A shipboard network active adaptive switching system and method

Abstract

The present application relates to the technical field of computer network communication, intelligent control and Internet of Things, and particularly relates to a shipborne network active adaptive switching system and method. A shipborne active adaptive network switching method comprises the following steps: during the sailing of a ship, current multi-dimensional context information of the ship and current network performance indicators of different network links are collected in real time according to a preset collection period; current environment influence factors corresponding to different network links are determined based on the current multi-dimensional context information; a history database is configured and updated in real time, and the history data comprises history multi-dimensional context information with a time stamp, history network performance indicators of different network links and history environment influence factors of different network links; the present application performs switching in advance for several minutes based on the prediction of future sailing states and history performance modes, especially the prediction switching for the coverage range of land-based cellular networks; and the present application avoids the command failure or safety hazards caused by network interruption at critical moments.

Description

Technical Field

[0001] This invention relates to the fields of computer network communication, intelligent control and Internet of Things, and in particular to a shipborne network active adaptive switching system and method. Background Technology

[0002] In maritime mobile platforms such as merchant ships, research vessels, and yachts, multiple network connectivity methods are typically required to ensure communication, such as 4G / 5G mobile communication networks in near-shore areas and satellite communication networks in the open ocean. Due to the varying coverage, bandwidth, latency, and cost characteristics of different networks, and the complex and ever-changing environment in which ships navigate, existing network connectivity faces numerous challenges.

[0003] First, network link performance exhibits significant volatility and unpredictability. Link performance is affected by a variety of factors, including the ship's geographical location, course, speed, and other navigational statuses; environmental obstacles such as the shielding effect of the ship's structure and port facilities; changes in network load; and the health of the communication equipment itself. These multidimensional factors lead to frequent fluctuations in link performance that are difficult to predict in advance.

[0004] Secondly, existing network switching solutions are mostly passive, triggering switching only when the current link performance deteriorates severely. This lag results in noticeable network interruptions perceived by users, causing issues such as VoIP call drops, video buffering, and data transmission failures. Especially during underlying physical link switching, the egress IP address of shipboard equipment typically changes, and standard remote network service terminals generally cannot seamlessly handle real-time changes in client IP addresses, leading to interruptions in existing TCP / UDP sessions and requiring users to reconnect or refresh applications.

[0005] Furthermore, existing technologies lack intelligent systems capable of comprehensively considering multiple factors and proactively predicting changes in link performance. On the client side, existing solutions struggle to minimize user-perceived interruptions without relying on modifications to remote network service terminals or their backend servers.

[0006] Therefore, there is an urgent need to propose a shipborne network active adaptive switching system and method, which aims to solve the technical challenge of how to accurately predict the trend of network performance changes during ship docking and navigation, and to carry out proactive switching accordingly. Summary of the Invention

[0007] The purpose of this invention is to provide a shipborne network active adaptive switching system and method. By integrating the ship's real-time navigation status, historical network performance data and intelligent prediction algorithms, it can predict the network link performance in advance and thus perform active switching, minimizing the impact of network interruption on user experience. It has the advantages of actively predicting network performance changes, reducing user-perceived interruptions and improving communication stability.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a shipborne active adaptive network handover method, comprising the following steps: S1. During the ship's journey, collect the ship's current multi-dimensional context information and the current network performance indicators of different network links in real time according to the preset collection cycle; S2. Determine the current environmental impact factors corresponding to different network links based on the current multidimensional context information; S3. Configure and update the historical database in real time. The historical data includes historical multidimensional context information with timestamps, historical network performance indicators of different network links, and historical environmental impact factors of different network links. S4. After predicting the future multidimensional context information of the ship at a preset future time, query the historical database for historical multidimensional context information clusters whose similarity to the future multidimensional context information exceeds a similarity threshold; S5. Calculate the initial predicted values ​​of network performance indicators for different network links at a preset future time based on the historical network performance indicator clusters corresponding to the historical multidimensional context information clusters; determine the future environmental impact factors based on the future multidimensional context information, use the future environmental impact factors as the penalty factors for the initial predicted values, and finally obtain the predicted values ​​of different network links at a preset future time. S6. Determine the real-time comprehensive value of different network links based on the current network performance indicators and current environmental impact factors, and formulate network link switching strategies based on the real-time comprehensive value and the predicted value.

[0009] As one possible implementation, multidimensional contextual information includes at least the ship's latitude, longitude, heading, and speed; different network links specifically include satellite network links and cellular network links; The environmental impact factor corresponding to the cellular network link is determined as follows: based on the distance between the ship's latitude and longitude and the preset geographical reference point, it is determined whether the ship exceeds the coverage area of ​​the cellular network link. If it does, the environmental impact factor is set to a negative number; otherwise, the environmental impact factor is set to a positive number. The environmental impact factor corresponding to the satellite network link is determined as follows: the azimuth of the target signal source relative to the bow is calculated based on the ship's heading, and it is determined whether the azimuth falls within the preset ship structure shielding sector. If so, the environmental impact factor is set to a negative number; otherwise, the environmental impact factor is set to a positive number. Multidimensional context information includes current multidimensional context information and historical multidimensional context information.

[0010] As one possible implementation, in a historical database, a cluster of historical multidimensional context information with a similarity exceeding a similarity threshold with future multidimensional context information is queried. The similarity threshold includes at least: geographical proximity, heading similarity, and / or historical time period matching.

[0011] As one possible implementation, for each network link, based on the queried historical multidimensional context information cluster, a weighted summation method is applied to calculate the initial predicted value of the network performance index at a preset future time. Based on the initial predicted value and the penalty factor, a summation method is used to calculate the predicted value of different network links at a preset future time.

[0012] As one possible implementation, network link switching strategies, from highest to lowest priority, include emergency failover, proactive predictive switching, and regular performance switching.

[0013] As one possible implementation method, the conditions for emergency failover are: the current network link is monitored to have no traffic or abnormal key indicators in multiple consecutive sampling periods, and the real-time comprehensive value of the backup network link is a positive score.

[0014] As one possible implementation method, the conditions for proactively predicting handover are: whether the minimum dwell time has been met since the last handover, and whether the predicted value of the backup network link is consistently better than the predicted value of the current network link during the time period from the current time to a preset future time.

[0015] As one possible implementation, the conditions for regular performance switching are: whether the minimum dwell time has been met since the last switch, and whether the real-time comprehensive value of the backup network link is better than the real-time comprehensive value of the currently active link for multiple consecutive sampling periods.

[0016] Secondly, the present invention provides a shipborne active adaptive network switching system, comprising a sensing module, a prediction module, a decision-making module and an execution module; The perception module is used to determine the environmental impact factors corresponding to different network links based on multidimensional context information. The multidimensional context information includes current multidimensional context information and historical multidimensional context information; the environmental impact factors include current environmental impact factors and historical environmental impact factors. The prediction module includes a historical database, a query unit, and a calculation unit. The historical database is configured with and updated in real-time with historical data, including timestamped historical multidimensional context information, historical network performance indicators for different network links, and historical environmental impact factors for different network links. The query unit retrieves historical multidimensional context information clusters from the historical database whose similarity to future multidimensional context information exceeds a similarity threshold. The calculation unit calculates initial predicted values ​​for network performance indicators of different network links at a preset future time based on the historical network performance indicator clusters corresponding to the historical multidimensional context information clusters. It then determines future environmental impact factors based on the future multidimensional context information, uses these factors as penalty factors for the initial predicted values, and finally obtains the predicted values ​​for different network links at the preset future time. The decision-making module determines the real-time comprehensive value of different network links based on the current network performance indicators and current environmental impact factors, and formulates network link switching strategies based on the real-time comprehensive value and the predicted value. The execution module generates control commands based on the network link switching strategy, and changes the network link routing at the operating system level to achieve network link switching.

[0017] As one possible implementation, the shipborne active adaptive network switching system also includes a management module for loading, saving and providing the system's operating parameters, which include at least the sampling period, network link switching strategy, geographical environment parameters, predicted values ​​and decision thresholds.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention proposes a shipborne active adaptive network handover method, which solves the problems of high lag and obvious user-perceived interruption in existing passive handover schemes by collecting multi-dimensional context information in real time, predicting future network performance, and dynamically adjusting the link handover strategy in combination with environmental impact factors. It transforms passive response into active prediction, reduces the risk of communication interruption caused by sudden deterioration of link performance, and has the advantages of actively predicting network performance changes, reducing user-perceived interruption, and improving communication stability.

[0019] 2. The shipborne active adaptive network switching system proposed in this invention can switch several minutes in advance based on predictions of future navigation status (geographical location, heading, distance from shore) and historical performance patterns, before the cellular network link performance has fully deteriorated. This is crucial for tugboats in complex port environments or long-distance towing, avoiding disruptions to operational continuity and safety due to communication interruptions, especially for predictive switching of land-based cellular network coverage; it avoids command errors or safety hazards caused by network interruptions at critical moments, particularly in tasks requiring high communication stability, such as towing large platforms; it makes switching decisions more forward-looking and reliable, reducing misjudgments caused by instantaneous fluctuations or insufficient prediction; it has high versatility and independence, applicable to various network service providers; it can adaptively learn network characteristics in different regions and environments (e.g., signal blind spots in specific port areas, signal attenuation patterns at different distances from shore, signal fluctuations in specific waterways), and the prediction accuracy will continuously improve with increasing operating time. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart of the shipborne active adaptive network switching method provided in an embodiment of the present invention; Figure 2 This is a structural diagram of the shipborne active adaptive network switching system proposed in an embodiment of the present invention. Detailed Implementation

[0021] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0022] It should be noted that in this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0023] In this invention, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one" or similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, "at least one of a, b, or c" can represent: a, b, c, a combination of a and b, a combination of a and c, a combination of b and c, or a, b, and c, where a, b, and c can be single or multiple.

[0024] Traditional shipboard network handover methods rely on passive handover mechanisms triggered by thresholds of real-time network performance metrics, lacking dynamic perception of multi-dimensional factors in the navigation environment and correlation analysis of historical performance patterns. Because contextual information such as ship location, heading, and speed is not quantified as environmental influencing factors, and a historical database with spatiotemporal correlations is not established, handover decisions are based solely on the current network state, failing to predict future link degradation trends. When ships enter areas at the edge of cellular network coverage or where satellite signals are obstructed, existing methods cannot identify the impact of environmental changes on link performance in advance, leading to delayed handover timing, communication interruptions, and data transmission failures.

[0025] For example, as a ship sails along a predetermined route from nearshore to the open ocean, the signal strength of cellular networks decreases non-linearly with increasing distance from the shore. However, traditional methods only monitor the current signal strength and packet loss rate, without incorporating historical data on performance degradation patterns in the same geographical area for prediction. When the ship's superstructure obstructs the azimuth angle of the satellite antenna due to a change in course, existing methods do not include the course and the obstructed sector in the evaluation system, resulting in the continued use of incorrect routing even when satellite link performance drops sharply. In this scenario, the passive switching mechanism waits for real-time performance indicators to exceed a preset threshold before triggering action. By this time, communication has been interrupted for several minutes, critical operational instructions cannot be transmitted in real time, and video surveillance streams experience stuttering and frame drops.

[0026] If these issues are not addressed, frequent communication interruptions will lead to delays in remote dispatch commands, impacting navigation safety and operational efficiency. In complex waterways or severe weather conditions, delayed handover may cause shipboard equipment to repeatedly attempt connection attempts in dead zones where both cellular and satellite networks are unavailable, exacerbating network resource waste and equipment energy consumption. Decision-making mechanisms lacking historical data support struggle to distinguish between instantaneous fluctuations and continuous degradation, easily triggering frequent link switching and reducing system stability and the continuity of user application sessions.

[0027] Faced with the aforementioned problems, this invention first considers how to dynamically correlate ship navigation status with environmental factors to the network performance evaluation system. Traditional methods rely solely on real-time network indicators, leading to delayed switching. This invention, by introducing multi-dimensional performance prediction technology and combining it with proactive switching strategies, significantly improves the "seamlessness" of the user experience without relying on modifications to remote network service terminals or their backend servers. It can comprehensively consider multiple factors, predict link degradation in advance, and proactively switch before link performance drops sharply, avoiding prolonged and noticeable network interruptions caused by complete link failures. It advances "hard switching" to a more optimal time, thereby minimizing the user's perception of network interruption; it improves the stability and reliability of network connections, shifting from passive response to proactive prediction, reducing the risk of communication interruptions caused by sudden deterioration of link performance, regardless of whether the cause is navigation status, environmental obstruction, network congestion, or other factors; and it makes decision-making more intelligent and forward-looking, with the decision engine integrating real-time performance, Multidimensional environmental context (including geographic location), historical data, and future predictions make switching decisions more forward-looking and reliable, reducing unnecessary switching and misjudgments. This invention has stronger universal applicability; its core innovations are entirely implemented on the client side, without relying on remote network infrastructure or the cooperation or modification of any specific network service provider, making it compatible with and applicable to terminal devices and corresponding backend services of various cellular and satellite network service providers. It possesses data-driven and adaptive learning capabilities; the prediction engine continuously accumulates and analyzes historical performance data, enabling the system to adaptively learn network characteristics in different regions and environments, and the prediction accuracy will continuously improve with increasing runtime. This invention attempts to introduce a multidimensional context information acquisition mechanism, constructing environmental influence factors by acquiring parameters such as geographic location and heading in real time, but it was found that a single real-time environmental judgment still cannot predict future performance changes. In further exploration, this invention analyzes network performance patterns under the same heading and geographic region in historical data, proposing to establish a spatiotemporally correlated historical database to support the prediction model; however, the initial solution did not solve the correlation problem between future context inference and historical pattern matching. To address this, this invention designs a method based on flight status to infer future location and combines similarity thresholds to match historical clusters; however, directly using linear prediction does not consider the dynamic impact of environmental occlusion. Ultimately, this invention integrates historical performance index clusters to calculate initial predicted values, and superimposes future environmental impact factors as penalty terms to form a dual correction mechanism. At the same time, it jointly decides on real-time comprehensive values ​​and predicted values ​​to construct a hierarchical switching strategy to balance the needs of immediate response and forward-looking switching.

[0028] In a first aspect, embodiments of the present invention provide a shipborne active adaptive network handover method, see [link to relevant documentation]. Figure 1 It includes the following steps: S1. During the ship's journey, collect the ship's current multi-dimensional context information and the current network performance indicators of different network links in real time according to the preset collection cycle; S2. Determine the current environmental impact factors corresponding to different network links based on the current multidimensional context information; S3. Configure and update the historical database in real time. The historical data includes historical multidimensional context information with timestamps, historical network performance indicators of different network links, and historical environmental impact factors of different network links. S4. After predicting the future multidimensional context information of the ship at a preset future time, query the historical database for historical multidimensional context information clusters whose similarity to the future multidimensional context information exceeds a similarity threshold; S5. Calculate the initial predicted values ​​of network performance indicators for different network links at a preset future time based on the historical network performance indicator clusters corresponding to the historical multidimensional context information clusters; determine the future environmental impact factors based on the future multidimensional context information, use the future environmental impact factors as the penalty factors for the initial predicted values, and finally obtain the predicted values ​​of different network links at a preset future time. S6. Determine the real-time comprehensive value of different network links based on the current network performance indicators and current environmental impact factors, and formulate network link switching strategies based on the real-time comprehensive value and the predicted value.

[0029] The preset data collection period refers to the interval between the periodic data collection by the system. It can be implemented by using a fixed time interval or by dynamically adjusting the interval, such as 5 seconds, to balance the real-time nature of the data and the consumption of system resources, and to ensure the continuity and timeliness of network status monitoring.

[0030] Among them, multidimensional contextual information refers to the collection of ship navigation status and environmental parameters, specifically including latitude and longitude, heading, and speed, which is used to reflect the real-time operation scenario of the ship, solve the limitations of traditional methods that rely on a single network indicator, and provide a spatial dimension analysis basis for environmental perception.

[0031] Among them, network performance indicators refer to quantitative parameters that reflect the quality of link communication, specifically including traffic, latency, packet loss rate, and signal strength. They are used to objectively assess link availability and provide data support for real-time decision-making.

[0032] Among them, the environmental impact factor refers to the quantitative value of the network link's impact on the environment. Specifically, positive and negative numbers are used to distinguish between available and blocked states. It is used to convert the navigation status into network performance evaluation parameters and solve the indirect impact of physical blockage and dynamic changes in geographical coverage on signal quality.

[0033] Among them, the historical database refers to a structured storage system that stores spatiotemporally related data. Specifically, it adopts an embedded database or file storage to record timestamped context and performance indicators, and supports pattern matching and predictive analysis based on spatiotemporal similarity.

[0034] The similarity threshold refers to the criteria for determining multidimensional context matching, which specifically includes geographic location radius, heading angle difference, and time period matching degree. It is used to filter historical data clusters and ensure the spatiotemporal relevance of the input data of the prediction model.

[0035] The initial forecast value refers to the performance prediction benchmark calculated based on historical data clusters. Specifically, it adopts weighted average or regression analysis methods to reflect historical patterns and provide a basis value for dynamic correction.

[0036] The penalty factor refers to the correction coefficient of environmental impact on the predicted value. Specifically, it is a multiplicative or additive factor used to incorporate future environmental changes into the prediction model and solve the problem of static prediction ignoring real-time environmental changes.

[0037] The real-time comprehensive value refers to the weighted score of current performance and environmental impact. It is calculated using normalized indicators and weight matrices to quantify the immediate availability of the link and ensure basic communication quality.

[0038] The predicted value refers to the comprehensive score of performance and environmental impact at future moments. Specifically, it is calculated by superimposing the initial predicted value and the penalty factor, and is used to proactively assess the link trend and support proactive switching decisions.

[0039] Among them, the network link switching strategy refers to the priority-based decision-making rules, which specifically include emergency fault switching, proactive predictive switching, and routine performance switching. These strategies are used to balance real-time anomaly response and long-term performance optimization, and to reduce communication jitter caused by frequent switching.

[0040] The core innovation of this invention lies in constructing a dynamic network evaluation system that integrates real-time environmental perception and historical data prediction. It generates predicted values ​​through multi-dimensional context matching and a dual correction mechanism, and combines this with a hierarchical decision-making strategy to achieve proactive switching, thus solving the communication interruption problem caused by network fluctuations and passive switching on maritime mobile platforms. The working process and principle are as follows: during navigation, the vessel collects multi-dimensional context information and performance indicators of different network links in real time according to a preset collection cycle. Multi-dimensional context information includes parameters such as the vessel's geographical location, heading, and speed; network performance indicators include bandwidth, latency, and packet loss rate. Based on the current multi-dimensional context information, the system determines the environmental impact factors corresponding to different network links. For example, it determines whether the vessel is outside the cellular network coverage area based on its position, and whether satellite signals are blocked by the ship's structure based on its heading.

[0041] This solution constructs a dynamic network performance evaluation system by integrating real-time environmental perception and historical data prediction. The collection of multi-dimensional contextual information provides a rich data foundation for subsequent analysis. The introduction of environmental impact factors quantifies navigation status as network performance evaluation parameters. The establishment of a historical database supports performance prediction based on spatiotemporal similarity. A dual correction mechanism ensures that predicted values ​​reflect both historical patterns and the impact of environmental changes. Joint decision-making based on real-time comprehensive values ​​and predicted values ​​constructs a switching strategy that considers both current status and future trends. The synergistic effect of these steps ultimately achieves the technical effect of proactively avoiding network degradation in complex navigation environments.

[0042] As one possible implementation, multidimensional contextual information includes at least the ship's latitude, longitude, heading, and speed; different network links specifically include satellite network links and cellular network links; The environmental impact factor corresponding to the cellular network link is determined as follows: based on the distance between the ship's latitude and longitude and the preset geographical reference point, it is determined whether the ship exceeds the coverage area of ​​the cellular network link. If it does, the environmental impact factor is set to a negative number; otherwise, the environmental impact factor is set to a positive number. The environmental impact factor corresponding to the satellite network link is determined as follows: the azimuth of the target signal source relative to the bow is calculated based on the ship's heading, and it is determined whether the azimuth falls within the preset ship structure shielding sector. If so, the environmental impact factor is set to a negative number; otherwise, the environmental impact factor is set to a positive number. Multidimensional context information includes current multidimensional context information and historical multidimensional context information.

[0043] Specifically, for the environmental impact factor judgment logic of cellular network links, the preset geographical reference point can be a set of location coordinates of cellular base stations. By calculating the distance between the ship's current position and the nearest base station, it is determined whether the distance exceeds the preset threshold distance for cellular network coverage. For example, when the distance exceeds 1.5 times the effective coverage radius of the base station, it is determined to be outside the coverage range. For the environmental obstruction judgment of satellite network links, the preset parameters for the ship's structural obstruction sector can be a sector area of ​​±45 degrees in the bow direction. When the azimuth angle of the satellite signal source is within this sector, it is determined to be obstructed by the ship's superstructure. Current multidimensional context information is used for real-time link assessment, and historical multidimensional context information is used to build a predictive model. The time window for historical data can be set to the navigation records of the most recent 30 days.

[0044] Specifically, during ship navigation, real-time collected latitude and longitude data is input into the cellular network link environmental impact factor calculation module, which uses a geofencing algorithm to determine whether the ship exceeds the preset coverage area. When the ship leaves the base station coverage area, the environmental impact factor is set to a negative value, triggering a link availability degradation warning. Simultaneously, the satellite network link environmental obstruction judgment module receives heading data and calculates the signal source azimuth angle based on satellite orbit parameters. If this azimuth angle falls within the ship's obstruction sector, it is set to a negative value to reflect the impact of physical obstruction. Current and historical multi-dimensional context information are stored separately. Historical data is analyzed through time series analysis to extract periodic patterns, used to predict future environmental obstruction status. By combining real-time judgment with historical patterns, misjudgments caused by temporary obstruction are avoided, while coverage failures caused by long-term geographical displacement can be identified. This differentiated environmental impact factor calculation mechanism allows the link performance evaluation of cellular and satellite networks to adapt to their respective physical characteristics, accurately quantifying the impact of environmental factors on communication quality.

[0045] Specifically, the latitude and longitude coordinates of multiple geographic reference points can be pre-set, representing the approximate locations of cellular base stations. As the ship navigates, the distance between the ship's current location and the nearest geographic reference point is calculated in real time. If this distance exceeds the preset cellular network coverage radius, for example, greater than 30 kilometers, the ship is determined to be outside the cellular network coverage area, and the environmental impact factor is set to -1. Conversely, if the distance is less than or equal to 30 kilometers, the ship is determined to be within the cellular network coverage area, and the environmental impact factor is set to +1.

[0046] The environmental impact factor corresponding to the satellite network link is determined as follows: The azimuth angle of the target signal source relative to the bow is calculated based on the ship's heading, and it is determined whether the azimuth angle falls within a preset ship structure shielding sector. If so, the environmental impact factor is set to a negative number; otherwise, it is set to a positive number.

[0047] Furthermore, a pre-defined shielding sector can be set for the ship's structure, such as a 120-degree sector extending aft from the bow. When the ship is sailing, the azimuth angle of the satellite signal source relative to the bow is calculated based on the ship's current course and the known location of the satellite signal source. If this azimuth angle falls within the pre-defined 120-degree shielding sector, the satellite signal is considered to be potentially blocked by the ship's structure, and the environmental impact factor is set to -1. Conversely, if the azimuth angle is not within the shielding sector, the satellite signal is considered not to be blocked by the ship's structure, and the environmental impact factor is set to +1.

[0048] Therefore, multidimensional context information includes current multidimensional context information and historical multidimensional context information. Current multidimensional context information reflects the real-time status of the vessel, while historical multidimensional context information is used to analyze long-term trends and patterns.

[0049] Through the above technical solutions, the embodiments of the present invention can design differentiated environmental impact judgment logic for different types of network links. For cellular networks, by comparing the distance between the ship's position and a preset geographical reference point, it is possible to effectively determine whether the ship has left the base station coverage area, avoiding link misjudgment caused by geographical coverage range. For satellite networks, by analyzing the relationship between the ship's course and the signal source's azimuth, it is possible to accurately identify signal obstruction caused by the ship's structure, avoiding link misjudgment caused by physical obstruction. In addition, by distinguishing between current and historical multi-dimensional contextual information, the system can simultaneously consider real-time status and long-term trends, further improving the accuracy and reliability of environmental impact judgment.

[0050] As one possible implementation, in a historical database, a cluster of historical multidimensional context information with a similarity exceeding a similarity threshold with future multidimensional context information is queried. The similarity threshold includes at least: geographical proximity, heading similarity, and / or historical time period matching.

[0051] Geographic proximity is achieved by setting a geographic radius threshold, such as matching historical data points within a 500-meter radius of the predicted future location to ensure consistency in physical environment characteristics. Heading similarity is determined using an angle difference threshold, such as setting a heading deviation of no more than 30 degrees to avoid missing historical data due to signal source obstruction caused by ship turning. Matching of historical time periods can be based on overlapping time windows or periodic patterns, such as matching historical data 15 minutes before and after a future time, or identifying network load differences between weekdays and holidays.

[0052] Specifically, after the system predicts the ship's position, course, and time information at a preset future time, it filters historical data clusters based on a multi-dimensional similarity threshold. For example, when a ship is about to turn to a specific course, the system not only matches historical data from the same sea area but also filters records with a course deviation of less than 30 degrees, thereby capturing related data where the ship's turn causes satellite signals to be blocked by superstructures. Simultaneously, it combines historical time period matching, such as filtering cellular network load data during the peak hours of 18:00-20:00 daily, to identify the periodic patterns of base station capacity decline. Through three-dimensional joint matching of geographical location, course, and time period, the system can eliminate interference data caused by sudden changes in course or time period misalignment, filtering out historical performance indicators highly correlated with future scenarios. This multi-dimensional matching mechanism synergizes with the network performance prediction model in the preceding claim, ensuring that when calculating the initial prediction value with weights, the selected historical data includes base station coverage characteristics of the same geographical area, signal blocking patterns related to course, and reflects network load fluctuation patterns over time, ultimately improving the reliability of the prediction value. For example, when the predicted future location is 2 kilometers outside a port and the ship's course deviates from the dock's direction, the system accurately predicts the performance degradation of the cellular network due to distance from the base station and peak load, and the performance improvement of the satellite network due to the ship's course adjustment to avoid obstruction, by matching historical data with a radius of 1 kilometer, a course deviation of 20 degrees, and the same evening time period. This triggers an active handover decision. First, based on the ship's current position, course, and speed, the system predicts the ship's approximate position and course after a specified time period (e.g., 5 minutes). Then, it searches the historical database for historical records that match the predicted location within a specified radius (e.g., 0.5 kilometers), with a course difference within a specified angle (e.g., 30 degrees), and within the same time period (e.g., both being weekday daytime). These matching historical records constitute a cluster of historical multidimensional contextual information.

[0053] Furthermore, weights can be assigned to different similarity dimensions. For example, the weight of geographical proximity may be higher than that of heading similarity, while the weight of historical time period matching may be somewhere in between. A comprehensive similarity score is calculated through weighted averages, and when the score exceeds a preset threshold, the historical record is included in the similar information cluster.

[0054] Thus, the system can identify historical data that is highly correlated with the predicted future state, providing a more accurate reference basis for subsequent network performance prediction.

[0055] Through the above technical solutions, the embodiments of the present invention improve the accuracy and relevance of historical data matching. By comprehensively considering multiple dimensions such as geographical location, heading, and time period, the system can more accurately identify historical data that is highly relevant to future contexts. This not only improves the accuracy of network performance prediction but also better captures the impact of heading changes on network performance, as well as the differences in network load patterns at different times. Therefore, the system can more accurately predict potential signal blockage problems or network load peaks, thereby making more reasonable network switching decisions in advance and reducing the risk of communication interruptions or performance degradation.

[0056] As one possible implementation, for each network link, based on the queried historical multidimensional context information cluster, a weighted summation method is applied to calculate the initial predicted value of the network performance index at a preset future time. Based on the initial predicted value and the penalty factor, a summation method is used to calculate the predicted value of different network links at a preset future time.

[0057] The weighted summation is based on the contextual similarity between historical data and the current predicted scenario, including at least one of geographical proximity, heading matching, and temporal correlation. Weights can be calculated using Euclidean distance or cosine similarity algorithms; for example, the geographical proximity weight can be set as a function inversely proportional to the square of the distance. The penalty factor is quantified by future environmental impact factors, including the probability of future environmental occlusion or the magnitude of coverage change. The occlusion probability is calculated using the geometric relationship between the heading and the signal source azimuth, and the magnitude of coverage change is determined by the distance difference between the current and future locations relative to the base station. The summation of the initial predicted value and the penalty factor uses a linear superposition method. For example, when the probability of future environmental occlusion exceeds 50%, the penalty factor is set to a correction of -20% of the initial predicted value.

[0058] Specifically, when predicting network performance indicators at a predetermined future time, the system first selects historical data clusters from the historical database that share similar multidimensional contextual information with the predicted scenario. For each historical data point, a weight is dynamically assigned based on its contextual similarity to the current predicted scenario. For example, if the geographical location of a historical data point is 0.3 kilometers away from the predicted location, its weight coefficient can be set to 0.8; if the heading difference is less than 15 degrees, the weight coefficient increases by 0.2. All historical network performance indicators are weighted and summed to generate an initial predicted value, effectively avoiding the dilution effect of the traditional arithmetic average method on key historical data. Subsequently, based on the environmental impact assessment results at the predetermined future time, the quantified penalty factor is algebraically superimposed on the initial predicted value. For example, if it is predicted that a ship will enter a satellite signal obstruction area in 5 minutes, the penalty factor is calculated as a signal attenuation of -15dB, which is directly applied to the initial predicted value. This phased calculation mechanism retains the basic predictive capability of historical data patterns while responding to future environmental changes through dynamic correction terms, enabling the final predicted value to reflect both historical patterns and future trends.

[0059] For each network link, based on the retrieved historical multidimensional context information clusters, a weighted summation method is applied to calculate the initial predicted values ​​of network performance indicators at a predetermined future time. First, historical records with a similarity exceeding a predetermined threshold to the future multidimensional context information are retrieved from the historical database, forming historical multidimensional context information clusters. The similarity threshold can include geographical proximity, heading similarity, and historical time period matching. For example, geographical proximity can be set to within 0.5 kilometers, heading similarity to within 30 degrees, and time period matching to the same time period (e.g., morning, afternoon, or evening). Next, for each network link, corresponding historical network performance indicators are extracted from the historical multidimensional context information clusters. These performance indicators can include data transmission rate, latency, packet loss rate, etc. Then, a weight is assigned to each historical data point, which can be determined based on its similarity to the future multidimensional context information. For example, historical data with closer geographical proximity, more similar headings, and more matching times can receive higher weights.

[0060] Furthermore, using these weighted historical performance metrics, initial predicted values ​​for network performance metrics at a predetermined future time are calculated through a weighted summation method. The specific calculation formula could be: Initial forecast value = Σ(historical performance indicators) (weight) / Σweight; The summation range is all relevant data points in the historical multidimensional context information cluster.

[0061] After obtaining the initial forecast values, future environmental impact factors are determined based on future multi-dimensional contextual information and used as penalty factors. Environmental impact factors can consider potential future shielding effects, weather conditions, etc. For example, if it is predicted that future course may cause the ship's structure to block satellite signals, a negative penalty factor can be set for the satellite network link.

[0062] Finally, by summing the initial predicted values ​​and the penalty factor, we obtain the final predicted values ​​for different network links at a predetermined future time: Final predicted value = Initial predicted value + Penalty factor; Therefore, by combining this weighted summation with a penalty factor, a more accurate and comprehensive prediction of network performance can be obtained.

[0063] Through the above technical solutions, the embodiments of the present invention can effectively solve the prediction bias problem caused by simple averaging or single-dimensional matching. By introducing a weighted summation mechanism, the system can more accurately reflect the differential contribution of different historical data to the current prediction scenario, improving the accuracy of prediction. Simultaneously, by introducing a penalty factor, the system can consider the dynamic impact of future environmental changes on network performance, further enhancing the accuracy and adaptability of prediction. This method not only improves the accuracy of network performance prediction but also enhances the system's adaptability to complex dynamic scenarios, providing a more reliable basis for subsequent network switching decisions, thereby effectively improving the stability of network communication and user experience.

[0064] As one possible implementation, network link switching strategies, from highest to lowest priority, include emergency failover, proactive predictive switching, and regular performance switching.

[0065] Emergency failover, with the highest priority, is triggered when the current network link detects no traffic or abnormal key indicators for multiple consecutive sampling periods, and the real-time comprehensive value of the backup network link has a positive score. Proactive predictive failover, with the next highest priority, is triggered when the minimum dwell time since the last failover is met, and the predicted value of the backup network link is consistently better than the predicted value of the current network link from the current time to a preset future time. Regular performance failover, with the lowest priority, is triggered when the minimum dwell time since the last failover is met, and the real-time comprehensive value of the backup network link is better than the real-time comprehensive value of the currently active link for multiple consecutive sampling periods. The failover conditions for each priority are logically mutually exclusive through time window constraints and continuous sampling verification mechanisms. For example, the minimum number of sampling periods for emergency failover is set to 2, the continuous sampling period for regular performance failover is set to 3, and the prediction time window for proactive predictive failover is set to 300 seconds, thereby avoiding policy conflicts. The priority stratification mechanism is combined with the prediction value generation method in the preceding claims, wherein the prediction value is obtained based on the similarity matching of historical multidimensional context information clusters and the correction of future environmental impact factors, so that the triggering basis for proactive prediction switching has historical data support and dynamic environmental adaptability.

[0066] Specifically, when network link performance fluctuates during ship navigation, the system first checks whether the emergency failover conditions are met. If the current link has no traffic or key indicators are abnormal for two consecutive sampling periods, and the real-time comprehensive value of the backup link is positive, a switchover is immediately triggered to ensure uninterrupted communication. If the emergency switchover threshold is not reached, the system further determines the proactive prediction switchover conditions: by querying historical performance data in the historical database that are similar to the context at a future preset time, the system calculates whether the predicted value of the backup link will be continuously better than the current link for the next 300 seconds, while also meeting the minimum dwell time constraint of 120 seconds, thus completing a smooth switchover before the link performance actually deteriorates. If none of the above conditions are met, a regular performance switchover verification is performed, requiring the real-time comprehensive value of the backup link to be better than the current link for three consecutive sampling periods and to meet the dwell time requirement, preventing misjudgments caused by instantaneous fluctuations. The execution results of each priority switchover strategy are fed back to the historical database to optimize the calculation accuracy of subsequent prediction values. For example, when proactive prediction switchover successfully avoids performance degradation, the relevant context features and prediction parameters will be marked as effective patterns, improving the prediction accuracy for similar future scenarios. This hierarchical decision-making mechanism, together with multi-dimensional context awareness and historical data-driven prediction methods, forms a closed-loop optimization, ensuring timely response to emergency faults while balancing the accuracy of forward-looking switching and the stability of routine switching.

[0067] Emergency failover has the highest priority. When the current network link is found to have no traffic or abnormal key indicators for multiple consecutive sampling periods, if the real-time composite value of the backup network link is positive, a switchover to the backup link is triggered immediately. For example, if a satellite communication link experiences signal interruption due to severe weather, resulting in no traffic for two consecutive sampling periods, and the cellular network link has a positive score at this time, a switchover to the cellular network link is triggered immediately.

[0068] Proactive predictive handover has the second highest priority. The system triggers a handover to the backup link when it predicts that the backup link will consistently outperform the current link for a period of time and meets the minimum dwell time requirement. For example, the prediction engine might analyze historical data and determine that a ship is about to leave cellular network coverage, predicting a significant drop in cellular network link performance in 5 minutes. If the satellite link prediction score consistently exceeds the cellular link prediction score, and the minimum dwell time has been met, a handover to the satellite link is triggered.

[0069] Regular performance handover has the lowest priority. A handover to the backup link is triggered when the real-time aggregated value of the backup link is better than the current link for multiple consecutive sampling periods and meets the minimum dwell time requirement. For example, if the real-time aggregated value of the cellular network link is higher than that of the satellite link for three consecutive sampling periods, and more than 120 seconds have passed since the last handover, a handover to the cellular network link is triggered.

[0070] Through the above technical solutions, this invention implements a hierarchical priority mechanism for network handover strategies. Emergency failover ensures rapid response in the event of complete link failure, avoiding communication interruptions. Proactive predictive handover executes handovers in advance based on future performance predictions, reducing the lag of passive handovers. Regular performance handover optimizes network performance in a stable state. This hierarchical strategy effectively coordinates the relationship between emergency response, forward optimization, and steady-state maintenance, improving the reliability of network handovers and the continuity of user experience. Simultaneously, by introducing constraints on minimum dwell time and the number of consecutive sampling periods, frequent handovers caused by instantaneous performance fluctuations are avoided.

[0071] As one possible implementation method, the conditions for emergency failover are: the current network link is monitored to have no traffic or abnormal key indicators in multiple consecutive sampling periods, and the real-time comprehensive value of the backup network link is a positive score.

[0072] The monitoring mechanism involving multiple consecutive sampling periods requires the fault state to persist for at least two sampling periods, with the sampling period set to 5 to 30 seconds. For example, if key indicators are detected as abnormal within three consecutive sampling periods, this means the fault condition is considered to be present. Abnormal key indicators include, but are not limited to, packet loss rate exceeding a preset threshold, latency exceeding an acceptable range, or complete traffic interruption. The packet loss rate threshold can be set to above 90%, and the latency threshold can be set to above 2000 milliseconds. The real-time comprehensive value of the backup link is obtained by weighted calculation of real-time performance indicators and environmental impact factors. For example, when the backup link is outside the cellular network coverage area or is obstructed by the ship's structure, its environmental impact factor is set to a negative number, causing the real-time comprehensive value to fail to meet the positive score requirement.

[0073] Specifically, when the primary link detects no traffic or a packet loss rate exceeding 95% for three consecutive sampling periods, the system determines that the emergency failover trigger condition is met. At this time, the real-time comprehensive value of the backup link needs to be calculated using a weighted average of real-time performance indicators and environmental impact factors. For example, a packet loss rate below 5%, latency below 500 milliseconds, and no environmental obstruction are required for a positive comprehensive score. By simultaneously verifying the continuous anomaly of the primary link and the availability status of the backup link, erroneous operations triggered by instantaneous signal fluctuations are avoided. For instance, a scenario where the primary link experiences a sudden increase in packet loss rate within a single sampling period but subsequently recovers will not trigger a switchover. When the backup link experiences environmental obstruction resulting in a negative environmental impact factor, even if its performance indicators are normal, the real-time comprehensive value is still judged as unavailable, thus preventing a switchover to the unavailable link. This mechanism works in conjunction with the calculation logic of the environmental impact factor in the pre-processing technology, quantifying the environmental obstruction status into a comparable numerical indicator through the real-time comprehensive value. This ensures that the switchover decision considers both link performance and environmental availability, effectively reducing the risk of communication service interruption.

[0074] Emergency failover conditions include two aspects: First, the current network link detects no traffic or abnormal key indicators for multiple consecutive sampling periods. The sampling period can be set to 5 seconds, and multiple consecutive sampling periods can be set to 2. No traffic means no data packet transmission is detected within one sampling period. Abnormal key indicators include, but are not limited to: round-trip time (RTT) exceeding 1000 milliseconds, packet loss rate exceeding 50%, or signal strength below -100 dBm. Second, the real-time composite value of the backup network link has a positive score. The real-time composite value is calculated by weighting multiple performance indicators; for example, composite value = 0.3. Normalized flow rate +0.3 Normalized RTT + 0.2 Normalized packet loss rate +0.2 Normalized signal strength. A positive score indicates that the overall value is greater than a preset threshold, such as 0.6.

[0075] When both conditions are met, the system will trigger an emergency failover. The failover process includes lowering the priority of the current network link and raising the priority of the backup network link. Priority adjustment is achieved by modifying the network interface configuration of the operating system. After the failover is complete, the system will record the failover event, including the failover time, the triggering reason, and the network status before and after the failover.

[0076] Through the above technical solutions, the embodiments of the present invention can effectively avoid erroneous handovers caused by momentary interference or the unavailability of backup links, thereby improving the reliability and stability of network handover. By requiring the current link to experience anomalies over multiple consecutive periods, unnecessary handovers due to brief fluctuations are avoided. Simultaneously, by verifying the availability of backup links, it ensures immediate restoration of communication after handover, reducing the risk of network service interruption. This mechanism is particularly suitable for scenarios requiring high-reliability communication, such as telemedicine, financial transactions, or industrial control systems, minimizing misoperations and unnecessary handovers while ensuring communication continuity.

[0077] As one possible implementation method, the conditions for proactively predicting handover are: whether the minimum dwell time has been met since the last handover, and whether the predicted value of the backup network link is consistently better than the predicted value of the current network link during the time period from the current time to a preset future time.

[0078] The minimum dwell time is configured to force the system to maintain the current link for a fixed duration after handover. This duration can be dynamically adjusted based on historical handover frequency and link stability data, for example, between 120 and 300 seconds. The sustained superiority condition requires that the predicted performance of the backup link at each sampling point is higher than that of the current link within a preset future time period. This time period can be set to 60 to 600 seconds, for example, 300 seconds. The sustained superiority is verified by comparing the values ​​of all predicted points within the future time period. A handover operation is triggered only if the backup link score at all predicted points is higher than that of the current link. The execution order of the above two conditions is set to a parallel verification mode; a handover command is only allowed to be generated when both conditions are simultaneously met.

[0079] Specifically, the minimum dwell time is implemented through the operating system kernel's timer module. This module records the timestamp of the most recent handover operation and calculates the time difference between the current time and the last handover within each decision cycle. When the time difference exceeds a preset threshold, the dwell condition is marked as satisfied. The prediction engine stores historical network link performance data with timestamps and multi-dimensional contextual information (including but not limited to geographical location, heading, and speed). Based on the ship's current multi-dimensional contextual information (such as current navigation status) and combined with performance patterns of similar multi-dimensional contextual regions in the historical database, the prediction engine predicts the performance trend of each link over a future period (e.g., the next few minutes). This prediction comprehensively considers the combined impact of multiple factors such as geographical location changes, environmental shading, and historical load on performance. The continued superiority of the predicted values ​​is verified through the future time period performance curve provided by the prediction engine, which contains predicted values ​​from multiple equally spaced sampling points. The decision engine combines the real-time performance of each link, the real-time environmental awareness status, and the future performance predictions provided by the prediction engine to calculate a comprehensive score for each link. Based on preset thresholds and strategies (including emergency failover, proactive predictive failover, and routine performance failover), it determines whether and when to perform a network link switchover. The decision engine compares the prediction curves of the backup link and the current link point by point. When it detects that the predicted values ​​of all sampling points of the backup link are higher than those of the current link, it generates a signal indicating that the superiority verification has passed. The synergistic effect of the two conditions is reflected in the fact that the time constraint mechanism suppresses erroneous switches caused by instantaneous prediction fluctuations, while continuous verification ensures that the switchover decision is based on long-term performance trends. For example, in a scenario where a ship is moving towards the edge of cellular network coverage, if the prediction engine calculates that the predicted value of the satellite link will be consistently better than that of the cellular link for the next 300 seconds, and the system has maintained the current link for more than 180 seconds, triggering a proactive switchover at this time can avoid communication interruptions caused by subsequent cellular signal attenuation, while also preventing invalid switches caused by short-term signal fluctuations.

[0080] The prediction engine stores historical data and predicts future link performance based on current multidimensional context information and historical patterns. The historical database utilizes a lightweight, non-independent embedded data storage solution (e.g., a file-based database) to permanently store timestamped multidimensional context information of vessels (position, heading, speed) as well as real-time performance metrics and environmental awareness status for each link. Data recording occurs when the data acquisition module acquires new data, which the prediction engine then writes to the historical database.

[0081] The first step in the prediction logic process is future context inference: based on the ship's current heading and speed, a preset physical model (e.g., uniform linear motion model) is used to infer the ship's approximate position and heading at the end of a specified time period in the future (e.g., the next 300 seconds), as well as other future context information.

[0082] The second step in the prediction logic process is similarity pattern query: the prediction engine searches the historical database for historical records that are similar to the predicted future context. This similarity matching can be based on multiple dimensions, such as: geographical proximity (within a specified radius), heading similarity (within a specified angular difference), and matching degree of historical time period (such as day of the week, whether it is nighttime), etc.

[0083] The third step in the prediction logic process is the calculation of prediction metrics: for the similar historical data retrieved, a weighted average is calculated as the prediction value for the performance metrics of each network link (such as average packet volume, average RTT, average packet loss rate, and average signal strength). The weights can be adjusted according to the similarity between the historical data and the current context (for example, historical data that is geographically closer, has a better heading, or is more recent in time is given higher weights).

[0084] The fourth step in the prediction logic process is to predict the environmental impact: using the calculated future context information, the logic of the environmental perception module is called again to predict whether the link may be affected by environmental occlusion in the future.

[0085] The fifth step in the prediction logic process is prediction score generation: using the same link scoring algorithm as the decision engine, but with the input parameters being the predicted performance indicators and the predicted environmental impact status, thereby calculating the prediction score for each link in the future time period.

[0086] The conditions for proactive predictive handover include two aspects. First, the system checks whether the minimum dwell time has been met since the last handover. Specifically, the system records the timestamp of each handover and calculates the difference between the current time and the last handover time. If this difference exceeds a preset minimum dwell time threshold, such as 120 seconds, the first condition is met. Second, the system compares the predicted performance of the backup network link and the current network link over a preset future time period. Specifically, the system uses a prediction model to calculate the predicted performance values ​​of the two links every 5 seconds for the next 300 seconds starting from the current moment, forming two sets of sequences each containing 60 predicted values. Then, the system compares these 60 pairs of predicted values ​​one by one. If the predicted value of the backup link is consistently higher than the predicted value of the current link throughout the entire sequence, the second condition is met. When both conditions are met simultaneously, the system triggers proactive predictive handover, switching the network connection from the current link to the backup link.

[0087] Through the above technical solutions, the embodiments of the present invention effectively solve the reliability and stability problems of predictive handover. Introducing a minimum dwell time avoids frequent handovers and improves system stability. It requires that the backup link consistently outperform the current link over a preset future time period, ensuring that handover decisions are based on long-term performance trends rather than instantaneous fluctuations, thus reducing the probability of invalid handovers. This proactive handover strategy based on multi-dimensional prediction can proactively address link degradation and ensure the robustness of handover decisions through time constraints and continuous verification, thereby achieving seamless optimization of the user experience.

[0088] As one possible implementation, the conditions for regular performance switching are: whether the minimum dwell time has been met since the last switch, and whether the real-time comprehensive value of the backup network link is better than the real-time comprehensive value of the currently active link for multiple consecutive sampling periods.

[0089] The minimum dwell time can be set from 60 to 300 seconds, for example, to 120 seconds. This time parameter is dynamically adjusted through the system configuration file to ensure that the link maintains a minimum stable observation window after each handover. The threshold for the number of consecutive sampling periods can be configured to 3 to 5 periods, for example, set to 3 periods, with the time interval between each sampling period being 5 to 30 seconds. The calculation method for the real-time comprehensive value includes a weighted sum of network performance indicators and environmental impact factors. For example, latency, packet loss rate, and signal strength are assigned weight coefficients of 0.3, 0.4, and 0.3 respectively, with the environmental impact factor added as a penalty.

[0090] Specifically, after a network handover is completed, a timer begins recording the dwell time of the current link. During the dwell time period, before reaching the minimum threshold, the system blocks all regular performance handover requests. Once the dwell time threshold is reached, the system initiates a multi-cycle verification mechanism: within each sampling cycle, real-time performance data of the currently active link and the backup link are simultaneously collected. This data, combined with the cellular network coverage status calculated based on the ship's current latitude and longitude, and the satellite signal obstruction status calculated based on the heading, generates a real-time comprehensive score for both links. When the backup link score exceeds the current link score for three consecutive sampling cycles, and the difference reaches a preset threshold (e.g., 10%), a regular performance handover is triggered. This mechanism works in conjunction with the calculation of environmental impact factors in the pre-processing information. For example, when a ship enters an area at the edge of cellular network coverage, the environmental impact factor dynamically adjusts the real-time score, ensuring link stability even in the early stages of signal attenuation and preventing erroneous handovers triggered by brief signal fluctuations. Through these dual constraints, the system ensures the necessity of handovers while reducing the number of handovers per unit time to less than 30% of the original scheme.

[0091] In the shipboard network handover system, the routine performance handover logic is configured to trigger a handover operation when two conditions are met. The minimum dwell time is set to 120 seconds, requiring that this time window must elapse since the last handover operation before the system enters the routine performance handover judgment process. The real-time comprehensive value is calculated based on a weighted sum of current network performance indicators and environmental impact factors. For example, the real-time comprehensive value of a cellular link is weighted by RTT, packet loss rate, and signal strength with coefficients of 0.5, 0.3, and 0.2 respectively, and then multiplied by the environmental impact factor. Three consecutive sampling periods are set, each lasting 5 seconds. The system must detect that the real-time comprehensive value of the backup link is higher than that of the currently active link for 15 consecutive seconds, and the difference in each period exceeds a preset threshold of 10%. When the above conditions are met simultaneously, the handover execution module, according to the instructions of the decision engine, triggers a priority adjustment command through the network interface priority adjustment mechanism of the operating system (such as Windows), thereby switching the network routing.

[0092] The decision engine makes optimal link switching decisions by comprehensively considering real-time performance, environmental awareness, and future performance predictions provided by the prediction engine. The link scoring process standardizes the real-time performance metrics (traffic, RTT, packet loss rate, signal strength) of each link (e.g., normalizes to the 0-1 range) and calculates a comprehensive real-time score for each link based on preset weights. Factors such as occlusion status output by the environmental awareness module are used as penalty items, lowering the score of the corresponding link.

[0093] The first step of the switching strategy (logical process) is emergency fault switching (highest priority): determine whether the current active link has detected no traffic or abnormal key indicators (such as 100% packet loss rate) for multiple consecutive sampling periods (e.g., 2 periods), and whether the backup link has a usable positive score. If "yes", immediately trigger the switch to the backup link.

[0094] The second step of the switching strategy (logic flow) is proactive switching prediction (medium priority): It determines whether the minimum dwell time (e.g., 120 seconds) has been met since the last switching, and whether the prediction engine shows that the prediction score of the backup link will continue to be significantly better than the prediction score of the current active link in the future for a specified period of time (e.g., 60 seconds). If "yes", switching to the backup link is triggered.

[0095] The third step of the switching strategy (logic flow) is a regular performance switching (low priority): it determines whether the minimum dwell time has been met since the last switching, and whether the real-time score of the backup link has been significantly better than the real-time score of the current active link for multiple consecutive sampling periods (e.g., 3 periods). If "yes", switching to the backup link is triggered.

[0096] The fourth step of the switching strategy (logic flow) is that if none of the above conditions are met, then no switching will be performed, and the current link will be maintained.

[0097] The hysteresis mechanism prevents frequent handovers caused by instantaneous fluctuations in network performance by introducing parameters such as minimum link dwell time and consecutively superior sampling period number.

[0098] Through the above technical solutions, the embodiments of the present invention effectively suppress erroneous handover operations caused by instantaneous fluctuations in network performance. The mandatory constraint of minimum dwell time avoids repeated handover actions within a short period, providing a necessary observation window for network stability. The continuous multi-cycle comparison mechanism, through data accumulation over time, filters out noise interference that may exist in a single sampling, ensuring the continuity of link performance advantages. This synergistic effect of dual conditions significantly improves the reliability of handover decisions, and the system reduces resource consumption caused by invalid handovers while maintaining network connection stability.

[0099] Secondly, this invention provides a shipborne active adaptive network switching system, see [link to relevant documentation]. Figure 2 It includes a perception module, a prediction module, a decision-making module, and an execution module; The perception module is used to determine the environmental impact factors corresponding to different network links based on multidimensional context information, including current multidimensional context information and historical multidimensional context information; environmental impact factors include current environmental impact factors and historical environmental impact factors; and based on the multidimensional context information of the vessel (including geographical location and heading), it determines whether there is physical shielding or other environmental impacts on each link, and uses it as a key influencing factor for link performance evaluation.

[0100] The prediction module includes a historical database, a query unit, and a calculation unit. The historical database is configured with and updated in real-time with historical data, including timestamped historical multidimensional context information, historical network performance indicators for different network links, and historical environmental impact factors for different network links. The query unit retrieves historical multidimensional context information clusters from the historical database whose similarity to future multidimensional context information exceeds a similarity threshold. The calculation unit calculates initial predicted values ​​for network performance indicators of different network links at a preset future time based on the historical network performance indicator clusters corresponding to the historical multidimensional context information clusters. It then determines future environmental impact factors based on the future multidimensional context information, uses these factors as penalty factors for the initial predicted values, and finally obtains the predicted values ​​for different network links at the preset future time. The decision-making module determines the real-time comprehensive value of different network links based on the current network performance indicators and current environmental impact factors, and formulates network link switching strategies based on the real-time comprehensive value and the predicted value. The execution module generates control commands based on the network link switching strategy, and changes the network link routing at the operating system level to achieve network link switching.

[0101] The switching execution module, based on instructions from the decision engine, modifies network routing at the operating system level to achieve link switching. This is implemented by interacting with network configuration interfaces provided by the host operating system (e.g., through system API calls or executing system command-line tools) to dynamically adjust the priority or routing metric values ​​of relevant network interfaces. Lower priority values ​​typically indicate higher routing priority. This operation usually requires system administrator privileges.

[0102] In this system, when the prediction module performs similarity matching, it obtains the current geographic environment parameters from the management module as the baseline value for the query radius. When formulating a switching strategy, the decision module reads the decision thresholds provided by the management module in real time for conditional judgment. A parameter persistence mechanism ensures that the system automatically restores the previous configuration after a restart, avoiding repeated initialization operations. When a vessel sails to the edge of cellular network coverage, the geographic parameters provided by the management module include the gradual attenuation coefficient of the base station coverage radius, and the prediction module adjusts the weight calculation method for future location similarity matching accordingly. When switching links, the execution module sets operating system-level routing table entries based on the routing priority rules saved by the management module. Through a dynamic parameter loading mechanism, satellite terminals of different vessel types can be independently configured with shielding sector angle parameters without modifying the core algorithm code. This parameter management system enables the system to maintain adaptability in various navigation scenarios, effectively solving the problems of prediction deviation and switching strategy failure caused by fixed preset parameters.

[0103] Through the above technical solutions, the embodiments of the present invention differ from existing technologies that rely solely on real-time performance indicators for passive switching or make simple judgments based on a single factor. The present invention introduces a "prediction engine," which records and analyzes historical network performance data of vessels in multi-dimensional contexts (including but not limited to geographical location, heading, speed, historical network load patterns, and environmental occlusion), and combines this with the vessel's current operating status to identify historical performance patterns, thereby predicting network performance on its future path. This allows the system to comprehensively and proactively detect link degradation trends, regardless of whether the cause is geographical change, increased network load, or other factors. This application implements dynamic adaptation and persistent storage of operating parameters, solving the prediction bias problem caused by fixed preset parameters. The management module updates geographical environment parameters in real time, enabling the prediction module to adjust similarity matching logic according to the current navigation area, avoiding prediction errors caused by mismatches in historical data. By dynamically loading decision thresholds and switching strategy parameters, the system can suppress erroneous switching behavior in extreme environments and reduce link jitter caused by parameter mismatches. The parameter linkage mechanism between multiple modules ensures coordinated optimization of the perception, prediction, decision-making, and execution stages, improving the accuracy and real-time performance of network switching in complex navigation environments.

[0104] As one possible implementation, the shipborne active adaptive network switching system also includes a management module for loading, saving and providing the system's operating parameters, which include at least the sampling period, network link switching strategy, geographical environment parameters, predicted values ​​and decision thresholds.

[0105] The management module loads initial operating parameters via a configuration file or application programming interface (API). The configuration file can be stored in JSON or XML format. Operating parameters are persistently stored via a local file system or an embedded database; for example, geographic environment parameters can be stored in an SQLite database. Examples include a sampling period range of 5-60 seconds and a cellular base station coverage radius of 20-50 kilometers. The storage function uses an embedded database or file system to persistently store parameters, such as using an SQLite database to record parameter update timestamps and version identifiers in real time. The provision function outputs parameter sets to the perception, prediction, and decision modules via standardized API interfaces. These interfaces can be based on RESTful or gRPC architectures; for example, providing a geographic matching threshold parameter query interface for the prediction module. The user interface module provides real-time status display, configuration management, and log viewing functions. Parameter updates receive new values ​​through the user interface module or external communication interfaces, with an update frequency that can be set to per second or per minute. The configuration management module is responsible for loading, saving, and providing all system operating parameters, including but not limited to data acquisition periods, link priority rules, geographic environment parameters, prediction parameters, and decision thresholds. Configuration data can be stored in local structured data files or an embedded database, and an interface is provided to allow dynamic parameter updates during system runtime. Geographic environmental parameters include the coverage radius of cellular base stations and the angle of obscured sectors, while decision thresholds include minimum dwell time and handover score difference thresholds. Parameters are provided using memory mapping or shared caching mechanisms to ensure that each module obtains the latest parameter values ​​in real time. Persistent parameter storage and real-time updates form a complementary mechanism. When a vessel enters polar waters, updated geographic parameters can correct the similarity matching radius of the prediction module, avoiding satellite link misjudgments caused by fixed preset parameters. Decision threshold adjustments affect the handover condition judgment of the decision module; for example, shortening the minimum dwell time from 120 seconds to 60 seconds can improve response speed in dense port environments. Hardware differences between different vessels are adapted by loading differentiated sampling period parameters; for example, vessels with lower satellite antenna sensitivity use longer sampling periods.

[0106] The user interface module provides an intuitive graphical user interface for displaying system status, configuration parameters, and viewing logs. The dashboard displays the status of each link obtained from the backend service in real time, including: link name, type, real-time traffic, RTT, packet loss rate, signal strength (if applicable), environmental awareness status (e.g., "obscured"), real-time comprehensive score, and predicted score. Link health and future trends are visually displayed using color coding, progress bars, or trend arrows. The name of the currently active link and the last switch time are displayed. The configuration management module provides user-friendly interface controls for setting and modifying all system parameters and sends new configurations to the backend service via an internal communication interface. The log viewer displays detailed operation logs, decision-making processes, and error messages generated by various system modules and supports log-level filtering and export. The manual control module provides the ability to manually start / stop the monitoring cycle and manually force a switch to a specified link.

[0107] Data acquisition involves periodically obtaining multidimensional real-time data required by the system from various sources. Network performance metrics are obtained through network interface information acquisition mechanisms provided by the operating system (e.g., parsing results from system APIs or command-line tools), acquiring the real-time packet transmission / reception volume of each network interface. Network probe requests (e.g., ICMP Echo Requests) are sent to preset stable network targets, and responses are analyzed to obtain the round-trip time (RTT) and packet loss rate for each link. Ship multidimensional context information is communicated with shipborne navigation sensors (such as GPS receivers and IMUs) via standard communication interfaces (e.g., serial port communication protocols). Data streams conforming to industry standard protocols (e.g., NMEA 0183) are parsed to extract real-time information such as the ship's latitude, longitude, heading, and speed. Signal strength (optional) attempts to obtain internal signal quality parameters (e.g., RSRP / SINR for cellular networks, Eb / No for satellite communications) through standardized interfaces provided by network communication terminal equipment (e.g., HTTP-based APIs, SNMP protocols, or sending vendor-specific command sets via communication interfaces). If the metric cannot be obtained, it can be configured to be invalid or assigned a lower weight.

[0108] In the dynamic parameter management mechanism, the dynamic configuration of the sampling period is achieved by modifying the numerical field in the configuration file. For example, in the complex waterway scenario of the port area, the sampling period field in the configuration file is updated to 5 seconds, while it is adjusted to 30 seconds during stable ocean navigation. The update of geographical environment parameters is completed by importing new base station coordinate datasets. For example, the latitude and longitude information of newly added cellular base stations is loaded into the management module in batches through CSV format files. The linkage adjustment of decision thresholds is achieved through the sliding window algorithm. For example, the minimum dwell time parameter in the handover strategy is dynamically adjusted from 120 seconds to 90 seconds to adapt to sudden changes in the network environment.

[0109] Specifically, the management module loads parameters from persistent storage upon system startup and initializes the configuration items for each functional module. During operation, parameter update commands are passed through the user interface module or external interface, and the management module writes the new parameters to memory and synchronizes them to the storage medium. Upon system startup, the management module reads the configuration file from local storage and loads preset parameters such as the sampling period and geographic matching threshold into the memory data structure. During operation, when changes in the navigation scenario or network service provider equipment updates are detected, parameter values ​​are written to the management module through the human-machine interface or automated scripts. The updated parameters are saved to the embedded database in real time, and a version control mechanism records the change history. Each functional module obtains the latest parameters by calling the unified interface provided by the management module. For example, when calculating the similarity threshold, the prediction module obtains the updated geographic proximity parameter value in real time.

[0110] When a vessel enters the port area, the management module receives an external parameter update command, adjusting the sampling period from 30 seconds to 5 seconds. The adjusted parameters take effect immediately. The perception module collects vessel position data based on the new sampling frequency, the prediction module queries historical data using the new geographic matching radius, and the decision module generates control commands by applying the updated switching strategy rules. The continuity of parameter adjustments is ensured through a database transaction mechanism, allowing the system to recover to the last persistently stored parameter state even after an abnormal system restart. Unified management of operational parameters eliminates inconsistencies in parameter transmission between modules. For example, when the decision threshold is adjusted from 0.7 to 0.6, the predicted value output by the prediction module is updated synchronously with the decision module's judgment criteria. The dynamic loading mechanism of geographic environment parameters enables the system to adapt to changes in regional network infrastructure. For example, after the coverage data of a newly added base station is imported, the environmental perception module can accurately identify the effective service area of ​​the cellular network.

[0111] The environmental perception module uses the ship's real-time multi-dimensional context information to determine the potential environmental impacts on each network link. Geographic distance determination calculates the distance between the ship's current geographical location and a preset geographic reference point (e.g., the approximate latitude and longitude of a known land-based cellular base station) to determine whether the cellular network link is outside its effective coverage area. Physical obstruction determination combines the ship's real-time heading to calculate the azimuth angle of the target signal source (e.g., a preset azimuth angle of a satellite or the relative azimuth of a cellular base station) relative to the bow. If this azimuth angle falls within a preset hull structure obstruction sector (e.g., a signal blind spot caused by the ship's superstructure or mast), it is determined to be physical obstruction; the determination result (e.g., whether the link is obstructed) is used as an input factor for link performance evaluation.

[0112] The system configures and updates a historical database in real time, recording timestamped historical multidimensional context information, network performance indicators, and environmental impact factors. By analyzing historical data, the system predicts the multidimensional context information of vessels at a preset future time. It then queries the historical database for historical data clusters whose similarity to the predicted future context information exceeds a threshold. This similarity threshold may include geographical proximity, heading similarity, etc. Based on the matched historical data clusters, the system calculates initial predicted values ​​for the performance indicators of different network links at the preset future time. Simultaneously, based on the predicted future context information, it determines future environmental impact factors and uses them as penalty factors to correct the initial predicted values, ultimately obtaining the performance predicted values ​​for each link at the future time. The system combines current network performance indicators and environmental impact factors to calculate a real-time comprehensive score for different links. Finally, based on the real-time comprehensive score and future predicted values, a network link switching strategy is formulated. The switching strategy may include multiple levels, such as emergency fault switching, proactive predictive switching, and routine performance switching.

[0113] The specific implementation of the solution in this embodiment of the invention is as follows: A parameter management service is deployed in the shipborne host computing platform. This service runs as an independent process and interacts with other modules of the system through an inter-process communication interface. The parameter management service is configured to load an initial parameter set from a local embedded database. The initial value of the sampling period is automatically set according to the network environment type currently in which the ship is located. For example, in a port scenario, the first configuration file is loaded and the sampling period is set to 5 seconds, while in a deep-sea scenario, the second configuration file is loaded and set to 30 seconds. The parameter saving function is implemented through an asynchronous write mechanism. Each parameter update operation generates a transaction log and is persistently stored in non-volatile memory, ensuring that the system can be restored to the latest parameter state through log replay after abnormal termination. The provision function is implemented through a publish-subscribe pattern. When the perception module requests geographic environment parameters, the parameter management service encapsulates the cellular base station coordinate set and satellite orbit occlusion sector data into a structured message and pushes it to the message queue, eliminating the resource consumption caused by inter-module polling queries. When the prediction module detects that the historical database query pattern has become rigid, the parameter management service receives a similarity threshold adjustment request from the prediction engine, dynamically expands the geographic proximity matching radius from the preset 500 meters to 1000 meters, and synchronously updates the switching strategy priority rules in the decision engine.

[0114] The management module is configured to store operational parameters via an embedded database and acquire real-time multi-dimensional contextual information about the vessel (e.g., geographical location, heading, speed) and real-time performance metrics of each network link (traffic, RTT, packet loss rate, signal strength, etc.). Geographical parameters include the coverage radius and shielding sector angle of the cellular base station. When the vessel enters polar waters, the management module receives updated geographic parameters via an external interface, reducing the cellular base station coverage radius from a preset 20 nautical miles to 15 nautical miles and adjusting the shielding sector angle from 180 degrees to 120 degrees. The updated parameters are loaded in real-time into the similarity matching logic of the prediction module, enabling the query unit to prioritize matching historical multi-dimensional contextual information clusters of the polar region when searching the historical database. The calculation unit calculates the initial predicted value of the satellite link based on the matching results and adds future environmental impact factors as penalty factors to generate a corrected predicted value. The decision module, based on the corrected predicted value and the real-time comprehensive value, increases the active handover score threshold from 5% to 8%, ultimately generating a link handover strategy to suppress frequent handovers. The execution module adjusts routing priorities according to the policy to maintain satellite link stability in the polar low-temperature environment.

[0115] Through the above technical solutions, the embodiments of the present invention achieve dynamic configuration of operating parameters and consistency maintenance between modules, solving the adaptability defects caused by parameter fixation. In near-shore complex waterway scenarios, the real-time performance of data acquisition is improved by shortening the sampling cycle, avoiding delays in occlusion judgment caused by sudden environmental changes; in areas with newly deployed cellular base stations, the matching accuracy of signal coverage is optimized by dynamically loading updated geographical parameters, reducing historical database query deviations; by adjusting predicted values ​​and decision thresholds in a coordinated manner, the differences in link characteristics among different network service providers are balanced, suppressing premature or delayed handover caused by historical data deviations, ultimately reducing the timeliness error of network handover actions to within seconds.

[0116] To facilitate understanding of the technical solution of this application, further explanation is provided below with reference to specific embodiments.

[0117] Example 1 As a preferred embodiment, the solution of the present invention is implemented as follows: Suppose a tugboat, equipped with a shipborne network active adaptive switching system and method (hereinafter referred to as "this system"), is performing a task: first completing auxiliary operations in port, and then towing a large floating structure from nearshore waters to a deep-sea operation area. The vessel is equipped with two main external network links: Link A (Cellular Network): Connects to shipboard cellular network communication terminals (e.g., 4G / 5G routers), suitable for ports, coastal and nearshore areas, featuring high bandwidth and low latency, mainly used for daily communication, receiving dispatch instructions, real-time video monitoring and crew internet access.

[0118] Link B (Satellite Network): Connects to shipborne satellite communication terminals (e.g., VSAT), suitable for ocean areas far from land, with wide coverage, but usually with low bandwidth and high latency, mainly used as backup and remote operation communication when cellular network signals are insufficient.

[0119] The tugboat is equipped with navigation sensors (GPS / IMU) capable of outputting accurate latitude, longitude, heading, and speed data in real time. This system is running as a background service on an onboard computing platform (e.g., a PC running a general-purpose operating system), with data acquisition, environmental awareness, prediction engine, and decision engine activated. The system's prediction engine has accumulated link performance data from the tugboat's historical operations in different geographical locations, headings, and port / coastal environments, specifically recording the 4G signal coverage attenuation patterns at different offshore distances.

[0120] Key system configurations (examples): Sampling interval: 5 seconds (the system collects data and makes a judgment every 5 seconds); Minimum dwell time: 120 seconds (after a link switch, it must remain switched for at least 120 seconds without switching again); Actively switch prediction duration: 300 seconds (for predicting performance in the next 5 minutes); Active handover score threshold: 5% (Active handover is triggered when the predicted score of another link is consistently 5% higher than that of the current link). Number of fault samples: 2 (no flow for 2 consecutive sampling periods is considered a fault); Historical data retention: 30 days; Context matching radius: 0.5 km (used for accurate matching of local occlusion when operating within the port. For offshore prediction, it relies more on the overall geographic region pattern). Context matching heading difference: 30 degrees.

[0121] Implementation scenarios and steps: Scenario: The tugboat has completed berthing assistance operations in the port and is now towing a large floating platform, sailing along the predetermined route towards the deep-sea operation area.

[0122] Initial phase: Operations within the port and nearshore waters (cellular network links are preferred). S1. System Startup and Initialization: This system service starts up, and according to the configuration, the cellular network link (link A) is set as the first choice, its operating system network interface priority is set to high, and the satellite link (link B) priority is set to low.

[0123] S2. Data acquisition module operation: Performs the following actions every 5 seconds: S21. Obtain real-time traffic and packet statistics for link A and link B from the host operating system; S22. Obtain the tugboat's current precise latitude and longitude (e.g., 25.0000N, 121.0000E), heading (e.g., 90°), and speed from the navigation sensors; S23. Perform network probing on public network targets to obtain the RTT and packet loss rate of link A; link B may not be fully activated because it is not the preferred choice, or its RTT and packet loss rate may be high. S24. (If the communication terminal supports this) Obtain the signal strength of link A (e.g., RSRP / SINR).

[0124] S3. Environmental Awareness Module Operation: Determines whether Link A and Link B are obstructed by the environment. In ports and nearshore waters, cellular network links typically have good signal strength and are not subject to fixed geographical obstruction (temporary obstruction is indirectly reflected by historical data from the prediction engine).

[0125] S4. Prediction engine operation: S41. Data Recording: Record the currently collected multi-dimensional context information (position, heading, speed, etc.) and link performance data into the historical database; S42. Performance Prediction (Initial Reference Real-Time): Based on the current context and referring to historical data, the prediction engine predicts that the score of Link A (cellular network) will continue to be higher than that of Link B (satellite).

[0126] S5. Decision Engine Operation: S51. Calculate the real-time scores of link A and link B. Because link A has good performance (low RTT, low packet loss, high traffic, high signal strength), its real-time score is much higher than that of link B; S52. Referring to the prediction score of the prediction engine, it is also shown that the prediction score of link A is higher; S53. No switching conditions were triggered (Link A is stable and has the highest score).

[0127] S6. User Interface Display: The user interface displays that the cellular network link is currently the preferred choice, with all indicators showing good performance and both real-time and predicted scores being high.

[0128] Key Phase: Leaving near-shore waters and entering the 4G signal attenuation zone (active predictive handover) S7. Tugboat Navigation: The tugboat tows the floating platform, continuously sailing at a predetermined course (e.g., 100°) and speed, gradually leaving the land and entering the edge area of ​​4G cellular network signal coverage. At this point, the tugboat's position reaches 24.9000N, 121.3000E. The real-time performance of Link A (cellular network) may still be acceptable, or only experience slight fluctuations, but has not yet deteriorated significantly enough to trigger a passive handover.

[0129] S8. Core operation of the prediction engine: S81. Future Context Estimation: Based on the tugboat's current position (24.9000N, 121.3000E), current heading (100°), and speed, the prediction engine estimates the tugboat's approximate position (24.8500N, 121.4000E) 300 seconds (5 minutes) in advance.

[0130] S82. Similar Pattern Query and Analysis: The prediction engine uses the inferred future context as a query condition to search the historical database for historical flight data with similar headings near the future location. The prediction engine finds that in the multi-dimensional context similarity historical records for the predicted location, the historical performance data of Link A (cellular network) generally shows that as the offshore distance increases, the RTT increases sharply (e.g., from 50ms to 1500ms), the packet loss rate increases significantly (e.g., from 5% to 80%), and the signal strength decreases sharply, indicating that it is about to exceed the reliable 4G coverage range; while the historical performance of Link B (satellite) remains stable in this area.

[0131] S83. Prediction Score Generation: Based on this, the prediction engine calculates that the predicted score for link A is expected to drop from the current 0.70 to 0.15 within 300 seconds. The predicted score for Link B is expected to remain at 0.65 within 300 seconds.

[0132] Results: The score of predicted link B (0.65) is significantly higher than the score of predicted link A (0.15), and its advantage (0.65 / 0.15-1) = 333% far exceeds the active handover threshold of 5%.

[0133] S9. Core functions of the decision engine: S91. Receive the prediction results from the prediction engine; S92. Determine if the current conditions for proactive predictive switching are met: The minimum dwell time has been met; the current real-time performance of Link A (cellular network) has not completely failed; most importantly, the predicted future score of Link B (0.65) is significantly higher than the predicted future score of Link A (0.15), meeting the active handover score threshold. S93. Trigger Decision: The decision engine decides to perform an active switch from link A to link B.

[0134] S10. Switch the execution module: S101. Receive decision engine instructions.

[0135] S102. Immediately use the network interface priority adjustment mechanism provided by the operating system to increase the priority of satellite link B and decrease the priority of cellular network link A.

[0136] S11. User Interface and Logs: S111. The user interface immediately updates to show that the currently preferred link is "Satellite Network"; S112. System log record: "Active handover: It was predicted that the cellular network link would experience a significant performance degradation within the next 5 minutes due to leaving near-shore waters, and a switch to the satellite link has been initiated." S113. Crew members and dispatchers on the tugboat may only perceive a momentary pause of less than 1 second, but this effectively avoids the long-term interruption caused by switching only when the 4G signal is completely lost (for example, after the tugboat enters the deep sea), ensuring the continuity of critical communications with the dispatch center (such as real-time location reporting and operation instruction interaction), especially in tasks with high requirements for communication stability, such as towing large platforms.

[0137] Subsequent phase: Ocean-going towing operations (satellite link preferred) S12. Link B operates stably: As the tugboat enters open ocean waters, the satellite link (Link B) maintains a stable connection. This system continuously monitors the real-time performance of Link B and records its multidimensional context and performance data.

[0138] S13. Validating the prediction: As the tugboat continued to sail, the real-time performance of the cellular network link did indeed gradually deteriorate and eventually disconnect completely; this validated the accuracy of the prediction engine and the necessity of proactive switching.

[0139] S14. Responding to emergencies (reflexive switching): Suppose that during ocean towing operations, a sudden severe weather event (such as a strong storm) occurs, causing the satellite communication antenna to be temporarily blocked by waves or the signal to be affected by heavy rainfall, resulting in a sudden deterioration of the satellite signal, or even two consecutive sampling cycles without data flow.

[0140] S141. Data Acquisition and Decision Making: The data acquisition module detects that there is no traffic in link B for two consecutive sampling periods. The decision engine identifies it as an emergency fault. If link A (cellular network, assuming the tugboat has completed its ocean towing mission and is about to return to port, and is about to enter the coverage area of ​​another coastal base station) has a positive score at this time; S142. Trigger Decision: The decision engine immediately triggers an emergency failover, raising the priority of link A and restoring it to the preferred position; S143. User Interface and Logs: The user interface update shows the cellular network link as the preferred link and logs "Emergency Switchover: The satellite link was interrupted due to a sudden failure (severe weather), and the connection has been switched to the cellular network link."

[0141] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, disclosure, and other materials. In this specification, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple components. A single processor or other unit can implement several functions listed in the specification. While certain measures are described in different embodiments, this does not mean that these measures cannot be combined to produce good results.

[0142] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A shipborne active adaptive network handover method, characterized in that, Includes the following steps: S1. During the ship's journey, collect the ship's current multi-dimensional context information and the current network performance indicators of different network links in real time according to the preset collection cycle; S2. Determine the current environmental impact factors corresponding to different network links based on the current multidimensional context information; S3. Configure and update the historical database in real time. The historical data includes historical multidimensional context information with timestamps, historical network performance indicators of different network links, and historical environmental impact factors of different network links. S4. After predicting the future multidimensional context information of the ship at a preset future time, query the historical database for historical multidimensional context information clusters whose similarity to the future multidimensional context information exceeds a similarity threshold; S5. Calculate the initial predicted values ​​of network performance indicators for different network links at a future preset time based on the historical network performance indicator clusters corresponding to the historical multidimensional context information clusters; Future environmental impact factors are determined based on multidimensional contextual information. These factors are then used as penalty factors for the initial predicted values ​​to ultimately obtain the predicted values ​​for different network links at a predetermined future time. The multidimensional contextual information includes at least the ship's latitude, longitude, heading, and speed. Different network links specifically include satellite network links and cellular network links. The environmental impact factors corresponding to cellular network links are determined as follows: based on the distance between the ship's latitude and longitude and a predetermined geographical reference point, it is determined whether the ship exceeds the coverage area of ​​the cellular network link. If it does, the environmental impact factor is set to a negative number; otherwise, it is set to a positive number. The predetermined geographical reference point is the set of location coordinates of cellular base stations. A geofencing algorithm is used to calculate the distance between the ship's current location and the nearest cellular base station to determine whether it exceeds the cellular network coverage area, which is 1.5 times the effective coverage radius. The environmental impact factor corresponding to the satellite network link is determined as follows: the azimuth angle of the target signal source relative to the bow is calculated based on the ship's heading, and it is determined whether the azimuth angle falls within the preset ship structure shielding sector. If so, the environmental impact factor is set to a negative number; otherwise, the environmental impact factor is set to a positive number. The preset parameter for the ship structure shielding sector is a sector area of ​​±45 degrees in the bow direction. The multidimensional context information includes the current multidimensional context information and the historical multidimensional context information. S6. Determine the real-time comprehensive value of different network links based on the current network performance indicators and current environmental impact factors, and formulate network link switching strategies based on the real-time comprehensive value and the predicted value; the priority is from high to low, and the network link switching strategies include emergency fault switching, proactive predictive switching and normal performance switching.

2. The shipborne active adaptive network handover method according to claim 1, characterized in that, In the historical database, query historical multidimensional context information clusters whose similarity to future multidimensional context information exceeds a similarity threshold, wherein the similarity threshold includes at least: geographical proximity, heading similarity, and / or historical time period matching degree.

3. The shipborne active adaptive network handover method according to claim 2, characterized in that, For each network link, based on the queried historical multidimensional context information cluster, a weighted summation method is applied to calculate the initial predicted value of the network performance index at a future preset time. Based on the initial predicted value and the penalty factor, the predicted value of different network links at a future preset time is calculated by summing.

4. The shipborne active adaptive network handover method according to claim 1, characterized in that, The conditions for emergency failover are: the current network link has detected no traffic or abnormal key indicators in multiple consecutive sampling periods, and the real-time comprehensive value of the backup network link has a positive score.

5. The shipborne active adaptive network handover method according to claim 1, characterized in that, The conditions for proactively predicting handover are: whether the minimum dwell time has been met since the last handover, and whether the predicted value of the backup network link is consistently better than the predicted value of the current network link during the time period from the current time to a preset future time.

6. The shipborne active adaptive network handover method according to claim 1, characterized in that, The conditions for normal performance switching are: whether the minimum dwell time has been met since the last switch, and whether the real-time comprehensive value of the backup network link is better than the real-time comprehensive value of the currently active link for multiple consecutive sampling periods.

7. A shipborne active adaptive network handover system, characterized in that, It includes a perception module, a prediction module, a decision-making module, and an execution module; The perception module is used to determine the environmental impact factors corresponding to different network links based on multidimensional context information. The multidimensional context information includes current multidimensional context information and historical multidimensional context information; the environmental impact factors include current environmental impact factors and historical environmental impact factors. The prediction module includes a historical database, a query unit, and a calculation unit. The historical database is configured with and updated in real-time with historical data, including timestamped historical multidimensional context information, historical network performance indicators for different network links, and historical environmental impact factors for different network links. The query unit retrieves historical multidimensional context information clusters from the historical database whose similarity to future multidimensional context information exceeds a similarity threshold. The calculation unit calculates initial predicted values ​​for network performance indicators of different network links at a preset future time based on the historical network performance indicator clusters corresponding to the historical multidimensional context information clusters. It then determines future environmental impact factors based on the future multidimensional context information, uses these factors as penalty factors for the initial predicted values, and finally obtains the predicted values ​​for different network links at the preset future time. The decision-making module determines the real-time comprehensive value of different network links based on the current network performance indicators and current environmental impact factors, and formulates network link switching strategies based on the real-time comprehensive value and the predicted value. The execution module generates control commands based on the network link switching strategy, and changes the network link routing at the operating system level to achieve network link switching.

8. The shipborne active adaptive network switching system according to claim 7, characterized in that, The shipborne active adaptive network switching system also includes a management module for loading, saving and providing the system's operating parameters, which include at least the sampling period, network link switching strategy, geographical environment parameters, predicted values ​​and decision thresholds.

Images