidth. Underwater acoustic channels are heavily affected by multipath effects and environmental noise, resulting in a bit error rate far exceeding that of radio communication. Simultaneously, their available bandwidth is extremely narrow and decreases sharply with increasing communication distance. This contradiction means that, on the one hand, simply increasing data redundancy or erasure coding would consume already valuable bandwidth resources, and the additional delay introduced by the data recovery process is unacceptable; on the other hand, transmitting large data packets over unreliable channels means that a single error can render the entire data packet unusable, leading to extremely low bandwidth utilization.

[0006] Secondly, there is the highly time-varying nature of link connectivity. Underwater communication nodes may experience random connection and disconnection due to temperature, ocean currents, their own movement, or malfunctions. Traditional routing strategies that react to changes based on the current network state are unable to keep up with the speed of link changes due to update and convergence delays, easily causing frequent intermittent communication link interruptions and failing to provide continuous and stable connections for critical services.

[0007] Furthermore, underwater nodes are typically energy-constrained. Nodes are mostly battery-powered, and recharging them is extremely difficult. Any transmission strategy that ignores energy constraints could lead to critical relay nodes running out of power prematurely, significantly shortening the entire network's lifespan.

[0008] However, existing transmission methods, including solutions specifically designed for underwater acoustic communication (e.g., patent document CN114584226A) and advanced terrestrial IoT solutions (e.g., patent document CN120201032A), all have the following fundamental limitations when facing the real-time reliable transmission requirements under the extreme conditions of the aforementioned underwater acoustic channels.

[0009] First, existing mechanisms are insufficient to meet the efficiency requirements of real-time services in dealing with high latency. Whether it's traditional underwater acoustic communication methods or newer network protocols based on the "tolerance for interruptions and delays" principle, their reliability relies heavily on acknowledgment and retransmission mechanisms, whose efficiency has been significantly reduced in high-latency environments. This results in extremely low utilization of the communication link during the lengthy waiting period for confirmation. Furthermore, while the "store-and-carry-forward" model used in some solutions can improve the final delivery rate, its inherent data retention characteristics make it completely unsuitable for latency-sensitive services such as real-time control.

[0010] Second, existing methods for improving the reliability of weak networks struggle to balance efficiency and real-time performance. Some solutions attempt to recover data through complex redundant coding and multi-node collaboration, but such solutions introduce unacceptable overhead and recovery delays in low-bandwidth, high-latency channels, failing to meet the timeliness requirements of services while ensuring reliability. Their routing strategies are mostly based on historical information or reactive adjustments to the current state, lacking the ability to proactively predict dynamic changes in the link, making it difficult to select and maintain a stable, optimal transmission path in rapidly changing channels.

[0011] Furthermore, existing solutions generally lack a comprehensive consideration of the global constraints of underwater acoustic networks. Many advanced solutions adapted from terrestrial networks fail to incorporate the core factor of node energy constraints, which determines the network's lifespan, into routing and transmission decisions. They also fail to effectively address the challenge of balancing conflicting optimization objectives such as bandwidth, latency, bit error rate, topology stability, and energy efficiency.

[0012] In summary, there is still a lack of a transmission method that can systematically solve the inherent problems of high latency, low bandwidth, high bit error rate, topology time-varying nature, and node energy constraints in underwater acoustic channels, while simultaneously meeting the comprehensive requirements of modern underwater services for data transmission reliability, real-time performance, and efficiency. Summary of the Invention

[0013] This application provides a weak network communication method suitable for underwater acoustic channel models, aiming to solve the inherent problems of high latency, low bandwidth, high bit error rate, topology time variation and node energy limitation in underwater acoustic channels, while meeting the comprehensive requirements of modern underwater services for data transmission reliability, real-time performance and efficiency.

[0014] This application provides a weak network communication method suitable for underwater acoustic channel models, including the following steps S1-S3.

[0015] Step S1: The communication node periodically acquires link status information and predicted link connectivity information.

[0016] Step S2: Based on the link status information, the link connectivity information, the data to be forwarded, and the system feedback parameters used for real-time adjustment of the strategy, a forwarding strategy is generated through a mapping relationship; wherein, the forwarding strategy includes the number of times data is repeatedly sent R, the number of available links L, the number of data fragments S, and the optimal path P.

[0017] Step S3: Process the data to be forwarded according to R, L, S and P in the forwarding strategy, and send it to the destination underwater acoustic device through the underwater acoustic communication module; wherein, the processing includes: dividing the data into S data fragments according to the number of data fragments S; copying and sending the S data fragments R times according to the number of times the data is repeatedly sent R; and determining the final transmission link for data transmission based on the number of available links L and the optimal path P.

[0018] Optionally, in the above scheme, the data is labeled with priority according to the urgency level U, and the sending order is arranged based on the priority.

[0019] Optionally, in the above scheme, the predicted link connectivity information includes the link connectivity probability predicted in the future time period T based on a graph neural network model.

[0020] In the above scheme, optionally, the data information to be forwarded includes at least the data type and data size; the system feedback parameters include at least one of node energy, transmission delay, link bit error rate, link congestion level and transmission success rate, which are used to adjust the subsequently generated forwarding strategy parameters in real time.

[0021] In the above scheme, optionally, the number of times the data is repeatedly sent, R, is determined by the following formula.

[0022] .

[0023] in, Indicates the basic number of repetitions; This represents the increment of the number of repetitions, determined by the bit error rate. Indicates the increment of the number of repetitions, which is determined by the data type; This indicates a reduction in the number of repetitions determined by the data size. This indicates the rounding up operation; This indicates the maximum number of repetitions.

[0024] In the above scheme, optionally, the number L of selectable links is determined by the following formula.

[0025] .

[0026] in, Indicates the total number of available links. This indicates the adjustment amount for the number of selected links, determined by bandwidth utilization requirements. It is calculated based on data size, allowed transmission time limit, bandwidth of each link, and link connectivity probability.

[0027] .

[0028] in, This indicates the allowed transmission time limit, which is set according to the data type and urgency level. Indicates the first Link bandwidth; Indicates the first The connectivity probability of the link, i.e. the link of the first link. Link connectivity prediction values ​​for each link. , ; Indicates the size of the data.

[0029] In the above scheme, optionally, the number of data fragments S is determined by the following formula.

[0030] .

[0031] in, This indicates the preset maximum number of shards. This represents the base number of shards calculated based on the data size and the shard size. , Indicates the data size. This represents the fragment size, which is related to the link's maximum transmission unit (MTU) and bit error rate, and is the smallest value among all links.

[0032] In the above scheme, the optimal path P can optionally be determined in the following way.

[0033] .

[0034] in, This indicates that set C is contained in all available link sets. , This indicates that set C contains L links; Represents the links in set C The utility value is determined by the system feedback parameters, the link status information, and the link connectivity information; i represents the sequence number of link l. .

[0035] In the above scheme, optionally, the urgency level U is divided into multiple priorities according to the data type; and when there are multiple data to be forwarded, a corresponding forwarding strategy is generated for each data to be forwarded; according to the forwarding strategy of each data, all data to be forwarded are processed into data blocks respectively; the method further includes: sorting the data blocks corresponding to all data to be forwarded according to their respective urgency levels, generating a global sending sequence and sending them sequentially, so that data blocks with higher priority are sent before data blocks with lower priority.

[0036] Optionally, the method further includes at least one of the following dynamic adjustment mechanisms:

[0037] a) Real-time dynamic adjustment: When the real-time status of the link is detected to deviate from the status information on which the forwarding strategy was generated, the deviation exceeds a preset threshold, at least one forwarding strategy parameter of the currently executing data transmission task is dynamically adjusted.

[0038] b) Perform policy calibration after a single transmission task: Based on the comparison between the actual transmission results of this task and the expected goals, update the mapping relationship or benchmark parameters used to generate forwarding policies in order to optimize the policy generation for subsequent tasks.

[0039] c) Iteratively train the graph neural network model used to predict link connectivity information: continuously update the parameters of the graph neural network model using accumulated historical transmission data and actual link status to improve its prediction accuracy.

[0040] d) Self-evolution of the decision rule base: Record strategy adjustment actions and their resulting performance effects, summarize effective adjustment rules through data analysis, and update the rule base.

[0041] Compared with the prior art, this application has at least the following beneficial effects.

[0042] Based on further analysis and research of existing technical problems, this application recognizes the current lack of a transmission method that can systematically solve the inherent challenges of high latency, low bandwidth, high bit error rate, time-varying link connections, and limited node energy in underwater acoustic channels, while simultaneously meeting the comprehensive requirements of modern underwater services for data transmission reliability, real-time performance, and efficiency. This application establishes an integrated, adaptive policy mapping mechanism that collaboratively maps multi-dimensional network conditions and service requirements (link conditions, link connection prediction, data characteristics, and system feedback) into a complete set of forwarding policy parameters. The core parameters include at least the number of data retransmissions (R), the number of available links (L), the number of data fragments (S), and the optimal path (P). This multi-dimensional policy coordination mechanism enables the system to achieve unified optimization of data transmission reliability, real-time performance, and efficiency in complex weak network environments.

[0043] Specifically, the beneficial effects of this application are reflected in the following aspects.

[0044] 1. A systematic solution is proposed: This application proposes a multi-input multi-output map() mapping model, which integrates multi-dimensional factors such as link status, routing, and data characteristics to generate a comprehensive and optimized forwarding strategy. The strategy parameters cover all dimensions of forwarding, including repetition count (R), number of links (L), number of fragments (S), optimal path (P), and urgency (U). Through the coordinated generation and execution of the R, L, S, P, and U parameters, a unified solution is constructed, which fundamentally avoids the contradictory problems of traditional solutions: relying on response mechanisms when dealing with high latency, failing to effectively utilize multiple links when dealing with low bandwidth, and simply adding redundancy when dealing with high bit error rates, leading to efficiency degradation.

[0045] 2. Closed-loop control and dynamic adaptation are achieved: This application introduces system feedback parameters (such as response delay, bit error rate, node energy, etc.) to form closed-loop control, which can make minute real-time adjustments to the strategy and has dynamic adaptation capabilities, thereby improving the robustness of the solution and the utilization rate of resources in time-varying network environments.

[0046] 3. Fundamental optimizations addressing core pain points: Addressing the core pain points of underwater acoustic communication, such as high latency, low bandwidth, and high bit error rate, a "non-acknowledgment-based retransmission mechanism (R)" is adopted. This avoids the overhead of high-latency acknowledgment while probabilistically ensuring transmission reliability. "Multi-link parallel transmission (L)" effectively increases the equivalent bandwidth, overcoming the bottleneck of low bandwidth in underwater acoustic channels. "Data fragmentation (S) coordinated with the number of links" is introduced, adapting large data blocks to multiple parallel links, further improving transmission success rate and efficiency. Through "optimal path selection (P)" that integrates multiple factors, the overall optimal balance between stability, latency, and energy consumption in data transmission paths is ensured. Attached Figure Description

[0047] Figure 1 This is a schematic diagram illustrating the application environment of a weak network communication method suitable for an underwater acoustic channel model, as provided in one embodiment of this application.

[0048] Figure 2 This is a flowchart illustrating a weak network communication method applicable to an underwater acoustic channel model, as provided in one embodiment of this application.

[0049] Figure 3 This is a generation logic diagram of a forwarding strategy for a weak network communication method applicable to an underwater acoustic channel model, provided in one embodiment of this application.

[0050] Figure 4 This is a data forwarding process diagram of a weak network communication method applicable to an underwater acoustic channel model, provided in the first embodiment of this application.

[0051] Figure 5 This is a data forwarding process diagram for a weak network communication method applicable to an underwater acoustic channel model, provided in the second embodiment of this application.

[0052] Figure 6 This is a data forwarding process diagram of a weak network communication method applicable to an underwater acoustic channel model, provided in the third embodiment of this application. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] In the description of this application: unless otherwise stated, "multiple" means two or more. Expressions such as "including", "comprising", and "having" also mean "not limited to" (certain units, components, materials, steps, etc.).

[0055] Before describing the specific technical solutions of this application, the relevant mathematical symbols and functions are defined as follows:

[0056] Explanation of the universality of logarithmic functions: In the fields of communication and information processing, using logarithmic functions for scaling is a common technique. Mathematically, for a given positive input value, calculations using logarithmic functions with different bases greater than 1, followed by appropriate linear scaling and translation, can map it to a predetermined target interval (e.g., between 0 and 1). This is because there is only one constant coefficient proportional relationship between logarithmic functions with different bases.

[0057] Specific definitions in this application: Although the above general principles exist, in order to ensure the clarity, consistency and reproducibility of the technical effects described in this specification, unless otherwise specified, throughout the entire application document:

[0058] The logarithmic function symbol "" appears in all formulas "" refers specifically to the natural logarithm with the natural constant e as its base, which is equivalent to " " or" ".

[0059] All embodiments and performance data provided in this application are calculated and verified based on this definition.

[0060] The specific parameters (such as scaling factor and offset) for transforming the output value of the logarithmic function to the target interval are designed and optimized for this natural logarithmic function with base e.

[0061] It should be noted that the core concept of this application lies in mapping input data to a target interval through a combination of logarithmic transformation and parametric scaling. Those skilled in the art will understand that, when implementing this application, other logarithmic functions with bases greater than 1 (e.g., 2, 10, etc.) can also be chosen. In this case, the technical effect of this application can be achieved by adaptively adjusting the scaling and translation parameters (e.g., based on the conversion relationship between the base and the natural constant e). Such adaptive parameter adjustment based on the teachings of this application should be considered to fall within the protection scope of this application.

[0062] This application addresses the typical challenges of weak network communication in underwater acoustic communication networks, such as high latency, low bandwidth, high bit error rate, time-varying topology, and limited node energy. It provides a method for generating forwarding strategy parameters based on a mapping of multiple factors, including link status, service type, network topology, and feedback mechanism. These parameters include the number of times data is repeatedly transmitted (R), the number of available links (L), the number of data fragments (S), the optimal path (P), and the urgency level (U). This method meets the requirements for forwarding efficiency, reliability, and real-time performance of data transmission under weak network communication conditions.

[0063] This application provides a hierarchical, configurable data transmission scheme. Its core inventive point lies in using a multi-factor mapping relationship and collaboratively executing a set of forwarding strategy parameters. This set of strategy parameters at least completely includes the number of data retransmissions (R), the number of available links (L), the number of data fragments (S), and the optimal path (P). [Reference] Figure 2These four parameters together form an organic whole. Reliability is improved through R, equivalent bandwidth and transmission success rate are enhanced through the synergy of L and S, and path optimization is ensured through P, thus systematically addressing the core challenges of weak underwater acoustic networks. Furthermore, the mapping relationship described in this application is based on a multi-factor collaborative decision-making model. This model comprehensively considers multiple dimensions such as link status, topology prediction, data characteristics, and energy constraints; it is not a simple linear superposition, but rather... The function realizes the coordination and trade-off among factors, and finally generates a set of mutually complementary and overall optimal forwarding strategy parameters (R, L, S, P).

[0064] Building upon this, this application further discloses urgency level (U) as an optional extended optimization parameter. Those skilled in the art should understand that these extended parameters can be introduced into the aforementioned core process based on the actual network environment, service requirements, and processing capabilities to achieve more refined transmission control. Regardless of whether urgency level (U) is introduced, as long as the core mapping steps are implemented to generate and coordinate the use of the parameter set R, L, S, and P, it falls within the scope of the technical solution constructed in this application. The embodiments and accompanying drawings described below are merely illustrations of the feasibility of the aforementioned core and extended solutions, and not the sole limitation of this application.

[0065] The communication nodes of an underwater acoustic communication network consist of modules for underwater acoustic communication, link management, routing management, and data forwarding. These, combined with modules for sensors, controllers, and data processing, form underwater acoustic equipment for detection, control, and processing. The link management module acquires link status information such as bandwidth, latency, packet loss rate, and bit error rate, and predicts link connectivity. The routing management module acquires information such as network topology, routing tables, and optimal paths. The data forwarding module performs functions such as data analysis, policy generation, forwarding execution, data processing, and data recovery. Policy generation involves generating forwarding policy parameters by acquiring routing information, link status, and system feedback parameters. The underwater acoustic communication module performs data transmission, data reception, and data recovery. Figure 1 This is a typical application scenario.

[0066] Therefore, this application first provides a communication node for an underwater acoustic communication network, including: an underwater acoustic communication module, a link management module, a routing management module, and a data forwarding module.

[0067] The link management module is used to acquire link status information and predict link connectivity information; the link management module periodically samples the attributes of each link, and the sampling period is denoted as [missing information]. (e.g., 120 s); The link management module saves the most recent W sampling records (e.g., W=50) for prediction and statistics.

[0068] The link status information includes at least one of bandwidth, latency, packet loss rate, and bit error rate.

[0069] The link connectivity prediction includes the probability of link connectivity within a future time period T, predicted based on a graph neural network model.

[0070] The preferred graph neural network model is a message-passing type, containing 3 message-passing layers, with a hidden dimension of 64, and a learning rate of [missing information]. , The activation function is ReLU. The edge output layer of the graph neural network model is used to predict the connectivity probability of each link within a future time window of T (e.g., 120 seconds).

[0071] The routing management module is used to obtain routing information, which includes at least one of network topology, routing table and optimal path.

[0072] The data forwarding module is communicatively connected to both the link management module and the routing management module. It is used to obtain link status information from the link management module and routing information from the routing management module, generate forwarding policy parameters based on the link status information and the routing information, and execute data forwarding operations based on the forwarding policy parameters. The module includes a controller and a data processing unit. The controller is used to generate forwarding policy parameters based on the link status information and the routing information. The data processing unit is used to process data based on the forwarding policy parameters and transmit it through the underwater acoustic communication module.

[0073] The underwater acoustic communication module is communicatively connected to the data forwarding module and is used to send data processed by the data forwarding module. In one embodiment, the underwater acoustic communication module is further configured to: receive underwater acoustic signals and recover data, and submit the recovered data to the data forwarding module.

[0074] like Figure 1 As shown, the underwater acoustic communication node in this embodiment adopts a layered architecture, including a control layer and a routing management module, a data forwarding module, a link management module, and an underwater acoustic communication module located thereunder. The data forwarding module communicates with the routing management module and the link management module, and ultimately accesses the underwater acoustic channel through the underwater acoustic communication module. The source underwater acoustic device and the destination underwater acoustic device communicate in the underwater acoustic channel through the above-described communication node structure.

[0075] based on Figure 1 In the application scenario where the source underwater acoustic device and the target underwater acoustic device communicate in the underwater acoustic channel through the above-mentioned communication node structure, this application proposes a weak network communication method suitable for the underwater acoustic channel model, which includes the following steps.

[0076] Step 1: Communication nodes periodically acquire link status and link connectivity prediction information. The communication nodes obtain link status information and link connectivity prediction information (predicted link connectivity information) between the underwater acoustic communication modules through the link management module. Link status information includes bandwidth, latency, packet loss rate, jitter, bit error rate, encoding method, link utilization, etc. The routing management module obtains routing information between communication nodes.

[0077] Step 2: The data forwarding module of the source underwater acoustic device generates forwarding strategy parameters, analyzes the received sensor data, processes the corresponding sensor data according to the strategy parameters, and then sends it to the destination underwater acoustic device through the underwater acoustic communication module. The destination underwater acoustic device recovers the data from the underwater acoustic communication module and sends it to the data processing module for processing through the data forwarding module.

[0078] Step 3 involves generating a forwarding strategy, with key factors including current link status information, link connectivity prediction information, routing information, data to be forwarded information, and system feedback parameters, as shown below.

[0079] (1) Link status ( Bandwidth, latency, packet loss rate, jitter, bit error rate, coding scheme, link utilization, MTU, etc., are used to evaluate channel quality.

[0080] (2) Link connectivity prediction information (Topology): The probability of link connectivity within the next time period T; the connectivity status of underwater acoustic nodes may change due to ocean currents, movement, or failures, resulting in extremely high latency in route updates. Therefore, based on the predicted link connectivity, it can be used to consider the stability of the path. The link connectivity prediction information is generated by a prediction module based on a graph neural network model. It uses historical link status data, node location information, etc., to predict link connectivity and outputs the probability of link connectivity within the next time period T, thereby enhancing link connectivity prediction and path stability assessment.

[0081] (3) Routing information to the target underwater acoustic communication node ( ); used to determine available paths.

[0082] (4) Data information to be forwarded: including data type ( ) and data size ( This includes various types of data and their sizes, such as real-time control commands, critical command responses, device status information, configurations, executable programs, and collected data. Data type and size information, along with other data to be forwarded, are used to determine transmission requirements.

[0083] (5) System feedback parameters ( The system feedback parameters, including node energy, transmission delay, link error rate, link congestion level, and transmission success rate, are used to make real-time minor adjustments to the generation strategy to optimize forwarding efficiency, reliability, and timeliness.

[0084] The forwarding strategy generation process is shown in the following formula.

[0085] .

[0086] The mapping method can be abstracted into a multi-input multi-output system model, such as Figure 3 As shown, this multiple-input multiple-output (MIMO) implementation can be achieved through neural networks or artificial intelligence methods, or through multi-input fuzzy matching. For example, when using a neural network, the training data for the neural network comes from at least one of historical transmission logs, simulation environment-generated data, and expert-annotated data. The neural network includes an input layer, at least one hidden layer, and an output layer. The input layer receives normalized link status, link connectivity prediction information (predicted link connectivity information), routing information, data type, data size, and system feedback parameters. The output layer generates the forwarding strategy parameters. When using fuzzy matching, the fuzzy rule base contains 20 core rules, covering at least four typical scenarios: high error rate scenarios, low bandwidth scenarios, energy-constrained scenarios, and real-time service scenarios.

[0087] The forwarding strategy includes parameters such as R, L, S, P, and U, which are described below.

[0088] 1) Number of Data Retransmissions (R): Taking into account factors such as bit error rate (including current and system feedback), data type, and data size, a non-acknowledgment mechanism for data retransmission is adopted to optimize reliability and latency. Given the high latency and unreliability of underwater acoustic channels, increasing the number of data retransmissions probabilistically improves transmission reliability. At the same time, the absence of an acknowledgment mechanism also reduces the effective transmission latency.

[0089] In underwater acoustic communication, fragmentation is typically performed before replication. The main reasons include: operating at the fragmentation level reduces peak memory usage; and fragmentation provides finer-grained redundancy due to the high bit error rate of underwater acoustic channels. Therefore, when processing data to be forwarded according to the forwarding strategy, the data is first divided into S data fragments based on the number of fragments S. Then, based on the number of times the data is repeatedly transmitted R, R copies are generated for each of the S data fragments. Finally, the resulting... Each data block unit determines the final transmission link for data transmission based on the number of available links L and the optimal path P.

[0090] The number of times the data is repeatedly sent, R, is determined by the following formula.

[0091] .

[0092] in: Indicates the base number of repetitions, which defaults to 1.

[0093] This represents the increment of the number of repetitions, determined by the bit error rate (BER).

[0094] ;

[0095] is the target reception success rate given by the system, which can be set to 0.99. BER is the bit error rate, which is usually taken as the maximum value of the link bit error rate in path P to ensure reliability in the worst case. The calculation method of path P is shown in 4) Optimal Path (P). This refers to the fragment size, and the calculation method is described in 3) Number of data fragments (S). It converts the number of fragment bytes into the number of fragment bits. This calculates the probability that a single fragment will be successfully received in a single transmission. This indicates the minimum number of repetitions required to achieve the target reception success rate.

[0096] This indicates the increment of the number of repetitions determined by the data type. For example, the increment of the number of repetitions for priority 1 data is 2, the increment for priority 2 data is 1, the increment for priority 3 data is 0, the increment for priority 4 data is -0.5, and the increment for priority 5 data is -1.

[0097] This represents a reduction in the number of repetitions determined by the data size. Its purpose is to avoid excessive network overhead when transmitting large amounts of data by appropriately reducing the number of repetitions. Specifically, this is determined by querying a pre-defined mapping table, for example... Bytes represent small data packets, the overhead of which is very small, so there is no need to reduce the size. ; This indicates a medium-sized data packet, at which point overhead is considered, and a slight reduction in packet size is implemented. ; This indicates that a larger data packet is being reduced in size. ; This indicates a large data packet, which needs to be significantly reduced in size. ; This indicates an extremely large data packet, which is significantly reduced to strictly control the total transmission volume. .

[0098] This indicates the maximum number of repetitions, for example, it can be set to 5 times.

[0099] This indicates the rounding up operation.

[0100] 2) Number of available links (L): Taking into account factors such as the type of data to be forwarded, the size of the data to be forwarded, the total number of available links, the link bandwidth, and the link connectivity probability, the equivalent bandwidth is improved and the transmission time is reduced by using multiple links in parallel transmission.

[0101] The number of selectable links L is determined by the following formula.

[0102] .

[0103] in, This indicates the total number of available links.

[0104] This indicates the bandwidth utilization demand, specifically the adjustment amount of the number of available links determined by the bandwidth utilization demand. The calculation is based on data size, allowed transmission time limit, bandwidth of each link, and link connectivity probability, as shown in the following formula.

[0105] .

[0106] This formula represents the ratio of the required bandwidth to the equivalent total bandwidth of all available links, reflecting how many links need to be used simultaneously to meet transmission time requirements under current network conditions. This indicates the allowed transmission time limit, which is set according to the data type and urgency level; details are as follows.

[0107] Priority 1: ;

[0108] Priority 2: ;

[0109] Priority 3: ;

[0110] Priority 4: ;

[0111] Priority 5: ;

[0112] Indicates the first Link bandwidth.

[0113] Indicates the size of the data.

[0114] Indicates the first The connectivity probability of the link, i.e. the link of the first link. Link connectivity prediction values ​​for each link. , .

[0115] 3) Number of data fragments (S): Taking into account information such as data size, link status, and the number of available links L, the data is fragmented to improve the transmission success rate, adapt to multi-link transmission, and optimize transmission efficiency.

[0116] The number of data fragments S is determined in the following way.

[0117] .

[0118] This represents the fragment size, which is related to MTU (Maximum Transmission Unit) and bit error rate, and is the minimum value among all links.

[0119] Slice size ( This is determined through the following steps.

[0120] Single-link constraint analysis: For each selected transmission link , Indicates link The serial number, Calculate the maximum allowed fragment size for each. This value is the minimum of the following two.

[0121] a) The effective maximum transmission unit of this link ( ).

[0122] b) The maximum fragment size, determined by the bit error rate constraint, is expressed by the formula: maximum fragment size. The calculation shows that, among which The error rate threshold is a system parameter (e.g., 0.1) set to ensure reliability. Indicates link The bit error rate.

[0123] Global unified decision: After completing the single-link constraint analysis of all L selectable links, the smallest allowed maximum fragment size is taken as the final data fragment size. This design ensures that each generated data fragment meets the requirements for transmission reliability and feasibility on the worst-quality link.

[0124] When allocating fragments to links, distribute them as evenly as possible. For example, each link should be allocated at least [amount missing]. The remaining S mod L fragments are distributed in turn to the first few links.

[0125] If the error rate is very high, it will lead to If the value of S is very small, then the number of fragments S might be very large. Therefore, it is necessary to limit S so that the number of fragments cannot exceed the maximum value. Therefore, the number of data fragments S is given by the following formula.

[0126] .

[0127] 4) Optimal Path (P): Based on factors such as the link status of available links, link connectivity prediction information, routing information, and system feedback parameters, the top L transmission paths are selected. The optimal path P is given by the following formula.

[0128] .

[0129] The above formula represents the set of available links. Choose a subset C containing L links such that the sum of the utilities of all links in this subset is maximized. In other words, the optimal path set P is the set of paths from all available links. In the process, select all possible subsets C containing L links, then calculate the sum of the utility values ​​of the L links in each subset C, and finally select the subset C with the largest total utility value.

[0130] Indicates link The effectiveness of a link is determined by factors such as link bandwidth, link congestion level, predicted link connectivity probability, historical transmission success rate, current transmission latency, remaining node energy, and link error rate.

[0131] ;

[0132] Where i represents the sequence number of link l, ; , indicating link Bandwidth normalization; Indicates link The effective bandwidth has already taken link quality into account; Represents a set of links The maximum bandwidth in.

[0133] Indicates link The predicted connectivity probability, . Indicates link Historical transmission success rate . Indicates link The bit error rate is usually very small, but if the bit error rate is very high, it will significantly reduce the utility. Indicates link Predicted connectivity probability The weight, Indicates link Historical transmission success rate The weight, Indicates bit error rate factor The weight, , ,and When the network changes rapidly and the link is unstable, historical data is trusted more, increasing [the effectiveness / advantage]. When the prediction model is accurate, we have more confidence in the predicted values ​​and increase [their confidence]. When more attention is paid to the bit error rate, its weight should be increased; for example, in the initial settings. , , The following table 1 shows an example of the dynamic adjustment rules.

[0134] ;

[0135] Indicates link The current transmission latency. The latency sensitivity coefficient is used for scoring. For services with high real-time requirements, A network with relatively large latency and small latency variation should be used. The system response latency can be relatively small, but the fluctuation is large. It should be moderate. For example, The initial value is set to ,in It is the median of all link delays, and the dynamic adjustment rules are shown in Table 2 below.

[0136] ;

[0137] Representing links respectively The ratio of remaining energy between the source node and the destination node. , . , indicating link Normalization of congestion levels Indicates link The degree of congestion, Represents a set of links The largest congestion in the system.

[0138] The weight parameters , , , and The assignment follows a hierarchical decision-making mechanism.

[0139] The first level is the determination of the baseline weights. The baseline weight values ​​are determined based on the network status. For example, the specific correspondence is shown in Table 3 below.

[0140] ;

[0141] The second level involves adjusting the baseline weight proportionally based on the type of data to be forwarded, for example:

[0142] ;

[0143] The third level involves dynamically fine-tuning the weights based on system feedback parameters. For example, if the recent average transmission success rate is lower than a threshold (e.g., 0.8), then the weights are increased. The weights of the data (e.g., multiplied by a factor of 1.2) are also considered. Meanwhile, to maintain a balance in the total weights and facilitate resource reallocation, the system will adjust the weights of other non-core dimensions differently based on the business type. For example, for businesses with extremely high real-time requirements (such as types 1 and 2), the weights will be appropriately reduced. The weighting is adjusted accordingly (e.g., multiplied by 0.9); for other business functions, the weighting is appropriately reduced. The weight is increased (e.g., multiplied by 0.9). If the ratio of average transmission delay to expected delay exceeds a threshold (e.g., 1.5), the weight is increased. The weight is increased (e.g., multiplied by 1.3) to more severely penalize high-latency links in subsequent decisions. As a coordination mechanism, the system will correspondingly reduce the weight slightly. The weight is calculated (e.g., multiplied by 0.9). If the energy consumption rate exceeds a threshold (e.g., 0.8), the weight is increased. The weight of the factor (e.g., multiplied by 1.4) influences the decision-making process, making it more inclined towards energy conservation. To allow room for improvements in the energy dimension, the system simultaneously reduces the weight of the factor. and To meet the requirements, the weights of both should be significantly reduced (e.g., multiplied by 0.8). If the overall network congestion exceeds a threshold (e.g., 0.7), then the weights should be increased. The weight is increased (e.g., multiplied by 1.3) to more proactively avoid congestion when selecting links. Accordingly, the system reduces... The weight (e.g., multiplied by 0.8) is used to suppress the excessive pursuit of high bandwidth in congested environments to some extent.

[0144] The fourth level is constraint guarantee, for example, ultimately ensuring that each weight value satisfies... Normalize all weights after constraint processing, so that .

[0145] 5) Urgency Level (U): Determines priority based on data type to optimize timeliness. The order in which multiple data items are forwarded simultaneously is determined by this urgency level parameter. Five priority levels are defined, from 1 to 5, with priority 1 being the highest and priority 5 being the lowest.

[0146] Priority 1 (highest): System safety and real-time control data, with extremely high real-time and reliability requirements, and typically small data volume. Examples include real-time control commands.

[0147] Priority 2: Critical task instructions, with high real-time and high reliability requirements and small data volume, such as critical instruction responses.

[0148] Priority 3: Important status monitoring information, with medium real-time and reliability requirements and a small data volume, such as equipment status information.

[0149] Priority 4: Routine task data, with lower real-time requirements but higher reliability requirements, and potentially larger data volume, such as non-real-time collected data like scientific observation data and image / video data.

[0150] Priority 5 (lowest): Background and non-real-time data, low real-time requirements, medium reliability requirements, and potentially large data volume, such as configuration information, executable programs, and other large files.

[0151] Due to the characteristics of underwater acoustic channels (high latency, low bandwidth, high bit error rate), strategy generation needs to comprehensively consider the above factors. For example: when the bit error rate is high, R and S are increased, while P focuses more on reliability; when the data volume is large, L and S are increased, R may be reduced, while P focuses more on bandwidth and congestion; when the data is real-time control type, U is increased, and a low-latency path P may be selected.

[0152] Step four, assuming there is The process of the data forwarding module using forwarding strategies for each piece of data to be forwarded is described by the following formula.

[0153] .

[0154] These are several fragments or duplicates of data generated according to strategy parameters. It is a method for generating forwarding data based on policy parameters. It is the first One forwarded data, .

[0155] Each piece of data to be forwarded corresponds to a set of forwarding strategy parameters. Based on the forwarding strategy parameters of fragmentation, replication, routing, and urgency labeling, the data to be forwarded is split into several data blocks, and each data block is labeled with routing and urgency.

[0156] when After processing all the data to be forwarded, all data blocks are sorted according to their urgency to form a forwarding sequence. Each of them sends out its own data through the corresponding link according to its chosen route.

[0157] In summary, the core of this application lies in the fact that, through the aforementioned multiple-input multiple-output... The mapping model generates a cooperative strategy consisting of five mutually coupled key parameters (R, L, S, P, U), and further enables the strategy to intelligently adapt to the high time-varying nature of the underwater acoustic channel through a closed-loop dynamic adjustment mechanism.

[0158] The selection of the optimal path set P is determined by the number of times data is repeatedly sent R and the size of the data fragments. The number of fragments, S, provides a key constraint. This is crucial in calculating the increment of repetitions determined by the bit error rate. At this time, the preferred bit error rate value is the maximum bit error rate of all links in the path set P, ensuring that even on the selected worst-quality link, data can still achieve the target success rate through repeated transmission. To ensure end-to-end reliability under worst-case scenarios. Data fragment size. The minimum value among the effective MTU and the bit error rate tolerance limits of all links in path set P is taken. Based on this, the size of the data to be transmitted is also considered. This allows us to determine the number of data fragments, S.

[0159] Data types Directly determines the urgency level U of the data and the corresponding allowable transmission time limit and generate data type adjustment amount Data size and Together, they determine the bandwidth required to complete the transmission, which is then expressed by the formula. The driver can calculate the number of links L to determine how many links need to work in parallel to meet the time limit requirements. Meanwhile, The path utility function is dynamically modified through a second-level weight adjustment mechanism. , , , and The weighting ratios of these factors influence the preference for selecting the optimal path set P (e.g., assigning a higher delay weight to real-time control commands). (to select the lowest latency path).

[0160] Parameters L and S are jointly responsible for rationally distributing the data load across multiple link resources. The number of available links, L, determines the number of links available for parallel transmission, while the number of data fragments, S, determines the amount of data to be transmitted. The system needs to process S data fragments and their generated R replicas (totaling...). Data blocks (units) should be evenly distributed across the selected L links as much as possible. Furthermore, the total load... This will affect the network congestion level. This information, as a system feedback parameter, will in turn affect the decision on L in subsequent policy generation and the path congestion level. The assessment.

[0161] The calculation of the number of times data is repeatedly sent, R, also takes into account the correlation with the number of data fragments, S. When the data is fragmented into many pieces (S is larger), it will be processed through... By appropriately reducing the number of repetitions of parameters, coordinated control of total transmission overhead can be achieved.

[0162] In addition, the strategy is dynamically optimized based on real-time system feedback parameters.

[0163] Real-time fine-tuning and rapid response are performed during transmission to quickly compensate for network emergencies and ensure the completion of each transmission task. For example, if the bit error rate of a used link is detected to be significantly higher than the value at the time of policy generation (e.g., exceeding the threshold of 50%) during data transmission, the number of retransmissions R of data fragments currently being transmitted on this link and those awaiting transmission is increased. If a link in path P is detected to be completely interrupted (e.g., the connectivity probability drops to zero), the faulty link is immediately removed from P, and the process is adjusted according to the utility function. In real time, a new set of effective paths is formed by selecting the best alternative from the set of backup links. This ensures that the number of parallel links L remains stable without interrupting the overall transmission, or that the transmission path is quickly reconstructed within the allowable range of L.

[0164] After a complete data transmission task is completed, the strategy generation logic is reviewed and calibrated based on the overall performance feedback. The actual transmission success rate is compared with the target success rate expected when generating the strategy. If the actual success rate remains below the target value, the base number of repetitions will be automatically increased. or adjustment The conservative coefficients in the calculations cause the system to tend to generate more redundant strategies in subsequent similar scenarios, thus gradually approaching and stabilizing at the target reliability level. Analyzing the macroscopic network state reflected in this task (such as excessively high average latency and frequent congestion) dynamically adjusts the parameters in the utility function. , , , and The baseline value allows the path selection preference to better match the long-term operating characteristics of the current network.

[0165] By accumulating historical operational data, the core model undergoes iterative evolution, enhancing the overall intelligence level of the system. The predicted connectivity probability (TP) output by the link connectivity prediction module (GNN model) is continuously compared with the actual link connectivity status. When the statistical value of the prediction error (such as root mean square error) exceeds a set threshold, the system automatically triggers an incremental learning or retraining process for the model, updating model parameters using the latest historical data to continuously improve prediction accuracy and provide more reliable forward-looking input for strategy generation. The data forwarding module records each triggered adjustment action and its final performance effect. Through long-term data analysis and mining, the system can automatically identify and eliminate inefficient or ineffective adjustment rules, while summarizing and refining new rules based on successful adjustment cases, enabling decision-making capabilities to continuously evolve over time.

[0166] Through the aforementioned deeply coupled parameter design and closed-loop adaptive adjustment mechanism, this application transforms the underwater acoustic network forwarding strategy from static configuration to a dynamic self-optimizing intelligent process. This effectively overcomes the core challenges of poor adaptability of traditional methods under high latency, high bit error rate, and time-varying channels, significantly improving communication reliability, real-time performance, and network resource utilization efficiency.

[0167] Example 1

[0168] A single piece of data, data1, is forwarded via two narrowband links with good link quality. Since the underwater acoustic channel is a time-varying narrowband channel, to improve communication efficiency, data is often split into blocks of several hundred bytes and forwarded sequentially, thereby improving channel utilization and data transmission success rate. In this embodiment, data1 is divided into several data blocks of default size based on the link status. Because the two links have similar states, parallel forwarding can be used to ensure the reliability of data forwarding. The processing of data1 is as follows... Figure 4 As shown.

[0169] Example 2

[0170] Two data sets, data2 (normal data) and data3 (real-time control commands), are forwarded via a well-maintained narrowband link. Data2 is split into several data blocks of default size, while data3, smaller than the default blocks, is not fragmented. Since data3 is a real-time control command and has a higher urgency than normal data, data3 blocks are forwarded with priority over data2 blocks. Furthermore, to improve the reliability of data forwarding, data3 is forwarded twice, while data2 is forwarded only once. The forwarding process for data2 and data3 is as follows: Figure 5 As shown.

[0171] Example 3

[0172] Feedback parameters in a communication system can fine-tune and optimize transmission strategies. In data transmission based on a forwarding strategy, the destination device replies with corresponding data status, link status, protocol status, etc. The source device collects and statistically analyzes these feedback parameters to fine-tune and optimize the transmission strategy. For example, when continuously forwarding control commands, as the link error rate decreases, the link utilization efficiency can be improved by reducing the number of subsequent control commands copied, thus enabling the transmission of more control commands per unit time. The working mechanism is as follows: Figure 6 As shown.

[0173] In this embodiment, the initial state (high bit error rate): At the start of communication, the link quality may be poor (high bit error rate). To ensure that control commands can be reliably delivered, the system adopts a relatively "conservative" strategy, namely, increasing the number of retransmissions (R). For example, each control command is copied and sent 3 times by default.

[0174] Feedback mechanism: After receiving data, the destination device will send a "feedback report" back to the source device. This report contains information such as "link status" (including the bit error rate).

[0175] Dynamic adjustment (core): The source device continuously monitors these feedback parameters.

[0176] When the system detects that the bit error rate is continuously decreasing (indicating that the link quality is improving), it will determine: "The channel conditions are now good, and reliable reception can be guaranteed without sending so many copies."

[0177] Therefore, the system dynamically reduced the number of times the "subsequent control command" was repeatedly sent (R), for example, from 3 times to 2 times, or even 1 time.

[0178] This reduces redundant transmissions, meaning that more different control commands can be sent per unit of time, thereby improving the "effective information transmission efficiency" of the link.

[0179] 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.

Claims

1. A weak network communication method suitable for underwater acoustic channel models, characterized in that, include: Step S1: The communication node periodically acquires link status information and predicted link connectivity information; Step S2: Based on the link status information, the link connectivity information, the data to be forwarded, and the system feedback parameters used for real-time adjustment of the strategy, a forwarding strategy is generated through a mapping relationship; wherein, the forwarding strategy includes the number of times data is repeatedly sent R, the number of available links L, the number of data fragments S, and the optimal path P; Step S3: Process the data to be forwarded according to R, L, S and P in the forwarding strategy, and send it to the destination underwater acoustic device through the underwater acoustic communication module; wherein, the processing includes: dividing the data into S data fragments according to the number of data fragments S; copying and sending the S data fragments R times according to the number of times the data is repeatedly sent R; and determining the final transmission link for data transmission based on the number of available links L and the optimal path P.

2. The weak network communication method according to claim 1, characterized in that, The forwarding strategy also includes an urgency level U; The processing of the data to be forwarded also includes assigning priority to the data according to the urgency level U, and arranging the sending order based on the priority.

3. The weak network communication method according to claim 1, characterized in that, The predicted link connectivity information includes the link connectivity probability predicted within the next time period T based on a graph neural network model.

4. The weak network communication method according to claim 1, characterized in that, The data to be forwarded includes at least the data type and data size; The system feedback parameters include at least one of node energy, transmission delay, link bit error rate, link congestion level, and transmission success rate, which are used to adjust the subsequently generated forwarding strategy parameters in real time.

5. The weak network communication method according to claim 1 or 2, characterized in that, The number of times the data is repeatedly transmitted, R, is determined by the following formula: ； in, Indicates the basic number of repetitions; This represents the increment of the number of repetitions, determined by the bit error rate. Indicates the increment of the number of repetitions, which is determined by the data type; This indicates a reduction in the number of repetitions determined by the data size. This indicates the rounding up operation; This indicates the maximum number of repetitions.

6. The weak network communication method according to claim 1 or 2, characterized in that, The number of selectable links L is determined by the following formula: ； in, Indicates the total number of available links. This indicates the adjustment amount for the number of selected links, determined by bandwidth utilization requirements. Calculated based on data size, allowed transmission time limit, bandwidth of each link, and link connectivity probability: ； in, This indicates the allowed transmission time limit, which is set according to the data type and urgency level. Indicates the first Link bandwidth; Indicates the first The connectivity probability of the link, i.e. the link of the first link. Link connectivity prediction values ​​for each link. , ; Indicates the size of the data.

7. The weak network communication method according to claim 1 or 2, characterized in that, The number of data fragments S is determined by the following formula: ； in, This indicates the preset maximum number of shards. This represents the base number of shards calculated based on the data size and the shard size. , Indicates the data size. This represents the fragment size, which is related to the link's maximum transmission unit (MTU) and bit error rate, and is the smallest value among all links.

8. The weak network communication method according to claim 1 or 2, characterized in that, The optimal path P is determined in the following way: ； in, This indicates that set C is contained in all available link sets. , This indicates that set C contains L links; Represents the links in set C The utility value is determined by the system feedback parameters, the link status information, and the link connectivity information; i represents the sequence number of link l. .

9. The weak network communication method according to claim 2, characterized in that, The urgency level U is divided into multiple priorities based on data type; and... When there are multiple pieces of data to be forwarded, a corresponding forwarding strategy is generated for each piece of data to be forwarded. According to the forwarding strategy of each data, all data to be forwarded are processed into data blocks respectively; the method also includes: sorting the data blocks corresponding to all data to be forwarded according to their respective urgency, generating a global sending sequence and sending them in sequence, so that data blocks with higher priority are sent before data blocks with lower priority.

10. The weak network communication method according to claim 1 or 2, characterized in that, The method also includes at least one of the following dynamic adjustment mechanisms: a) Real-time dynamic adjustment: When the real-time status of the link is detected to deviate from the status information on which the forwarding strategy was generated, the deviation exceeds a preset threshold, at least one forwarding strategy parameter of the currently executing data transmission task is dynamically adjusted. b) Perform policy calibration after a single transmission task: Based on the comparison between the actual transmission results of this task and the expected goals, update the mapping relationship or benchmark parameters used to generate forwarding policies in order to optimize the policy generation for subsequent tasks. c) Iteratively train the graph neural network model used to predict link connectivity information: continuously update the parameters of the graph neural network model using accumulated historical transmission data and actual link status to improve its prediction accuracy. d) Self-evolution of the decision rule base: Record strategy adjustment actions and their resulting performance effects, summarize effective adjustment rules through data analysis, and update the rule base.

Images