An intelligent routing and maintenance method for underwater Ad Hoc network facing dynamic topology

Abstract

The application provides a kind of underwater Ad Hoc network intelligent routing and maintenance method for dynamic topology, comprising: collecting and normalizing the data forwarding success rate, time delay fluctuation and neighbor survival time and other characteristics of node, adopting exponential decay mechanism to generate timeliness local credit score, and synchronizing lightweight between neighbors through structured credit update package;Based on multi-source weighted fusion and abnormal filtering, the comprehensive credit evaluation of adjacent nodes is realized;Combined with the relative mobility of nodes, the neighborhood topology fracture probability is predicted, and the high fracture risk link is dynamically identified;When detecting the risk link, the upstream node realizes implicit transmission by embedding early warning signal in the data packet header, without additional signaling;After receiving the early warning, the routing layer starts the adaptive switching of backup high-credit, low-fracture-risk path, and the application can efficiently avoid the influence of degraded link, significantly enhancing the network robustness and service continuity.

Description

Technical Field

[0001] This invention relates to the field of intelligent routing and maintenance technology for underwater ad hoc networks, and in particular to an intelligent routing and maintenance method for underwater ad hoc networks with dynamic topology. Background Technology

[0002] In the field of intelligent routing and maintenance technology for Underwater Ad Hoc Networks (UANs), with the rapid development of underwater sensor networks, unmanned underwater vehicle (UUV) swarm communication, and marine engineering applications, network connectivity and routing reliability in dynamic topology environments have become critical issues that urgently need to be addressed. Mainstream technical solutions generally employ distributed routing mechanisms (such as AODV and DSR), combined with link failure detection and path maintenance strategies to cope with underwater environments characterized by frequent node movement, unstable links, and limited bandwidth. Some research has introduced link failure detection algorithms based on node local state information, identifying link failures by tracking neighbor visibility, data forwarding failure rate, or link quality indicators, and triggering a path reconstruction process to restore communication capabilities upon detecting a link failure. Recent industry trends favor lightweight, distributed information processing methods that balance real-time performance with signaling load control, supporting continuous network operation in high-latency and dynamic topology environments. However, existing link failure detection methods generally rely on each node's local or neighbor state awareness, which only provides limited local information. For example, nodes typically determine link failure based on their own periodic neighbor table maintenance, single-hop link QoS parameters (such as signal strength, latency, and packet loss rate), or data forwarding results within a short time window. This approach is prone to false positives and false negatives in scenarios involving high-speed node movement, changing environments, or frequent entry and exit of neighbors from the communication range. On the one hand, when nodes are in local observation blind spots, they cannot promptly obtain global topology changes and link deterioration trends of non-directly connected nodes, making it difficult to provide timely warnings of certain link failure events. On the other hand, false alarms triggered by single-point anomalies or temporary noise disturbances may lead to unnecessary route reconstruction and signal overhead, further affecting network throughput efficiency and stability. Furthermore, current distributed trust management and acoustic channel-based control information broadcasting methods primarily improve accuracy by synchronizing complete trust tables or periodically exchanging global information. However, in underwater environments, due to extremely low communication bandwidth and large propagation delays, these methods incur significant signaling overhead, making efficient operation difficult in practical deployment scenarios. While some existing technologies introduce multi-hop recommendation mechanisms, they lack effective dynamic integration of node historical behavior, making them susceptible to being misled by malicious or faulty nodes that forge recommendations, leading to unstable routing decisions. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an intelligent routing and maintenance method for underwater Ad Hoc networks with dynamic topology.

[0004] The technical solution of this invention is implemented as follows: A method for intelligent routing and maintenance of underwater Ad Hoc networks oriented towards dynamic topology, comprising: S1: Based on the historical interaction data of each node in the underwater Ad Hoc network, obtain the data forwarding success rate, latency fluctuation value and neighbor survival time of each node as input parameters for local credit assessment; S2: Based on the historical interaction data, the various indicators are weighted using an exponential decay mechanism to generate a local credit score vector that reflects recent behavioral dynamics, thereby enhancing the timeliness of the response to changes in node status. S3: Using the local credit scoring vector, construct a lightweight credit update packet, which includes the node ID, credit change amount and timestamp, and send it to the direct neighbor via single-hop broadcast when the link quality deteriorates significantly or when a successful forwarding is completed; S4: Receive credit update packets from multiple neighboring nodes, and perform weighted fusion of the credit changes in each CUP based on the historical credibility of the information source node to generate a comprehensive credit assessment value for the neighboring nodes, preventing misjudgments caused by malicious recommendations. S5: Combining the comprehensive credit assessment value with the relative movement vector of nodes in the neighborhood, calculate the neighborhood breakage probability index of the neighborhood connection stability at the next moment, and form a local topology evolution trend prediction result. S6: Determine whether the comprehensive credit assessment value of any hop node in the active path is lower than the preset threshold and whether the probability of its neighborhood breakage is on the rise. If this condition is met, the link is determined to be a high-risk link for breakage. S7: When a link with a high risk of breakage is determined to exist, the upstream node embeds an implicit link breakage warning signal in the header of the normal service data packet, and uses the existing transmission channel to achieve forward warning transmission without additional signaling overhead. S8: After receiving a data packet containing an implicit link break warning signal, the routing layer initiates the backup path detection process, prioritizing alternative links with higher comprehensive credit assessment values ​​and lower probability of neighboring link breakage, completing path switching and maintaining network connectivity.

[0005] The present invention provides an intelligent routing and maintenance method for underwater Ad Hoc networks with dynamic topology, which has the following beneficial effects: (1) By constructing a local credit scoring vector based on an exponential decay mechanism and combining it with an incremental credit summary propagation strategy, this invention significantly improves the response sensitivity to dynamic changes in node behavior. Compared with the centralized trust model, which requires global information collection and frequent state synchronization, this invention only triggers a single-hop broadcast of a lightweight credit update packet (CUP) when the link quality deteriorates significantly or when a successful forwarding is completed. This greatly reduces the signaling load while ensuring the timeliness of credit assessment. The receiving node adopts an information fusion strategy based on the credibility of the recommendation source, which enhances the ability to resist malicious recommendations and false scores and improves the reliability and security of routing decisions. (2) This invention introduces local topology evolution trend analysis and forward early warning mechanism, realizing a paradigm shift from "passive link recovery" to "active path pre-maintenance", which significantly enhances the network connectivity stability and service continuity in a highly dynamic environment. Through joint analysis of neighbor node credit sequences and relative movement vectors, each node can predict the risk of breakage of neighborhood connections, and embed implicit early warning signals into the header of service data packets by upstream nodes before the actual path is interrupted, realizing early warning transmission without additional overhead; the routing layer starts backup path detection in advance based on this, and prioritizes alternative links with high credit accumulation and stable topology, thereby avoiding the transmission interruption and delay surge caused by rerouting after link failure in traditional methods, which significantly improves the robustness and throughput efficiency of end-to-end communication; (3) This invention designs an event-driven credit synchronization anchor mechanism, which enables rapid convergence of local credit consensus in abnormal scenarios such as large-scale topology reconstruction, effectively alleviating credit misjudgment and decision-making disagreement caused by information propagation delay, and further improving overall consistency and fault tolerance. This mechanism only triggers a short-cycle, low-frequency local credit exchange window when there are significant disturbances such as multiple nodes losing connection in the network, avoiding the resource consumption caused by continuous global synchronization. It is particularly suitable for extreme communication environments such as bandwidth-limited, high propagation delay, and energy-sensitive underwater wireless sensor networks. The entire scheme forms a closed-loop adaptive trust enhancement system of "perception-evaluation-early warning-adjustment-calibration", which has the characteristics of lightweight, high responsiveness and strong robustness, and can be adapted to various dynamic network scenarios without complex parameter tuning. Attached Figure Description

[0006] Figure 1 This is a flowchart of an intelligent routing and maintenance method for underwater Ad Hoc networks with dynamic topology according to the present invention; Figure 2 This is a sub-flowchart of an intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology according to the present invention; Figure 3 This is another sub-flowchart of the present invention, which describes an intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology. Detailed Implementation

[0007] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0008] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0009] like Figure 1 As shown, this invention provides an intelligent routing and maintenance method for underwater Ad Hoc networks with dynamic topology, specifically including: S1: Based on the historical interaction data of each node in the underwater Ad Hoc network, obtain the data forwarding success rate, latency fluctuation value and neighbor survival time of each node as input parameters for local credit assessment; S2: Based on the historical interaction data, the various indicators are weighted using an exponential decay mechanism to generate a local credit score vector that reflects recent behavioral dynamics, thereby enhancing the timeliness of the response to changes in node status. S3: Using the local credit scoring vector, construct a lightweight credit update packet, which includes the node ID, credit change amount and timestamp, and send it to the direct neighbor via single-hop broadcast when the link quality deteriorates significantly or when a successful forwarding is completed; S4: Receive credit update packets from multiple neighboring nodes, and perform weighted fusion of the credit changes in each CUP based on the historical credibility of the information source node to generate a comprehensive credit assessment value for the neighboring nodes, preventing misjudgments caused by malicious recommendations. S5: Combining the comprehensive credit assessment value with the relative movement vector of nodes in the neighborhood, calculate the neighborhood breakage probability index of the neighborhood connection stability at the next moment, and form a local topology evolution trend prediction result. S6: Determine whether the comprehensive credit assessment value of any hop node in the active path is lower than the preset threshold and whether the probability of its neighborhood breakage is on the rise. If this condition is met, the link is determined to be a high-risk link for breakage. S7: When a link with a high risk of breakage is determined to exist, the upstream node embeds an implicit link breakage warning signal in the header of the normal service data packet, and uses the existing transmission channel to achieve forward warning transmission without additional signaling overhead. S8: After receiving a data packet containing an implicit link break warning signal, the routing layer initiates the backup path detection process, prioritizing alternative links with higher comprehensive credit assessment values ​​and lower probability of neighboring link breakage, completing path switching and maintaining network connectivity.

[0010] Step S1: Based on the historical interaction data of each node in the underwater Ad Hoc network, obtain the data forwarding success rate, latency fluctuation value, and neighbor survival time of each node as input parameters for local credit assessment. Specifically, this includes: S1.1: Collect the data packet forwarding records of each node in the underwater Ad Hoc network during the historical communication cycle, extract the number of data packets that were actually successfully forwarded and the total number that should have been forwarded based on the original communication logs, and calculate the data forwarding success rate at the node level to quantify its reliability performance in multi-hop transmission. This study collects packet forwarding records from each node in an underwater Ad Hoc network during historical communication periods. The input is the raw data forwarding log generated by the node communication module, which contains fields such as timestamp, packet identifier, forwarding status code, and target node ID. A log parsing algorithm (parameters: field filtering rules, time window definition) is used to extract all forwarding records that match the current analysis period from the raw log. Furthermore, a packet state matching algorithm (parameters: success flag value, error code mapping table) is used to determine the forwarding result of each forwarding record and obtain the actual number of successfully forwarded packets within the period, outputting an integer success count. ; Furthermore, using an index counting algorithm (parameters: total number of log entries, period definition rules), the total number of data packets to be forwarded within the period is counted and recorded as the total count. ; Furthermore, the success rate calculation formula is adopted:

[0011] To perform a preliminary calculation of the node-level data forwarding success rate, the formula is as follows: This represents the number of data packets that were actually successfully forwarded. This represents the total number of data packets that should be forwarded. Furthermore, an outlier correction algorithm (parameters: consecutive failure threshold, packet loss statistics window) is used to robustly correct the success rate calculation results, remove distorted data caused by individual abnormal periods, and generate the final node-level data forwarding success rate metric. ; Through the above log parsing and calculation processing methods, the original multi-hop transmission records are transformed into index data that quantifies the reliability performance of nodes, thereby providing a forwarding success rate parameter that can be directly input into the local credit assessment model. For example, in a series of underwater Ad Hoc network tests comprising 60-minute communication cycles, node A's communication log recorded a total of There are 10 packet forwarding events, of which the number of events that meet the periodic filtering rules is 1. The number of successful events determined using a state matching algorithm is [number]. The initial success rate was obtained. The result of the calculation is An outlier correction algorithm is applied to eliminate temporary failure counts caused by short-term link jitter. The revised number of successful events is [number]. Recalculate the success rate. To obtain the final indicators for In the input phase of the credit assessment model, this metric significantly improves the measurement accuracy of multi-hop transmission reliability at node A, thereby enhancing the accuracy of subsequent link break prediction. S1.2: Based on the timestamp sequence of data packets received by a node during a continuous communication period, the standard deviation of delay between adjacent arrival intervals is calculated using the sliding window method to generate a delay fluctuation value that reflects the dynamic changes in link quality, serving as a key indicator for measuring communication stability. Based on the timestamp sequence of data packets received by a node during a continuous communication period, a sliding window statistical method (parameters: window length w, step size s) is used to realize the batch extraction and time series construction of the arrival interval of adjacent data packets; Furthermore, through difference operations (parameter: timestamp) and This enables the arrival time interval of adjacent data packets. The calculation yields an interval sequence of length w as the raw input data for the dynamic characteristics of the link. Furthermore, statistical dispersion analysis is performed on the arrival interval sequence within the window using the standard deviation calculation formula to generate time delay fluctuation values. The standard deviation The calculation formula is:

[0012] Where μ is the value within the window. The mean of the samples, where w is the number of samples within the window; Furthermore, the standard deviation calculation is performed on the entire timestamp sequence using a rolling window mechanism to generate a delay fluctuation value sequence covering the entire continuous communication period, in order to characterize the dynamic change process of link quality; By using a normalization method (parameters: maximum value normalization or Z-score), the time delay fluctuation value sequence is mapped to a unified dimension interval [0,1], thereby achieving comparability of fluctuation levels between different links; Through the above sliding window statistics and standard deviation analysis, timestamp data is transformed into a latency fluctuation index that reflects the dynamic changes in link quality, thereby quantifying the key characteristics of link stability and providing important input variables for local credit assessment. For example, in an underwater Ad Hoc network node, the timestamp sequence sampling frequency is 1Hz, the sliding window length w is set to 10, and the step size s is set to 1 second. Differential operations are performed on the timestamps of 50 consecutively received data packets to obtain an arrival interval sequence such as [1.01, 0.99, 1.02, 1.00, 0.98, 1.03, 1.01, 0.99, 1.00, 1.02, ...] seconds. The mean μ = 1.005 seconds is calculated within the first window and substituted into the formula: get ≈0.014 seconds. A set of time delay fluctuation values ​​can be obtained by rolling window calculation, with fluctuation range between 0.012 and 0.020 seconds. After Z-score normalization and mapping to the [0,1] interval, the normalized value of the fluctuation value sequence position corresponding to the maximum value is 1, and the normalized value corresponding to the minimum value is 0. The final output normalized time delay fluctuation sequence is embedded as input feature into the subsequent local credit assessment model, which shows significant sensitivity to the precursors of link failure in the link stability analysis test; S1.3: Analyze the update log of the neighbor table maintained by each node, count the continuous visibility status of the same neighbor node in two adjacent beacon cycles, calculate the neighbor survival time based on the maximum continuous coexistence duration, and use it to evaluate the durability level of the adjacency relationship; S1.4: Normalize the three indicators of data forwarding success rate, latency fluctuation value and neighbor survival time to the same dimension range to form a structured local behavior feature vector, which serves as the initial input condition for the local credit assessment model. S1.5: Cache the local behavioral feature vector to the node's local storage unit and mark it with timestamp information for use by the exponential decay weighting mechanism in the next processing stage, ensuring that the credit assessment process is both time-sensitive and historically dependent.

[0013] Step S2: Based on the historical interaction data, an exponential decay mechanism is used to weight each indicator to generate a local credit score vector reflecting recent behavioral dynamics, thereby enhancing the timeliness of response to changes in node status. Specifically, this includes: S2.1: Based on the three historical interaction parameters of data forwarding success rate, latency fluctuation value and neighbor survival time of each node obtained in S1, normalization processing is performed on the parameters to eliminate the influence of differences in different dimensions on the comprehensive score, and a standardized set of input parameters under a unified dimension is obtained as the input condition of the exponential decay weighted model. S2.2: Based on the normalized standardized input parameters, the historical records of each indicator are weighted using the exponential decay function, and the observations of each indicator are summed in a weighted manner to generate a time-weighted indicator value that highlights the contribution of recent behavior, ensuring that sudden changes in node state can be captured quickly. For a standardized set of input parameters with unified dimensions, an exponentially weighted algorithm based on time decay (parameters: decay coefficient λ, time interval Δt) is used to assign time sensitivity weights to each historical observation. This is achieved by constructing the decay function expression:

[0014] in, As a time-sensitivity weight, For the preset attenuation coefficient, The difference between the current evaluation time and the historical time stamp is used to ensure that newer observation data has a higher weight; Furthermore, by iterating through the historical time series of each indicator, the index weighting function is called to calculate the weighted value one by one, and the values ​​are accumulated to generate the time-series weighted cumulative sum of the indicator. The formula is as follows:

[0015] in, This represents the value of the k-th indicator in the i-th historical record. This is the weighted cumulative sum of the indicator; Furthermore, by performing time-series weighted accumulation and normalization, the bias effect caused by differences in the number of observations is eliminated, as shown in the formula:

[0016] in, The normalized time-weighted index value reflects the concentrated performance of the index's recent contribution. Through the above exponential weighting and normalization processing, the set of input parameters is transformed into a time-series weighted index vector, ensuring that the node credit assessment process is more sensitive to recent state changes and achieving rapid response capability to sudden changes in node state. For example, in a set of historical interaction data of underwater nodes, the data forwarding success rate sequence is [0.93, 0.85, 0.88], the latency fluctuation value sequence is [0.12, 0.20, 0.16] seconds, and the neighbor survival time sequence is [35, 40, 38] seconds. All three sequences have been standardized to a unified dimension range. An attenuation coefficient is set. = The time intervals Δt between the current moment and the three historical records are [5, 10, 15] seconds, respectively. The exponential weights are calculated as follows: The values ​​obtained are [0.6065, 0.3679, 0.2231]. Taking the data forwarding success rate sequence as an example, the time-weighted cumulative sum... =0.6065×0.93+0.3679×0.85+0.2231×0.88=0.5630+0.3127+0.1963=1.072. Normalizing this and dividing by the sum of the weights (0.6065+0.3679+0.2231=1.1975), we get... =0.8949. The latency fluctuation value and neighbor lifetime are calculated using the same process. , This yields a time-weighted index vector [0.8949, 0.1566, 38.3749] (units are standardized intervals). This vector is directly used in the subsequent weighted linear fusion stage to generate the initial local credit score. The results show that the evaluation system's response to the latest data is significantly improved in the event of recent behavioral changes in nodes, supporting the accuracy and real-time performance of the chain break prediction process. S2.3: Based on the obtained time-series weighted index values, a weighted linear fusion strategy is adopted to assign preset fusion weights to the weighted values ​​corresponding to data forwarding success rate, latency fluctuation value and neighbor survival time, respectively, and to perform linear combination calculation to generate an initial local credit score value, so as to realize the mapping and transformation of multi-dimensional behavioral features into a single credit measurement space. S2.4: Apply upper and lower limit threshold constraints to the initial local credit score value, map it to the preset credit score range, generate a normalized local credit score value, prevent outliers or extreme behaviors from causing credit value distortion, and improve the robustness and comparability of the scoring system. The initial local credit score generated by weighted linear fusion An interval threshold constraint method is adopted (parameter: preset lower limit). Preset upper limit This allows the score to be limited to an acceptable range to prevent extreme values ​​from distorting subsequent path decisions. Furthermore, through the linear normalization algorithm (parameter: input) Boundary values and This process maps the score values ​​to the standardized interval [0,1] and obtains intermediate normalized results. This improves the comparability of scores across different nodes; Furthermore, the normalized result is then scaled using a transformation function. Remapped to preset credit score range And generate standardized local credit scores. To ensure that the scoring maintains a consistent dimension under different operating environments; Furthermore, an outlier detection filtering mechanism (parameters: sliding window length w, detection threshold θ) is employed to achieve [the following]. The system continuously monitors periodic changes and eliminates transient jumps caused by equipment failures or link measurement errors to obtain a stable score output. By using interval constraints, normalization, and filtering, the data generated in the previous step is... Transform into This enables robust and standardized processing of local credit scores, supporting subsequent credit differential calculation and dynamic credit evolution. For example, in an underwater Ad Hoc network test environment consisting of 20 nodes, the following settings are configured: =0.2, =0.9, sliding window length w=5, detection threshold θ=0.15. For a certain node, the following values ​​are calculated through previous steps: =1.15, using the interval threshold constraint method, the cutoff value is obtained. =0.9. Using the linear normalization formula:

[0017] in =0.9, the normalized value is calculated. =1.0, and rescale it to the interval [0.2, 0.9] to get =0.9. Subsequently, a sliding window anomaly detection was performed; within the window, the credit score fluctuation did not exceed 0.15, confirming score stability. Final output Used for calculating credit differentials in the next cycle, and significantly improves the stability and reference value of credit information during the distribution process between nodes; S2.5: The normalized local credit score value is calculated by difference with the local credit score status of the previous period to obtain the current credit change. The change is then combined with the node ID and timestamp to form a local credit score vector. This vector is output for the construction and propagation of the subsequent lightweight credit update package (CUP), forming a data foundation that supports dynamic credit evolution.

[0018] like Figure 2As shown, step S3 involves using the local credit scoring vector to construct a lightweight credit update packet (CUP), which includes the node ID, credit change amount, and timestamp. This packet is then sent to the direct neighbor via single-hop broadcast when link quality significantly deteriorates or a successful forwarding is completed. Specifically, this includes: S3.1: Based on the changing trends of three indicators in the local credit scoring vector—data forwarding success rate, latency fluctuation value, and neighbor survival time—calculate the credit change of the current node. Use the differential comparison method to compare the credit scoring vectors of the two evaluation periods one by one to obtain the credit offset of each dimension. Then, generate the comprehensive credit change through weighted summation, which serves as the core incremental information for credit updates. Local credit score vector generated based on previous steps The difference comparison method is adopted (parameters: the period t of the scoring vector and t0). 1) Perform item-by-item comparison of each dimension indicator, extract the single-cycle offset of three indicators: data forwarding success rate, latency fluctuation value, and neighbor survival time, and realize the quantification of indicator change trends; Furthermore, through the weighted bias calculation method (parameter: weight set) The three offsets are then linearly weighted and summed to generate the overall credit change. This ensures that the influence of higher-weighted indicators is amplified in credit volatility; Furthermore, using a sliding trend smoothing algorithm (parameter: window length w=3), the smoothing is applied to... Short-term sequence smoothing is performed to reduce the spikes in credit changes caused by occasional anomalies, thereby obtaining a stable and reliable amount of credit volatility. Furthermore, for the smoothed Perform threshold clipping (parameter: , This limits the value to a preset range to prevent numerical out-of-bounds errors when building the credit update package, thereby improving the robustness of subsequent propagation and decoding processes; Through the above-described differential comparison, weighted summation, smoothing, and threshold pruning processes, the credit score vector from the previous step is transformed into a single-cycle comprehensive credit change. This enables a time-sensitive characterization of node credit evolution, serving as the core incremental information for constructing credit update packages; For example, in a node of an underwater Ad Hoc network, the previous period t The credit score vector of 1 is The credit score vector for the current period t is: The offsets of the three indicators were calculated using the differential comparison method: data forwarding success rate offset = -0.05, latency fluctuation value offset = 0.004, and neighbor survival time offset = -0.02. Weight parameters were then set. =0.5, =0.3, =0.2, apply the weighted summation formula:

[0019] The first term represents the contribution of the weighted data forwarding success rate change, the second term represents the contribution of the weighted latency fluctuation change, and the last term represents the contribution of the weighted neighbor survival time change. The initial... =0.0248, and then a stable value was obtained after smoothing with a sliding window of w=3. =-0.022, after threshold clipping ( =-0.5, After setting the value to 0.5, keep the original value and use it as input for the CPU build. In subsequent neighborhood broadcasts, it can be used to warn of potential link performance degradation, significantly improving the foresight and stability of path maintenance; S3.2: Based on the credit change amount, combined with the current node's unique identifier (Node ID) and the timestamp of the generation time, construct a structured credit update packet (CUP) to ensure information integrity and traceability, and use TLV (Type-Length-Value) encoding to serialize the fields to adapt to the low bandwidth transmission requirements of underwater acoustic channels; Based on the input local credit scoring vector A field extraction mechanism (parameters: index 0, index 2, index 3) is used to extract the unique identifier of a node from the credit score vector. Credit change and generate timestamps ; Furthermore, by using the triplet structured encapsulation method (parameters: fixed field order, tuple length = 3), the three core parameters are combined in an orderly manner to generate basic data units that conform to the CUP format definition, and to ensure the integrity of information fields and the traceability of data location; Furthermore, the TLV (Type-Length-Value) encoding method is adopted (parameters: the Type field indicates the data type, the Length field identifies the byte length, and the Value field stores the parameter value) to realize the serialization processing of triplet data units and obtain a compact binary CUP message structure adapted to the low bandwidth transmission environment of underwater acoustic channels; Furthermore, by using a byte alignment optimization algorithm (parameters: byte boundary = 8 bits, alignment padding strategy = zero padding), the boundary consistency of each field in the binary message is achieved, avoiding cross-byte access overhead, and generating aligned CUP messages suitable for parsing embedded network protocol stacks; Through the above TLV encoding and byte alignment processing, the triplet structure of the previous step is transformed into serialized and low-overhead credit update packet binary data, enabling the underwater Ad Hoc network to stably and traceably propagate credit change information under low bandwidth conditions. For example, on a certain underwater Ad Hoc network node, a local credit score vector [1023, 0.78, -0.05, 1682572800] has been obtained, and a field extraction mechanism is used to obtain... =1023、 =-0.05、 =1682572800. The basic unit {1023, -0.05, 1682572800} is generated through triplet structured encapsulation. The Type field is set to 0x01 (node ​​ID), 0x02 (credit change amount), and 0x03 (timestamp), respectively, and the Length field is 2 bytes, 4 bytes, and 4 bytes respectively. For the encoding of ΔC, the IEEE 754 32-bit floating-point format is used, converting -0.05 into binary representation, and the Value field length is fixed at 4 bytes. During TLV serialization, the Type, Length, and Value sub-segments are sequentially concatenated to form the CUP message, with a total message length of 2+1+2+4+1+4+1+4=19 bytes. A byte alignment optimization algorithm is used to extend the 19-byte message to 24 bytes through zero padding, ensuring that all fields are aligned at 8-bit boundaries. This optimized CUP message significantly reduces transmission latency in underwater acoustic channel environments and greatly improves the credit change synchronization rate between nodes. S3.3: Real-time monitoring of link quality status, performing sliding window statistical analysis based on received signal strength (RSSI) and historical bit error rate (BER) data to determine whether the link has deteriorated significantly; when the link quality degradation is detected to exceed the preset threshold and the duration reaches the judgment window period, the CUP generation process is triggered to ensure that credit warning information is released to the public in a timely manner in the early stage of abnormal link degradation. The real-time monitoring process of link quality status takes the acoustic communication channel between underwater Ad Hoc network nodes as the object, and obtains the current received signal strength RSSI and historical bit error rate BER dataset as the analysis input conditions. Using the sliding window statistical analysis method (parameter: window length) For RSSI, window length For BER, the mean and variance are calculated on the RSSI time series to obtain the link energy attenuation trend index; the mean is calculated on the BER time series and combined with the standard deviation to generate the link error rate fluctuation index, realizing the dynamic quantification of the physical layer transmission characteristics of the link. Furthermore, the aforementioned decay trend indicators and error rate fluctuation indicators are mapped to a unified performance dimension interval [0,1] through normalization processing, and a weighted fusion method is adopted (parameter: weight). , Based on adaptive link type determination, a comprehensive link quality score is generated. Used to determine the current status of the link; Furthermore, using a significant degradation discriminant function ,when Below the preset threshold Output degradation flags at time, where the discriminant function It can be defined as follows: ,in This is the threshold for determining link quality. Furthermore, a persistence detection is performed on the degradation marker sequence of consecutive judgment periods. When the degradation condition is met for P consecutive periods, it is confirmed that the duration of link degradation has reached the set judgment window period. This forms an event flag indicating that the triggering conditions have been met; By triggering the event flag logic, the above link quality degradation detection results are transmitted to the CUP generation module to prepare for the release of credit warning information in the early stage of abnormal link degradation. For example, in an underwater Ad Hoc network with 5 nodes, the nodes communicate using an acoustic modulated signal with a center frequency of 12 kHz, and the window length for sampling RSSI is... Set to 10 seconds, BER statistics window length Set to 20 seconds, weight It is 0.6. Set the threshold to 0.4 for the overall link quality score. The value is 0.35, and the window period is continuously determined. The interval is 30 seconds. During one running cycle, the mean RSSI received by the node continuously decreased to 0.5 times its original value, while the variance increased to 2 times its original value, corresponding to a normalized decay trend index of 0.3; simultaneously, the mean BER increased to 1.5 times its original value, and the standard deviation of fluctuation increased to 1.8 times its original value, corresponding to a normalized error rate fluctuation index of 0.4. A comprehensive quality score is obtained through fusion calculation. = 0.6 × 0.3 + 0.4 × 0.4 = 0.34, according to the discriminant function Output degradation flags. If the above conditions are met for three consecutive judgment cycles, persistent detection is triggered, confirming that the link is in a significantly degraded state, activating the CUP generation process, and the signal warning information is broadcast to the direct neighbor in a single hop when a persistent event is detected for the first time, reducing the risk of false link loss due to information delay, and significantly improving link loss avoidance capability and routing stability; S3.4: After a successful end-to-end data forwarding is completed, the forwarding execution node calls the credit update module to automatically generate a CUP, which includes the positive credit gain caused by this action in the update packet. This positive incentive mechanism enhances the credit propagation frequency of reliable nodes, thereby improving the information visibility of high-quality paths in the network. S3.5: The generated CUP is sent to all directly connected neighbor nodes via single-hop broadcast. Channel access scheduling is performed using a CSMA / CA compatible underwater media access control protocol to avoid collisions. At the same time, the CUP lifetime (TTL=1) is set to limit the propagation range and prevent flooding. This achieves lightweight credit status synchronization only for local neighborhoods and effectively controls network signaling overhead.

[0020] like Figure 3 As shown, step S4 involves receiving credit update packets from multiple neighboring nodes, weighting and fusing the credit changes in each CUP based on the historical credibility of the information source node, and generating a comprehensive credit assessment value for the neighboring nodes to prevent misjudgments caused by malicious recommendations. Specifically, this includes: S4.1: Based on the lightweight credit update packet (CUP) generated in the previous step S3, obtain the CUP data received by the current node from each direct neighbor node. Each CUP contains the sending node ID, credit change amount and timestamp, which serves as the input data source for the weighted fusion process to start the distributed credit information aggregation process. S4.2: Based on the sending node ID carried in the CUP data, query the locally maintained 'historical credibility record table'. This table is dynamically constructed based on the consistency of the node's past forwarding behavior and the stability of the recommendation, and outputs the historical credibility weight coefficient of the corresponding information source node, which is used for subsequent weighted processing of the credit change amount to ensure that the recommendation of high credibility nodes has a higher influence. Based on the CUP data input from step S4.1, a node identifier matching retrieval method (parameter: sending node ID, index structure is a hash mapping table) is used to achieve efficient location of the sending node's identity. Furthermore, by using a credibility table query algorithm (parameter: input node ID, query range is the local historical credibility record table), the historical interaction behavior information of the corresponding node is extracted, and its forwarding success rate consistency index and recommendation stability index are obtained; Furthermore, by utilizing a normalized weighted calculation method for consistency and stability indices (parameters: weight coefficients are set based on long-term network performance statistics), an initial value for the historical credibility of nodes is achieved. The calculation is performed, and structured credibility data records are generated; Furthermore, by employing a time decay correction method (parameters: the decay coefficient λ is based on the network's dynamic change rate, and the time interval Δt is derived from the record table timestamps), the correction is achieved. The timeliness adjustment yields a time-weighted credibility coefficient α, ensuring a rapid response to changes in node status; Furthermore, an abnormal behavior penalty mechanism is adopted (parameter: threshold τ is set based on the standard deviation of credit fluctuation) to penalize nodes that have recently made abnormal recommendations or failed to forward content. The values ​​are weighted down, and the final credibility weight coefficients are generated. ; By using an algorithm based on hash retrieval, normalized weighting, time decay correction, and anomaly weighting, the historical behavior data of nodes retrieved in the previous step is transformed into credibility weight coefficients, thereby giving high-credibility nodes priority in the subsequent weighted calculation of credit change. For example, in a set of underwater Ad Hoc network test scenarios, node A receives a CUP from node B, whose Node ID is B123. The corresponding forwarding success rate consistency in the historical reliability record table is 0.92, and the recommended stability is 0.85. A normalized weighted calculation formula is used:

[0021] The calculation output The value is 0.89. Considering the time difference between the current timestamp and the record table update time Δt = 120 seconds, the attenuation coefficient λ is set to 0.001, and the time-weighted correction formula is applied:

[0022] The calculation output The value is approximately 0.879. Node B was detected to have exhibited abnormal recommendation behavior within the last three update cycles. The abnormality penalty threshold τ is set to 0.05, and the weighting reduction formula is executed:

[0023] Finally, the credibility weight coefficient is obtained. It is approximately 0.829. This... The weighted summation used in subsequent S4.3 can significantly improve the performance of high-reliability nodes. Influence, reduce the interference of malicious recommendations on the overall credit assessment value; S4.3: The historical credibility weighting coefficient is used to calculate the credit change in each CUP by weighting and performing a weighted summation operation to generate the initial fusion credit increment sum as the basic input for comprehensive credit assessment, reflecting the synergistic effect of multi-source credit information; Based on the historical credibility weight coefficients obtained in the preceding step S4.2 Credit changes in multiple received CUP data sets As input conditions, a weighted fusion calculation method is adopted (parameters: The historical credibility weight of the information source node. To correspond to the credit change amount of CUP, linear combination processing of multi-source credit update information is achieved; Furthermore, each credit change is calculated using a weighted product operation. With the corresponding weighting coefficients Multiplying these together forms a weighted credit increment term, which is then summed using the following formula:

[0024] in, Let be the historical credibility weight coefficient of the i-th information source node. This refers to the amount of credit change sent by the information source node; Furthermore, the above weighted product terms are truncated and rounded through a floating-point precision control mechanism to reduce the cumulative data error in the low-bandwidth underwater acoustic channel and ensure the consistency and reproducibility of the subsequent fusion results. Furthermore, using vectorized matrix operations (parameter: weight vector) Increment vector This enables batch parallel computation of all weighted product terms, and the summation result is scaled using a normalization function to match the interval requirements of the preset credit measurement standard. By using a weighted fusion algorithm, the multi-source CUP data from the previous step is transformed into the sum of credit increments after preliminary fusion, thereby achieving a unified quantitative expression of credit suggestions from different neighboring nodes and providing a stable basic input for subsequent overlay of benchmark values ​​and verification of abnormal fluctuations. For example, three neighboring nodes are set in the network, with historical credibility weight coefficients of 0.85, 0.65, and 0.90, respectively, corresponding to the received credit changes. The values ​​are 0.04, -0.02, and 0.05 respectively. Substituting the parameters into the weighted summation formula, we first calculate the weighted product term: Node 1 is... = Node 2 is = Node 3 is = Summing the three results: = By mapping to the [0,1] interval using a normalization function, the sum of the initial credit increments is obtained as follows: In subsequent steps, it is superimposed on the local credit benchmark value, which significantly improves the stability of the comprehensive credit assessment value and its sensitivity to fluctuations. S4.4: The sum of the credit increments after the initial fusion is superimposed with the historical credit benchmark values ​​of the target neighboring nodes stored locally to generate an updated intermediate credit status value. The updated intermediate credit status value is then mapped to the standard credit scoring range [0,1] through a normalization function to obtain an intermediate credit output result under a unified dimension, so as to support cross-node comparison and path selection decision-making. S4.5: Apply an anomaly detection filtering mechanism to the intermediate credit output results, and statistically analyze its changing trend based on sliding window. If the abrupt change exceeds the preset threshold and there is no continuous multi-source consistency support, it is determined to be caused by potential malicious recommendation or transmission disturbance. The weight reduction correction logic is activated, and the final comprehensive credit assessment value is output, which effectively suppresses the spread of false credit and improves the stability and security of the assessment results.

[0025] Step S5: Combining the comprehensive credit assessment value with the relative movement vectors of nodes within the neighborhood, calculate the neighborhood breakage probability index of the neighborhood connection stability at the next time step, forming a local topology evolution trend prediction result. Specifically, this includes: S5.1: Based on the comprehensive credit evaluation value of neighboring nodes generated in the previous steps, obtain the credit dynamic sequence of each direct neighbor node in the most recent communication cycle, and use it as the basic input parameter for neighborhood stability analysis to characterize the temporal evolution characteristics of node behavior reliability. S5.2: Perform differential processing on the relative position information output by the acoustic positioning module of each node in the underwater Ad Hoc network, calculate the displacement increment between adjacent sampling times, and generate the relative movement vector of the nodes in the neighborhood by combining the time interval, so as to quantify the dynamic change rate of the spatial relationship between the nodes. S5.3: The time series of the comprehensive credit assessment value and the relative movement vector are used to form a two-dimensional feature input space. A neighborhood stability joint criterion function is constructed through a weighted linear fusion mechanism. The credit weight coefficient α and the mobility weight coefficient β are adaptively adjusted according to the current network density to balance the influence of behavioral credibility and physical topology dynamics. S5.4: Map the output of the joint neighborhood stability criterion function to the [0,1] interval, and use the Sigmoid nonlinear transformation to generate a neighborhood breakage probability index. This index reflects the degree of probability of local adjacency relationships breaking in the next prediction period, and serves as a quantitative expression of the local topological evolution trend. In the local topological evolution trend prediction step, the output value of the neighborhood stability joint criterion function calculated in the preceding S5.3 is used as the input parameter for this step. This joint criterion function integrates the sequence of comprehensive credit assessment values ​​of adjacent nodes and the relative movement vector information to form a stability assessment quantity with unified dimensions. A normalization mapping method (parameter: target interval [0,1]) is used to linearly compress the output value of the joint criterion function to the standard probability space. This process is achieved by setting an upper limit. and lower limit Perform the interval linear transformation formula:

[0026] Where U is the original output value of the joint criterion. and These are the minimum and maximum values ​​of the criterion within the current observation period, respectively. This is the result of normalization; Furthermore, through a Sigmoid-type nonlinear transformation (parameter: curvature coefficient) This transform achieves a probabilistic mapping of the normalized criterion value, enhancing sensitivity within the range approaching the critical threshold of fracture risk. The transformation uses the formula:

[0027] in, It is a neighborhood fracture probability index. is the curvature adjustment coefficient, and μ is the center position parameter of the Sigmoid function, which controls the shift of the discrimination sensitivity; Furthermore, an adaptive parameter selection algorithm is used (parameters: network node density ρ, frequency of historical disconnection events). ),accomplish and The dynamic adjustment is adapted to the probability mapping characteristics under different topological densities and broken link modes, and to ensure that the discrimination resolution is effectively improved in high-risk areas. Furthermore, the output neighborhood breakage probability index P is buffered and stored in the node's local prediction cache and bound to a timestamp for use as input in the subsequent S5.5 trend analysis step. By combining normalization with a Sigmoid-type nonlinear function, the output of the neighborhood stability joint criterion function is transformed into a neighborhood fracture probability index in the [0,1] interval. This realizes the mapping of multi-source stability assessment quantities into quantifiable and comparable fracture risk probabilities, thereby improving the accuracy and usability of local topological degradation trend prediction. For example, with a node density ρ of 0.35 and a historical link breakage event frequency... In an underwater Ad Hoc network scenario with 4 cycles / hour, the joint criterion function output value is 0.68. It is 0.92. The value is 0.41. After normalization, the result is:

[0028] The result is 0.529. Based on ρ and... Combinations, dynamically set =8.5, =0.5, will Substituting 0.529 into the Sigmoid function:

[0029] The calculated value P≈0.560 indicates that the risk level of neighborhood connection breakage in the next cycle is in the medium-to-high range. This probability maintains an increasing trend across multiple consecutive cycles, providing a reliable signal of topology degradation for subsequent risk link identification modules. S5.5: Perform sliding window trend analysis on the neighborhood breakage probability index calculated for multiple consecutive cycles, extract its first-order difference symbol sequence, and determine whether the index shows a continuous upward trend. If the positive difference ratio within the preset length exceeds the threshold, it is confirmed to enter the topology degradation warning state, and the local topology evolution trend prediction result is output for risk link identification. Based on the time series of neighborhood fracture probability indices generated in step S5.4, a sliding window trend analysis method was used (parameter: window length). Based on the dynamic period setting of the network link, the system can identify the pattern of change in the probability of network disconnection within multiple consecutive periods. Furthermore, the change direction of the fracture probability sequence is quantified by using a first-order difference algorithm (parameter: the difference period Δt is the length of a single evaluation period), and a symbol sequence dataset is generated as the basic input for trend discrimination. Furthermore, by using the symbol sequence statistical method (parameter: statistical period synchronized with sliding window), the number of positive difference symbols in each window is counted, and the proportion of positive difference is obtained to characterize the intensity of the continuous upward trend of the fracture probability index. Furthermore, a threshold comparison algorithm is employed (parameter: threshold). Based on historical topological stability experience values, the positive difference ratio is compared with a preset threshold to obtain a trend determination flag. By using the correlation processing method between trend judgment flags and fracture probability index time series, the results of the previous step are transformed into local topology deterioration early warning status data, thereby realizing the local topology evolution trend prediction results required for risk link identification. For example, in the neighborhood stability analysis of an underwater Ad Hoc network, the time series of breakage probabilities is [0.42, 0.45, 0.47, 0.50, 0.54], and the sliding window length is... Set to 5, and the difference period Δt to 1 period, using the first-order difference formula:

[0030] in, This represents the fracture probability value. This represents the difference result for period t. The difference sign sequence is [+, +, +, +], and the proportion of positive differences is... Preset threshold The threshold is 0.75, and the threshold comparison formula is:

[0031] in, The proportion of positive differences The number of positive signs. This is the window length. In this example... =1, greater than The system determines that the network has entered a topology degradation warning state and outputs a local topology evolution trend prediction result for risk link identification. In this scenario, the trend warning mechanism significantly improves the accuracy of link failure prediction, ensuring that the routing layer completes the preparation of alternative paths before the link actually breaks, thus greatly improving network stability and throughput efficiency.

[0032] Step S6: Determine whether the comprehensive credit assessment value of any hop node in the active path is lower than a preset threshold and whether the probability of its neighboring node breaking is increasing. If this condition is met, the link is determined to be a high-risk link for breaking. Specifically, this includes: S6.1: Based on the comprehensive credit assessment value of adjacent nodes generated in the previous steps, obtain the current credit score of each active path hop point as the basic input parameter for link stability analysis, so as to quantify the reliability of the node's historical behavior; S6.2: Based on the time series data of the comprehensive credit assessment value, the slope of its change is calculated using a sliding window mechanism to generate a credit trend derivative index, which is used to characterize the dynamic evolution direction of the node's credit status. S6.3: Combining the credit trend derivative index with the neighborhood break probability index output by the previous step, execute the dual threshold joint discrimination logic: when the comprehensive credit assessment value is lower than the preset credit threshold and the credit trend derivative is greater than zero (i.e. the credit continues to decline), the node is determined to be in the credit deterioration stage. The comprehensive credit assessment value C(t) based on active path jump points and the credit trend derivative index output by sub-step S6.2 A dual-threshold joint discrimination logic is used to identify the stage of node credit deterioration. The input conditions include the normalized credit score sequence of the most recent communication period, the credit trend derivative, and the neighborhood break probability index from the previous step S5.4. ; Static threshold screening method is used (parameter: preset credit threshold) This allows for the initial screening of credit scores for the current period, outputting a set of candidate nodes with low credit ratings. ; Furthermore, through a symbol detection algorithm (parameter: credit trend derivative) ), to implement the collection Analysis of the direction of credit change in internal nodes, and obtaining the set of nodes with continuously declining credit. ; Furthermore, through the set intersection operation algorithm (parameter: set...) With sets This allows for the joint screening of nodes with low and continuously declining credit, resulting in a set of nodes with deteriorating credit. This ensures that the set contains only nodes that satisfy the double threshold condition; Furthermore, through the state flag appending algorithm (parameter: neighborhood breakage probability) ), is a set Each node records the current neighborhood fracture risk status and generates a credit deterioration stage judgment result table, which serves as the input for S6.4 processing; By using a dual-threshold joint discrimination algorithm, the credit slope index and credit threshold screening results from the previous step are transformed into credit deterioration status judgment data, enabling the identification of high-risk nodes in the early stage and providing an accurate target set for subsequent calculation of the probability change rate of breakage. For example, in an 8-node underwater Ad Hoc network, a preset credit threshold is set. The value is 0.35, and the assessment period Δt is 5 seconds. The credit scores of node N3 for the most recent three periods are 0.42, 0.38, and 0.32, respectively. The credit trend derivative is calculated as follows: when period t=3... The result is A negative value indicates a downward trend; because its current score of 0.32 is below the 0.35 threshold and the trend continues to decline, it is added to the set. Combined with the neighborhood fracture probability = This is marked as a deterioration in credit status. In actual operation, The set includes nodes N3 and N5 in this period. The subsequent calculation of the rate of change of their failure probability in S6.4 shows that both are on the rise, which eventually forms a high failure risk link flag. This effectively avoids routing to unstable node paths and realizes rapid risk link identification under the conditions of marine acoustic channel delay. S6.4: Perform first-order difference processing on the neighborhood breakage probability index to obtain the breakage probability change rate. If the change rate is positive for two consecutive periods, it is confirmed that the neighborhood connection stability is deteriorating, forming evidence of breakage risk at the topological level. S6.5: Based on the logic and calculation results of the credit degradation status and the deterioration trend of the neighborhood break probability, a high break risk link flag is generated and output to the routing control module as the direct basis for triggering the implicit link break warning signal.

[0033] Step S7: When a high-risk link breakage is detected, the upstream node embeds an implicit link breakage warning signal in the header of normal service data packets, utilizing the existing transmission channel to achieve forward warning transmission without additional signaling overhead. Specifically, this includes: S7.1: Based on the link information that has been identified as having a high risk of failure in the active path, obtain the identity of the upstream adjacent node of the link and the normal service data packet to be forwarded, and use it as the operation object for embedding implicit early warning signals. S7.2: Based on the predefined lightweight warning field encoding rules, generate an implicit link break warning signal that occupies only a few bits. This signal uses a binary marking mechanism to represent the 'high risk link' state, so as to minimize the amount of modification to the original data packet header structure. Based on the high-risk link information and normal service data packets to be forwarded obtained by the upstream node in step S7.1, a lightweight warning field encoding method (parameters: warning bit width 3 bits, status code set {000,001}) is adopted to achieve minimal encoding processing for high-risk link states. Furthermore, a structured representation of the high-risk status of the link is achieved through a binary tagging mechanism (parameter: status bit = 001), and the original data stream of the binary early warning field is obtained; Furthermore, a bit mapping table generation algorithm is adopted (parameters: mapping target = protocol header reserved bit index, bit order mode = low bit priority) to achieve precise position binding of binary warning field to protocol header destination field and generate mapped implicit link break warning bit string; Furthermore, redundant bits are removed from the effective information segment of the warning bit string using a bit-width compression method to obtain the compressed warning payload, thereby minimizing the modification range and overhead of the original data packet header structure. Furthermore, a codec consistency verification method (parameter: CRC polynomial width = 8 bits) is adopted to add a check code to the end of the compressed payload, thereby realizing the verifiability verification of the payload during cross-node transmission and generating the final implicit link break warning signal. Through the above encoding and mapping processing, the link risk judgment result of the previous step is transformed into a warning bit string with location binding and compression optimization, so as to achieve the effect of link break risk warning embedded in the header of business data packets without affecting normal parsing. For example, when an underwater Ad Hoc network node performs implicit link failure warning coding, the warning bit width is set to 3 bits, and only 001 is allocated in the status code set to represent a high-risk link status. The node binds the warning field to the 5th to 7th reserved bits of the IPv6 extension header according to the bit mapping table, and writes the value 001 using a low-bit-first sequence pattern. When applying the bit width compression method, the detection bit pattern is 001, and no redundant bits need to be discarded; the compressed payload length remains 3 bits. During the verification phase, a CRC-8 polynomial is used. Encoding is performed to generate an 8-bit checksum, which is appended to the end of the warning payload, forming a complete implicit link break warning signal sequence. This sequence does not cause structural conflicts when subsequently embedded in the header of data packets, and its integrity can be directly verified through the checksum after transmission to the next hop node. This ensures accurate transmission of high-risk link conditions without additional signaling, significantly improving the timeliness and reliability of link break warnings. S7.3: Perform field expansion processing on the protocol header of the normal service data packet, embed the generated implicit link break warning signal into the reserved bit or the unused subfield of the Quality of Service (QoS) field, and form an enhanced data packet with link break warning capability to ensure transmission compatibility and parsing reversibility; S7.4: A data transmission scheduling mechanism based on underwater acoustic communication links, which performs single-hop forwarding of enhanced data packets, enabling implicit early warning signals to propagate naturally to the next hop node along with the service traffic, and realizing forward notification of link breakage risk under the condition of no independent control message. S7.5: Maintain a local early warning cache table at the sending end to record the target path and timestamp of the early warning signal that has been issued, and set a suppression window in combination with the credit update cycle to prevent the same high-risk link from triggering repeated embedding operations, thereby optimizing the spatiotemporal distribution density and signaling efficiency of early warning information.

[0034] Step S8: After receiving a data packet containing an implicit link failure warning signal, the routing layer initiates a backup path detection process, prioritizing alternative links with higher comprehensive credit assessment values ​​and lower neighborhood failure probabilities to complete path switching and maintain network connectivity. Specifically, this includes: S8.1: Based on the service data packets carrying implicit link failure warning signals received by the routing layer, the warning flag bit embedded in the header is parsed, the source node ID that triggered the warning, the high-risk link segment identifier and timestamp information are extracted, and a link failure precursor event record is generated to start the backup path detection process. S8.2: Based on the link failure precursor event record, broadcast a lightweight path probe request packet (PRP) to neighboring nodes, which includes the destination node address, the current path interruption location, and credit constraints for alternative links; the PRP packet adopts a single-hop forwarding mechanism to activate the credit-aware route discovery process within the local topology. S8.3: Receive path response messages from multi-hop neighbors, call the constructed comprehensive credit evaluation value database for each hop node on each candidate path, and combine the neighborhood breakage probability index output by the local topology evolution trend prediction module to calculate the end-to-end stability score of each path and obtain the output result of quantified path reliability. S8.4: Sort the candidate paths based on their end-to-end stability scores and select the path with the highest score as the alternative link; if there are multiple paths with similar scores, further compare their hop count cost and cumulative propagation delay, perform multi-attribute decision optimization, generate the optimal switching path decision table entry, and update the local routing table. S8.5: Issues a path switching instruction to the upstream node of the data flow, directing subsequent business traffic to the newly selected high-reputation, low-breakage-risk link; at the same time, it retains a short-term buffer window of the original path to handle out-of-order packets, and releases the old path resources after confirming stable transmission of the new path for three cycles, completing the seamless switching process to ensure network connectivity and service quality.

[0035] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0036] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and rules of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent routing and maintenance of underwater Ad Hoc networks with dynamic topology, characterized in that, Includes the following steps: S1: Based on the historical interaction data of each node in the underwater Ad Hoc network, obtain the data forwarding success rate, latency fluctuation value and neighbor survival time of each node; S2: Based on the historical interaction data, the various indicators are weighted using an exponential decay mechanism to generate a local credit score vector that reflects recent behavioral dynamics. S3: Using the local credit scoring vector, construct a credit update packet and send it to the direct neighbor via single-hop broadcast when the link quality deteriorates significantly or when a successful forwarding is completed; S4: Receive credit update packets from multiple neighboring nodes, and perform weighted fusion of the credit changes in each CUP based on the historical credibility of the information source node to generate a comprehensive credit assessment value for the neighboring nodes. S5: Combining the comprehensive credit assessment value with the relative movement vector of nodes in the neighborhood, calculate the neighborhood breakage probability index of the neighborhood connection stability at the next moment, and form a local topology evolution trend prediction result. S6: Determine whether the comprehensive credit assessment value of any hop node in the active path is lower than the preset threshold and whether the probability of its neighborhood breakage is on the rise. If this condition is met, the link is determined to be a high-risk link for breakage. S7: When it is determined that there is a high risk of link breakage, the upstream node embeds an implicit link breakage warning signal in the header of the normal service data packet.

2. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology as described in claim 1, characterized in that, Following step S7, the following is also included: S8: After receiving a data packet containing an implicit link break warning signal, the routing layer initiates the backup path detection process, prioritizing alternative links with higher comprehensive credit assessment values ​​and lower probability of neighboring link breakage, completing path switching and maintaining network connectivity.

3. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology as described in claim 1, characterized in that, Step S1 specifically includes: The data packet forwarding records of each node in the underwater Ad Hoc network during the historical communication period are collected. Based on the original communication logs, the number of data packets that were actually successfully forwarded and the total number that should have been forwarded are extracted, and the data forwarding success rate at the node level is calculated. Based on the timestamp sequence of data packets received by a node during a continuous communication period, the standard deviation of delay between adjacent arrival intervals is calculated using the sliding window method to generate a delay fluctuation value that reflects the dynamic changes in link quality. Analyze the update logs of the neighbor table maintained by each node, count the continuous visibility status of the same neighbor node in two adjacent beacon cycles, and calculate the neighbor survival time based on the maximum continuous coexistence duration. The data forwarding success rate, the latency fluctuation value, and the neighbor survival time are normalized to the same dimension range to form a structured local behavior feature vector; The local behavior feature vector is cached in the node's local storage unit and marked with timestamp information for use by the exponential decay weighting mechanism in the next processing stage.

4. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology as described in claim 3, characterized in that, Step S1 further includes collecting data packet forwarding records within historical communication cycles, calculating the forwarding success rate through log parsing, data packet status matching, and index counting, removing distorted data caused by individual abnormal cycles through outlier correction, and generating the final node-level data forwarding success rate index.

5. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology according to claim 1, characterized in that, Step S2 specifically includes: Based on the three historical interaction parameters of each node obtained in step S1, namely, the data forwarding success rate, latency fluctuation value and neighbor survival time, the historical interaction parameters are normalized to obtain a set of standardized input parameters under a unified dimension. Based on the standardized set of input parameters, the historical records of each indicator are weighted using an exponential decay function, and the observations of each indicator are summed in a weighted manner to generate a time-weighted indicator value that highlights the contribution of recent behavior. Based on the time-series weighted index values, a weighted linear fusion strategy is adopted to assign preset fusion weights to the weighted values ​​corresponding to data forwarding success rate, latency fluctuation value and neighbor survival time, respectively, and to perform linear combination calculation to generate an initial local credit score. Apply upper and lower limit threshold constraints to the initial local credit score value, map it to a preset credit score range, and generate a normalized local credit score value. The normalized local credit score value is compared with the local credit score status of the previous period to obtain the current credit change, and the local credit score vector is formed by combining the node ID and the timestamp.

6. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology according to claim 1, characterized in that, Step S3 specifically includes: Based on the changing trends of the three indicators—data forwarding success rate, latency fluctuation value, and neighbor survival time—in the local credit scoring vector generated in step S2, the credit change of the current node is calculated. The differential comparison method is used to compare the credit scoring vectors of the two evaluation periods one by one to obtain the credit offset of each dimension, and the comprehensive credit change is generated by weighted summation. Based on the comprehensive credit change, combined with the current node's unique identifier and the timestamp of the generation time, a structured credit update package is constructed, and the fields are serialized. The link quality status is monitored in real time. Based on the received signal strength and historical bit error rate data, a sliding window statistical analysis is performed to determine whether the link has deteriorated significantly. When the link quality degradation is detected to exceed the preset threshold and the duration reaches the judgment window period, the CUP generation process is triggered. After a successful end-to-end data forwarding is completed, the forwarding execution node calls the credit update module to automatically generate a CUP, which includes the positive credit gain caused by this action in the update package; The generated CUP is sent to all directly connected neighbor nodes via single-hop broadcast, and channel access scheduling is performed using a CSMA / CA compatible underwater media access control protocol.

7. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology as described in claim 6, characterized in that, Step S3 further includes serializing the field using Type-Length-Value encoding, generating a compact binary message through byte alignment optimization, limiting the packet lifetime (TTL=1), broadcasting only in the local neighborhood, and dynamically controlling the signaling overhead of the local credit.

8. The intelligent routing and maintenance method for underwater Ad Hoc networks oriented towards dynamic topology according to claim 1, characterized in that, Step S4 specifically includes: Based on the credit update packet generated in step S3, obtain the CUP data received by the current node from each direct neighbor node; Based on the sending node ID carried in the CUP data, query the locally maintained historical credibility record table and output the historical credibility weight coefficient of the corresponding information source node; The credit changes in each CUP are weighted using the historical credibility weighting coefficients, and a weighted summation operation is performed to generate the initial fusion credit increment sum. The sum of the initial credit increments is superimposed with the historical credit benchmark values ​​of the target neighboring nodes stored locally to generate an updated intermediate credit status value. This value is then mapped to the standard credit scoring range using a normalization function to obtain an intermediate credit output result under a unified scale. An anomaly detection filtering mechanism is applied to the intermediate credit output results. The changing trend is statistically analyzed based on a sliding window. If the abrupt change exceeds a preset threshold and there is no continuous multi-source consistency support, it is determined to be caused by potential malicious recommendation or transmission disturbance. The weight reduction correction logic is then activated, and the final comprehensive credit assessment value is output.

9. A method for intelligent routing and maintenance of underwater Ad Hoc networks oriented towards dynamic topology as described in claim 1, characterized in that, Step S5 specifically includes: Based on the comprehensive credit assessment value generated in step S4, obtain the credit dynamic sequence of each direct neighbor node in the most recent communication period; The relative position information output by the acoustic positioning module of each node in the underwater Ad Hoc network is differentially processed to calculate the displacement increment between adjacent sampling times, and the relative movement vector of the nodes in the neighborhood is generated by combining the time interval. The time series of the comprehensive credit assessment value and the relative movement vector are used to form a two-dimensional feature input space, and a joint criterion function for neighborhood stability is constructed through a weighted linear fusion mechanism. The output of the joint neighborhood stability criterion function is mapped to a unified interval, and a neighborhood fracture probability index is generated by Sigmoid-type nonlinear transformation. A sliding window trend analysis is performed on the neighborhood fracture probability index calculated over multiple consecutive periods. Its first-order difference symbol sequence is extracted, and it is determined whether the index shows a continuous upward trend. If the positive difference ratio within the preset length exceeds the threshold, it is confirmed to enter the topology degradation warning state, and the local topology evolution trend prediction result is output.

10. A method for intelligent routing and maintenance of underwater Ad Hoc networks oriented towards dynamic topology, as described in claim 9, is characterized in that... Step S5 further includes a joint criterion that fuses the time series of neighbor credit assessment values ​​with the relative movement vector between nodes. The credit weight coefficient and the mobility weight coefficient are adaptively adjusted according to the current network density. The neighborhood break probability after Sigmoid transformation is output. When the proportion of positive differences for multiple consecutive periods exceeds a set threshold, it is determined to be a topology deterioration state.

Images