Adaptive threshold dynamic optimization method for VDE signal satellite and related device

Abstract

The invention relates to an adaptive threshold dynamic optimization method for a VDE signal satellite and a related device, and the method comprises the steps: obtaining the real-time state information and historical state information of each dynamic cluster in a satellite coverage area, the real-time state information and historical state information comprising a negotiation success rate, a conflict rate and an idle time slot proportion; based on the real-time state information and the historical state information, using a preset machine learning model to train and learn an operation mode of each dynamic cluster so as to generate a dynamic optimization threshold matched with the current operation state of each dynamic cluster; comparing the real-time state information of each dynamic cluster with the dynamic optimization threshold to determine the resource demand state of each dynamic cluster; and dynamically adjusting the communication time slot allocated to each dynamic cluster in the next period according to the resource demand state. The method has the advantages that the method adapts to busy and complex port or channel area environments, and the communication efficiency can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of satellite communication resource management, and in particular to an adaptive threshold dynamic optimization method and related devices for VDE signal satellites. Background Art

[0002] The Very High Frequency (VHF) Data Exchange System (VDES) is a significant upgrade to the traditional Automatic Identification System (AIS), designed to provide global, real-time, two-way data communications. Using satellite relay, VDES enables ocean-going vessels to maintain reliable data exchange with shore-based vessels even in areas beyond the coverage of shore-based VHF base stations, significantly expanding the range of maritime communications and increasing data transmission rates.

[0003] To manage uplink channel access for a large number of ships in an orderly and efficient manner and avoid communication conflicts, a dynamic access control method has been proposed in the prior art. This method generally involves the satellite dynamically clustering the ships based on their real-time location distribution and density within its coverage area, pre-allocating a portion of communication time slots to each cluster. Within each cluster, ships then independently allocate and use these reserved time slots through a local negotiation mechanism. The negotiation results and access status are periodically reported back to the satellite. Finally, based on the received status reports, the satellite dynamically optimizes the time slot pre-allocation for the next period.

[0004] However, the inventors discovered during their research that the above-mentioned method has certain limitations when performing dynamic optimization. Specifically, its optimization decision-making process relies on comparing real-time status information (such as negotiation success rate, number of conflicts, and number of idle time slots) with a set of pre-set, fixed, static thresholds. This fixed-threshold-based judgment mechanism lacks sufficient adaptability to cope with the highly dynamic and complex ocean communication environment. For example, a conflict rate threshold set for sparsely populated sea areas may be too stringent in busy ports or waterways, causing the system to frequently misjudge congestion and overallocate resources. Conversely, a threshold set for busy areas may be too lenient in sparse areas, failing to promptly identify resource redundancy and resulting in wasted channel resources. This rigid decision-making logic limits the flexibility and accuracy of system resource scheduling, resulting in suboptimal resource allocation results, making it difficult to maximize the overall throughput efficiency of the communication system, and making it impossible to proactively predict and respond to changes in network load based on historical patterns. Summary of the Invention

[0005] In order to adapt to busy and complex port or waterway areas, the present application provides an adaptive threshold dynamic optimization method and related devices for VDE signal satellites.

[0006] In a first aspect, the present application provides an adaptive threshold dynamic optimization method for VDE signal satellites, which adopts the following technical solutions: A method for dynamically optimizing an adaptive threshold for a VDE signal satellite comprises the following steps: S1 obtains real-time status information and historical status information of each dynamic cluster within the satellite coverage area, wherein the real-time status information and historical status information include negotiation success rate, conflict rate and idle time slot ratio; S2. Based on the real-time status information and historical status information, the operation mode of each dynamic cluster is trained and learned using a preset machine learning model to generate a dynamic optimization threshold that matches the current operation state of each dynamic cluster; S3 compares the real-time status information of each dynamic cluster with the dynamic optimization threshold to determine the resource demand status of each dynamic cluster; S4. Dynamically adjust the communication time slots allocated to each dynamic cluster in the next cycle according to the resource demand status.

[0007] By employing this technical solution, the method first continuously collects historical and real-time operational data from each ship cluster. It then uses a machine learning model to deeply analyze this data, learning and understanding each cluster's unique operating patterns. Based on this learning, the model dynamically generates a set of optimized thresholds for each cluster. This intelligent threshold is then used to accurately determine each cluster's true resource needs (shortage, redundancy, or balance). Ultimately, based on this determination, the time slot allocation for the next cycle is precisely adjusted.

[0008] Optionally, step S1 includes the following sub-steps: S11. Select a central ship in each dynamic cluster to aggregate the status information of the cluster; S12. Statistics of the center ship cluster within a communication cycle negotiation success rate, conflict rate and idle time slot ratio to form cycle status data; S13. At the end of the communication cycle, the center ship sends the cycle status data as real-time status information to the satellite via the uplink channel; S14. The satellite terminal receives and stores real-time status information of multiple consecutive communication cycles to form the historical status information for training the machine learning model.

[0009] By adopting the above technical solution, a central ship is first selected in each dynamic cluster to delegate the data aggregation task. The central ship serves as the representative of the cluster and is responsible for subsequent statistics and reporting work, thus avoiding channel congestion that may be caused by a large number of ships reporting directly to the satellite at the same time.

[0010] The central vessel then calculates the negotiation success rate, collision rate, and idle time slot ratio within its cluster over a complete communication cycle, generating standardized cycle status data. This allows the satellite to receive preliminarily processed and summarized information, rather than large amounts of fragmented raw data, simplifying data processing. At the end of the cycle, the central vessel transmits this status data to the satellite as real-time information.

[0011] Ultimately, the satellite accumulates and stores the real-time status information received over multiple consecutive periods, building a rich time series database, or historical status information. This historical database is used to train and learn machine learning models and discover underlying patterns.

[0012] Optionally, step S2 includes the following sub-steps: S21. The machine learning model analyzes the historical status information to identify and learn the periodic operation patterns and traffic trends of each dynamic cluster in different time dimensions; S22. Based on the periodic operating rules and traffic trends, a dynamic performance baseline is established for each dynamic cluster, wherein the dynamic performance baseline is used to characterize the expected normal performance range of the cluster at the current time and geographic location; S23. Combining the dynamic performance baseline with the real-time status information, calculate and output dynamic optimization thresholds for negotiation success rate, conflict rate and idle time slot ratio, respectively, so that the thresholds can adapt to the instantaneous load fluctuation of the cluster.

[0013] By employing the aforementioned technical solution, this method uses a machine learning model to analyze accumulated historical status information to identify and learn the unique periodic operating patterns and traffic trends of each dynamic cluster across different time dimensions. Based on these learned patterns and trends, the method establishes a dynamic performance baseline for each dynamic cluster. This dynamic baseline characterizes the expected normal performance range of the cluster at a specific time and geographic location, providing an intelligent, non-fixed reference standard for subsequent judgments. Finally, the method combines the dynamic performance baseline with real-time status information received from the satellite to calculate and output dynamically optimized thresholds for the negotiation success rate, conflict rate, and idle time slot ratio. Through this combination, the generated thresholds reflect long-term historical trends while quickly responding to immediate load fluctuations within the cluster, thereby achieving precise threshold adaptation.

[0014] Optionally, step S3 includes the following sub-steps: S31. If the negotiation success rate in the real-time status information is lower than the corresponding dynamic optimization threshold, or the conflict rate in the real-time status information is higher than the corresponding dynamic optimization threshold, the resource demand state of the dynamic cluster is determined to be insufficient resources; S32. If the idle time slot ratio in the real-time status information is higher than the corresponding dynamic optimization threshold, the resource demand state of the dynamic cluster is determined to be resource redundancy; S33. If the conditions of S31 and S32 are not met, the resource demand state of the dynamic cluster is determined to be resource balanced.

[0015] By adopting the above technical solution, the method compares the real-time negotiation success rate and conflict rate with the dynamic optimization threshold generated in the previous step. If the negotiation success rate is lower than the threshold, or the conflict rate is higher than the threshold, the resource demand state of the cluster is determined to be insufficient resources. This step is used to diagnose clusters with current communication channel congestion or severe resource contention. At the same time, the method also compares the real-time idle time slot ratio with its corresponding dynamic threshold. If the idle time slot ratio is higher than the threshold, the resource demand state is determined to be resource redundancy, thereby identifying clusters with excessive resource allocation and channel waste. If neither of the above two conditions is met, the state of the cluster is determined to be resource balanced, which indicates that the current time slot allocation can basically match the communication needs of the cluster.

[0016] Optionally, step S4 includes the following sub-steps: S41. When the resource demand state of the dynamic cluster is determined to be insufficient, a portion of the time slots is allocated from the preset global elastic time slot pool to increase the number of communication time slots allocated to the dynamic cluster in the next cycle; S42. When the resource demand state of the dynamic cluster is determined to be resource redundant, the dynamic cluster currently allocated part of the communication time slot is recovered and returned to the global elastic time slot pool to reduce the number of communication time slots allocated to the dynamic cluster in the next cycle; S43. When the resource demand state of the dynamic cluster is determined to be resource balanced, the number of communication time slots allocated to the dynamic cluster in the next cycle is maintained unchanged.

[0017] By employing the above technical solution, differentiated time slot adjustments are performed based on the three resource demand states determined in the previous step. The entire adjustment process relies on a pre-set global elastic time slot pool, which serves as a central resource buffer and scheduling center. When a dynamic cluster is determined to be resource-scarce, the method allocates a portion of time slots from the global elastic time slot pool to increase the number of communication slots allocated to the cluster in the next cycle. This operation is intended to quickly alleviate channel congestion in high-load areas and ensure smooth communication. Conversely, when a cluster is determined to be resource-redundant, the method reclaims some of its currently allocated communication slots and returns them to the global elastic time slot pool. This resource recovery and reuse mechanism effectively improves channel resource utilization across the entire satellite coverage area. Finally, for clusters determined to be resource-balanced, the method maintains the same number of time slots allocated for the next cycle, ensuring stable network operation and avoiding unnecessary resource scheduling overhead.

[0018] Optionally, the process of forming the training set of the machine learning model includes the following sub-steps: S201 extracts the status data of each communication cycle from the historical status information, and associates each status data point with the corresponding cluster identifier, timestamp and geographic location information to form a feature vector; S202. Retrospectively analyze or manually annotate the historical status information, and annotate each feature vector with a corresponding resource allocation evaluation label, wherein the evaluation label represents the degree of optimization of resource allocation at the historical moment corresponding to the feature vector; S203. Combine the feature vector and the resource allocation evaluation label into a data pair to construct a labeled training set for training the machine learning model.

[0019] By adopting the above technical solution, the state data of each historical communication cycle is first extracted from the stored historical state information. Subsequently, these data points are associated with their corresponding contextual information such as cluster identification, timestamp and geographic location, so as to construct a feature vector that can comprehensively describe the network state at a certain moment in history. Then, by retrospectively analyzing the historical data or based on manually input expert knowledge, a resource allocation evaluation label is labeled for each feature vector. This label is used to evaluate whether the actual optimization level of the resource allocation strategy is high or low under the current network state, which is equivalent to assigning a judgment result to the historical data. Finally, the feature vector containing the contextual state is combined with the evaluation label representing its optimization level to form a one-to-one corresponding data pair, that is, a labeled training set for training the machine learning model is constructed.

[0020] In a second aspect, the present application provides a computer device that adopts the following technical solution: A computer device comprising: one or more processors; Memory; One or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs being configured to: The above-mentioned adaptive threshold dynamic optimization method for VDE signal satellite is executed.

[0021] In a third aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium stores a computer program that can be loaded by a processor and execute the above method.

[0022] The storage medium stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement: As described above, the adaptive threshold dynamic optimization method for VDE signal satellites.

[0023] In summary, this application includes at least one of the following beneficial technical effects: 1. By introducing a machine learning model to dynamically generate optimized thresholds, this approach overcomes the lack of adaptability inherent in existing technologies that rely on fixed thresholds. This application learns the unique operating patterns of each cluster based on its historical data and real-time status, generating dynamic thresholds that precisely match the current scenario. This makes resource allocation decisions more targeted and significantly improves the system's dynamic adaptability in complex and changing ocean environments.

[0024] 2. Based on more precise state judgment, this application can more accurately identify resource-deficient clusters earlier and replenish time slots in a timely manner, effectively alleviating channel congestion in high-density areas. It can also keenly detect resource-redundant clusters and reclaim idle time slots, avoiding resource waste. This peak-shaving and valley-filling resource scheduling maximizes the overall system throughput and time slot utilization.

[0025] 3. By learning from periodic patterns and traffic trends in historical data, the machine learning model gives the system a certain degree of predictive capability. This enables proactive resource deployment based on foreseeable load changes, rather than simply responding to problems after they occur. This improves the operational stability and service quality reliability of the entire network. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1A flowchart of a method for dynamically optimizing an adaptive threshold of a VDE signal satellite according to an embodiment of the present invention is shown.

[0027] Figure 2 FIG. 4 is a flow chart illustrating sub-step S1 in an embodiment of the present invention.

[0028] Figure 3 A flowchart illustrating steps for forming a training set for a machine learning model according to an embodiment of the present invention is shown.

[0029] Figure 4 FIG. 4 is a flow chart illustrating the S2 sub-step in an embodiment of the present invention.

[0030] Figure 5 FIG. 4 is a flow chart illustrating the S3 sub-step in an embodiment of the present invention.

[0031] Figure 6 FIG. 4 is a flow chart illustrating the S4 sub-step in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The present application will be further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0033] In the following description, for the purpose of explanation, many specific details are set forth in order to provide a thorough understanding of the inventive concepts. Some of the figures in the drawings of the present disclosure, which are part of this specification, represent structures and devices in block diagram form to avoid making the disclosed principles complicated and obscure. For the sake of clarity, not all features of an actual implementation are necessarily described. In addition, the language used in this disclosure has been selected primarily for readability and instructional purposes and may not have been selected to delineate or limit the subject matter of the invention, thereby resorting to the necessary claims to determine such inventive subject matter. References in this disclosure to "one embodiment" or "an embodiment" mean that the specific features, structures or characteristics described in conjunction with that embodiment are included in at least one embodiment, and multiple references to "one embodiment" or "an embodiment" should not be understood to necessarily all refer to the same embodiment.

[0034] Unless expressly limited, the terms "a", "an" and "the" are not intended to refer to a singular entity, but rather to include a general class of which a specific example may be used for illustration. Thus, the use of the term "a" or "an" may mean any number of at least one, including "one", "one or more", "at least one", and "one or more than one". The term "or" means any of the alternatives and any combination of the alternatives, including all, unless the alternatives are expressly indicated to be mutually exclusive. The phrase "at least one of" when combined with a list of items refers to a single item in the list or any combination of the items in the list. The phrase does not require all of the listed items unless expressly limited to that.

[0035] In a typical VDE satellite communications system, the architecture primarily consists of a satellite serving as a central control node and multiple vessels located within the satellite's coverage area. The satellite is responsible for broad communications coverage and macro-management of network resources, while the vessels, as mobile communication terminals, access the network under the unified scheduling of the satellite to exchange data. The system typically operates in TDMA mode, dividing time into consecutive frames, each of which is further divided into several basic time slots. Vessels request and occupy these time slots for uplink communications.

[0036] In specific operational scenarios, such as the Strait of Malacca, a region with significant vessel traffic, the satellite can perceive the location, distribution, and density of hundreds or even thousands of ships within its signal coverage area in real time. To manage this large-scale concurrent access demand, a leading existing solution employs a two-tiered access control strategy. First, the satellite dynamically divides ships within the strait into several logical clusters based on their location and density, using algorithms such as density clustering. For example, the densely packed ships near Singapore Port are grouped into one cluster, and ships in the middle of the channel are grouped into another. The satellite then pre-allocates a dedicated TDMA time slot for each cluster. Within each cluster, ships independently apply for and occupy specific time slots within the allocated time slots through local channel monitoring and negotiation mechanisms (such as the three-step handshake). Status information, such as negotiation success rate, collision rate, and idle time slot ratio, is periodically aggregated and reported to the satellite by a representative within the cluster (the central ship).

[0037] However, this operational mode presents a significant technical problem. After receiving periodic status reports from each cluster, the satellite needs to dynamically optimize and adjust the time slot allocation for the next cycle. Existing solutions rely on comparing reported performance metrics with a set of preset, fixed, static thresholds. For example, when the collision rate exceeds 10%, time slot allocation is increased, while when the idle time slot ratio exceeds 30%, it is reduced. This fixed threshold-based judgment mechanism is particularly rigid in the highly dynamic and unevenly loaded maritime communication environment. In scenarios like the Strait of Malacca, a 10% collision rate may be normal and acceptable during peak shipping hours; however, the same collision rate at night, when shipping traffic is sparse, may indicate severe channel interference. Using a single set of static thresholds cannot accurately adapt to traffic patterns at different times and locations, nor can it meet communication needs under various emergencies. Ultimately, this results in inaccurate and inflexible resource allocation, reducing spectrum utilization and throughput of the entire communication system.

[0038] Therefore, refer to Figure 1 The embodiment of the present application discloses an adaptive threshold dynamic optimization method for a VDE signal satellite, including the following steps S1-S4.

[0039] S1. Acquire real-time status information and historical status information of each dynamic cluster within the satellite coverage area, wherein the real-time status information and historical status information include negotiation success rate, conflict rate and idle time slot ratio.

[0040] Specifically, in one embodiment, referring to Figure 2 , step S1 includes the following sub-steps S11-S14.

[0041] S11. Select a central ship in each dynamic cluster to aggregate the status information of the cluster.

[0042] S12. The central ship collects statistics on the negotiation success rate, conflict rate and idle time slot ratio of its cluster in a communication cycle to form cycle status data.

[0043] S13. At the end of the communication cycle, the central ship sends the cycle status data as real-time status information to the satellite end via an uplink channel.

[0044] S14. The satellite terminal receives and stores real-time status information of multiple consecutive communication cycles to form the historical status information for training the machine learning model.

[0045] A dynamic cluster does not refer to a fixed geographic area, but rather a dynamically changing logical set calculated by the satellite using a density clustering algorithm based on the real-time location distribution and access density of ships. For example, in a scenario involving the Strait of Malacca, as a fleet navigates, a ship originally belonging to a cluster near Singapore Port may be reassigned to a new cluster in the middle of the waterway after a few hours. The status information of a dynamic cluster refers to a set of key performance indicators that characterize the communication resource usage and congestion level within the cluster during a communication cycle.

[0046] This embodiment selects the negotiation success rate, collision rate, and idle time slot ratio as core status information. The negotiation success rate reflects the proportion of ships within the cluster that can successfully obtain a communication channel and is a direct reflection of the service quality. The collision rate refers to the proportion of negotiation failures caused by multiple ships competing for the same time slot at the same time and is a key indicator of the degree of network channel congestion. The idle time slot ratio indicates how much of the resources pre-allocated to the cluster are unused and is an indicator of the degree of resource waste. These three pieces of information are collected because they comprehensively provide the subsequent machine learning model with the multi-dimensional input data required to judge the health of the network from the three dimensions of service quality, congestion status, and resource utilization.

[0047] Real-time status information collected over multiple consecutive communication cycles is accumulated and stored to form historical status information. Its primary function is to provide the data foundation necessary for training and learning machine learning models. Real-time status information from a single cycle only reflects the momentary situation. However, a time series database composed of a large amount of historical status information can help models discover underlying patterns hidden in the data, such as the cyclical changes in traffic flow on a specific waterway at different times of the day or traffic trends during specific holidays.

[0048] Therefore, steps S11 through S14 together constitute an efficient and reliable data collection and preparation phase. By establishing a central vessel within the cluster, the task of collecting raw data is decentralized to the edge of the network. This significantly reduces the computational and communication burden on the satellite, acting as the central node. The central vessel periodically reports statistically standardized status data, which the satellite then compiles into a historical information database containing temporal context.

[0049] S2. Based on the real-time status information and historical status information, the operation mode of each dynamic cluster is trained and learned using a preset machine learning model to generate a dynamic optimization threshold that matches the current operation state of each dynamic cluster.

[0050] It should be noted that the training and learning of the operating mode referred to here does not mean retraining the model at every communication cycle. Instead, it refers to the use of a fully trained machine learning model to quickly infer and calculate the input real-time data during the real-time operation of the system to output decision results. The model training process itself is based on the historical state information constructed in S14 and is completed periodically in the background or offline state to ensure that the model can grasp various historical operating patterns.

[0051] In its internal processing flow, the pre-set machine learning model first generates a dynamic performance baseline based on its knowledge learned from a vast amount of historical information, combined with contextual information such as the current input cluster ID and time. This baseline represents the model's prediction of the cluster's normal communication status in the current situation. The model then uses the cluster's real-time status information obtained from S1—namely, the actual negotiation success rate, collision rate, and idle time slot ratio—as core input, comparing and calculating this dynamic performance baseline. Through this series of complex nonlinear calculations, the model ultimately outputs a set of dynamically optimized thresholds that closely match the current cluster's real-time load and historical patterns. For example, for a cluster determined to be experiencing peak shipping season based on historical data, the model might output a relatively relaxed collision rate threshold, while for a cluster experiencing sparse traffic, it might output a more stringent threshold.

[0052] The model's training process is typically initiated periodically in the system's background to ensure that its decision-making logic can continuously learn and evolve from the latest network environment. The training process first requires structuring the vast amount of historical state information accumulated in S14 to construct a labeled training set for the model to learn from. This process involves associating each historical data point with contextual information such as the corresponding cluster identifier, timestamp, and geographic location to form a feature vector that comprehensively describes the network state at the time. Furthermore, the system will label each feature vector with a resource allocation evaluation label through retrospective analysis or the introduction of expert rules. This label is a quantitative evaluation of the effectiveness of the resource allocation strategy at the time, for example, an optimization score that integrates multiple indicators such as throughput and communication latency.

[0053] Specifically, in one embodiment, referring to Figure 3 , the process of forming the training set of the machine learning model includes the following sub-steps S201-S203.

[0054] S201. Extracting status data of each communication cycle from the historical status information, and associating each status data point with a corresponding cluster identifier, timestamp, and geographic location information to form a feature vector.

[0055] S202. Retrospectively analyze or manually label the historical status information, and label each feature vector with a corresponding resource allocation evaluation label, where the evaluation label represents the degree of optimization of resource allocation at the historical moment corresponding to the feature vector.

[0056] S203. Combine the feature vector and the resource allocation evaluation label into a data pair to construct a labeled training set for training the machine learning model.

[0057] For example, the system retrieves a record from the historical database. The contextual information for this record includes: cluster identifier Cluster-A7, representing a busy area near the entrance to Singapore Port; timestamp "2025-06-27 09:00 JST," corresponding to the Friday morning shipping peak; and geographical center point 1.25 degrees north latitude, 103.8 degrees east longitude. The corresponding status data at this time is: negotiation success rate 88%, conflict rate 15%, and idle time slot ratio 5%. The system integrates this information into a high-dimensional feature vector. The timestamp can be further processed into more regular features such as Friday and 9:00.

[0058] Next, step S202 is executed, where a human or automated analysis program evaluates and labels the feature vector. The analysis reveals that while the conflict rate of 15% was relatively high at the time, considering it was peak shipping time, the negotiation success rate remained high at 88%. Furthermore, the cluster's total data throughput also reached its peak for the day. Overall, it is concluded that the resource allocation strategy at the time effectively supported peak traffic. Therefore, the feature vector is assigned a higher resource allocation evaluation label, for example, 0.9 on a 0 to 1 scale.

[0059] Finally, in step S203, the system combines the feature vectors and evaluation labels obtained in the previous two steps to form a complete labeled data pair: ([Cluster-A7, 2025-06-27 09:00, 1.25N / 103.8E, 88%, 15%, 5%], 0.9). The system repeats this process, processing a large number of records in the historical database, ultimately constructing a large and high-quality labeled training set.

[0060] After the labeled training set is constructed, it is fed into a pre-set machine learning algorithm for iterative training. During training, the model continuously adjusts its internal parameters, aiming to minimize the error between its predicted evaluation results for any input feature vector and the true evaluation labels in the training set. Training is complete when the model's performance on the validation set reaches the pre-set convergence criterion. Ultimately, this new model, trained and optimized with the latest data, is deployed in the satellite's real-time decision-making system, replacing the old one.

[0061] Specifically, in one embodiment, referring to Figure 5 , step S2 includes the following sub-steps S21-S23.

[0062] S21. The machine learning model analyzes the historical status information to identify and learn the periodic operation rules and traffic trends of each dynamic cluster in different time dimensions.

[0063] S22. Based on the periodic operation rules and traffic trends, establish a dynamic performance baseline for each dynamic cluster, wherein the dynamic performance baseline is used to characterize the expected normal performance range of the cluster at the current time and geographical location.

[0064] S23. Combining the dynamic performance baseline with the real-time status information, calculate and output dynamic optimization thresholds for negotiation success rate, conflict rate and idle time slot ratio, respectively, so that the thresholds can adapt to the instantaneous load fluctuation of the cluster.

[0065] Continuing with the previous example, the pre-trained model has learned specific patterns by studying historical data. For example, the model has learned that for Cluster-A7 near Singapore Port, the peak traffic period is from 8:00 AM to 11:00 AM every Friday, with a collision rate typically between 12% and 15%, and an idle time slot ratio below 5%. This is the model's identification of the operational patterns and traffic trends in this area.

[0066] When executing step S22, assuming the current time is 9:05 AM on Friday, the model will establish a dynamic performance baseline for Cluster-A7 at this specific moment based on the learned patterns. This baseline might be expressed as a set of expected performance values, such as an expected collision rate of 14%, an expected negotiation success rate of 90%, and an expected idle time slot ratio of 4%. This baseline is the model's quantitative prediction of a healthy network state under the current conditions.

[0067] Next, step S23 is executed. The model combines the latest real-time status information just received from the cluster (for example, the actual collision rate is 16% and the actual negotiation success rate is 87%) with the dynamic performance baseline established in S22 and calculates the results. The model analysis finds that the actual collision rate is slightly higher than expected, and the success rate is slightly lower than expected, indicating that the current congestion is even more severe than the typical peak period. Therefore, the model outputs a set of fine-tuned dynamic optimization thresholds for the next minute. For example, the collision rate threshold is set to 18%, the negotiation success rate threshold is set to 85%, and the idle time slot ratio threshold is set to 10%.

[0068] S3. Compare the real-time status information of each dynamic cluster with the dynamic optimization threshold to determine the resource demand status of each dynamic cluster.

[0069] This step is used to apply the dynamic, quantitative threshold data output by the model to the previously acquired real-time status information for judgment, thereby converting complex network performance indicators into a clear and unambiguous qualitative status label.

[0070] Specifically, in one embodiment, referring to Figure 5 , step S3 includes the following sub-steps S31-S33.

[0071] S31. If the negotiation success rate in the real-time status information is lower than the corresponding dynamic optimization threshold, or the conflict rate in the real-time status information is higher than the corresponding dynamic optimization threshold, the resource demand state of the dynamic cluster is determined to be insufficient resources.

[0072] S32. If the idle time slot ratio in the real-time status information is higher than the corresponding dynamic optimization threshold, the resource demand state of the dynamic cluster is determined to be resource redundancy.

[0073] S33. If the conditions of S31 and S32 are not met, the resource demand state of the dynamic cluster is determined to be resource balanced.

[0074] Continuing with the previous example, after the satellite obtains the model-generated dynamic optimization thresholds (collision rate threshold 18%, negotiation success rate threshold 85%, and idle time slot ratio threshold 10%), it begins the judgment process in step S3. First, in step S31, the system compares Cluster-A7's real-time status information (actual collision rate 16%, actual negotiation success rate 87%) with the thresholds. Because the actual negotiation success rate of 87% is no less than 85%, and the actual collision rate of 16% is no more than 18%, the resource shortage condition is not met. Next, in step S32, the actual idle time slot ratio of this cluster is 3%, which is no more than the 10% threshold, and therefore the resource redundancy condition is also not met.

[0075] Ultimately, because neither S31 nor S32 conditions are met, the system proceeds to step S33, finalizing Cluster-A7's resource demand status for the current cycle as balanced. Although the conflict rate is high, it falls within the normal peak fluctuation range predicted by the model based on historical patterns. Therefore, the system does not hastily determine that resources are insufficient, thus avoiding unnecessary resource adjustments.

[0076] S4. Dynamically adjust the communication time slots allocated to each dynamic cluster in the next cycle according to the resource demand status.

[0077] This step introduces a global elastic time slot pool as a central resource buffer, establishing a complete control logic for adding, reducing, and maintaining time slots. This enables the satellite to dynamically shift resources from areas with low load and redundancy to areas with high load and resource shortages, thereby achieving continuous dynamic balance and global optimization of communication resources across the entire coverage area.

[0078] Specifically, in one embodiment, referring to Figure 6 , step S4 includes the following sub-steps S41-S43.

[0079] S41. When the resource demand state of the dynamic cluster is determined to be insufficient, a portion of time slots is allocated from a preset global elastic time slot pool to increase the number of communication time slots allocated to the dynamic cluster in the next cycle.

[0080] S42. When the resource demand state of the dynamic cluster is determined to be resource redundancy, part of the communication time slots currently allocated to the dynamic cluster is reclaimed and returned to the global elastic time slot pool to reduce the number of communication time slots allocated to the dynamic cluster in the next cycle.

[0081] S43. When the resource demand state of the dynamic cluster is determined to be resource balanced, the number of communication time slots allocated to the dynamic cluster in the next cycle is maintained unchanged.

[0082] The execution process is explained using the example above. For Cluster-A7, which is determined to be in a resource-balanced state, the system will execute the logic of step S43, that is, maintain the number of time slots allocated to it in the next communication cycle unchanged to ensure the stability of its network service. At the same time, assume that during the same communication cycle, another cluster, Cluster-B3, located in a sparsely populated area, is determined by S3 to be resource redundant due to its extremely high idle time slot ratio. Then, the system will execute the logic of step S42 on it, for example, reclaim 50 of its originally allocated 200 time slots and return these 50 time slots to the global elastic time slot pool. Assume again that another cluster, Cluster-C5, is determined to be resource-scarce due to sudden communication demands. The system will then execute the logic of step S41 on it, allocating 40 time slots from the global elastic time slot pool to supplement Cluster-C5, thereby increasing its number of time slots.

[0083] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0084] In one embodiment, a computer device is provided, which may be a server. The computer device includes a processor, memory, a network interface, and a database connected via a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data related to a method for adaptive threshold dynamic optimization of a VDE signal satellite. The network interface of the computer device is configured to communicate with an external terminal via a network connection. When executed by the processor, the computer program implements a method for adaptive threshold dynamic optimization of a VDE signal satellite.

[0085] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, the adaptive threshold dynamic optimization method for a VDE signal satellite according to the above embodiment is implemented. To avoid repetition, the method will not be described here in detail.

[0086] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program implements the adaptive threshold dynamic optimization method for VDE signal satellites according to the above embodiment. To avoid repetition, the above description is omitted here.

[0087] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the various embodiments of this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0088] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0089] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for dynamic optimization of adaptive thresholds for VDE signal satellites, characterized in that: The following steps are involved: S1 obtains real-time status information and historical status information of each dynamic cluster within the satellite coverage area, wherein the real-time status information and historical status information include negotiation success rate, conflict rate and idle time slot ratio; S2. Based on the real-time status information and historical status information, the operation mode of each dynamic cluster is trained and learned using a preset machine learning model to generate a dynamic optimization threshold that matches the current operation state of each dynamic cluster; S3 compares the real-time status information of each dynamic cluster with the dynamic optimization threshold to determine the resource demand status of each dynamic cluster; S4. Dynamically adjust the communication time slots allocated to each dynamic cluster in the next cycle according to the resource demand status.

2. The adaptive threshold dynamic optimization method for VDE signal satellite according to claim 1, characterized in that: The step S1 includes the following sub-steps: S11. Select a central ship in each dynamic cluster to aggregate the status information of the cluster; S12. Statistics of the center ship cluster within a communication cycle negotiation success rate, conflict rate and idle time slot ratio to form cycle status data; S13. At the end of the communication cycle, the center ship sends the cycle status data as real-time status information to the satellite via the uplink channel; S14. The satellite terminal receives and stores real-time status information of multiple consecutive communication cycles to form the historical status information for training the machine learning model.

3. The adaptive threshold dynamic optimization method for VDE signal satellite according to claim 1, characterized in that: The step S2 includes the following sub-steps: S21. The machine learning model analyzes the historical status information to identify and learn the periodic operation patterns and traffic trends of each dynamic cluster in different time dimensions; S22. Based on the periodic operating rules and traffic trends, a dynamic performance baseline is established for each dynamic cluster, wherein the dynamic performance baseline is used to characterize the expected normal performance range of the cluster at the current time and geographic location; S23. Combining the dynamic performance baseline with the real-time status information, calculate and output dynamic optimization thresholds for negotiation success rate, conflict rate and idle time slot ratio, respectively, so that the thresholds can adapt to the instantaneous load fluctuation of the cluster.

4. The adaptive threshold dynamic optimization method for VDE signal satellite according to claim 1, characterized in that: The step S3 includes the following sub-steps: S31. If the negotiation success rate in the real-time status information is lower than the corresponding dynamic optimization threshold, or the conflict rate in the real-time status information is higher than the corresponding dynamic optimization threshold, the resource demand state of the dynamic cluster is determined to be insufficient resources; S32. If the idle time slot ratio in the real-time status information is higher than the corresponding dynamic optimization threshold, the resource demand state of the dynamic cluster is determined to be resource redundancy; S33. If the conditions of S31 and S32 are not met, the resource demand state of the dynamic cluster is determined to be resource balanced.

5. The adaptive threshold dynamic optimization method for VDE signal satellite according to claim 4, characterized in that: The step S4 includes the following sub-steps: S41. When the resource demand state of the dynamic cluster is determined to be insufficient, a portion of the time slots is allocated from the preset global elastic time slot pool to increase the number of communication time slots allocated to the dynamic cluster in the next cycle; S42. When the resource demand state of the dynamic cluster is determined to be resource redundant, the dynamic cluster currently allocated part of the communication time slot is recovered and returned to the global elastic time slot pool to reduce the number of communication time slots allocated to the dynamic cluster in the next cycle; S43. When the resource demand state of the dynamic cluster is determined to be resource balanced, the number of communication time slots allocated to the dynamic cluster in the next cycle is maintained unchanged.

6. The adaptive threshold dynamic optimization method for VDE signal satellite according to claim 1, characterized in that: The process of forming the training set of the machine learning model includes the following sub-steps: S201 extracts the status data of each communication cycle from the historical status information, and associates each status data point with the corresponding cluster identifier, timestamp and geographic location information to form a feature vector; S202. Retrospectively analyze or manually annotate the historical status information, and annotate each feature vector with a corresponding resource allocation evaluation label, wherein the evaluation label represents the degree of optimization of resource allocation at the historical moment corresponding to the feature vector; S203. Combine the feature vector and the resource allocation evaluation label into a data pair to construct a labeled training set for training the machine learning model.

7. A computer device, characterized in that: It includes: one or more processors; Memory; One or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to: execute the adaptive threshold dynamic optimization method for a VDE signal satellite according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the adaptive threshold dynamic optimization method for a VDE signal satellite according to any one of claims 1 to 6.

Images