Edge cache migration method based on cross-terminal trajectory prediction

Abstract

The present application relates to the technical field of satellite communication and edge computing, and discloses an edge cache migration method based on cross-terminal trajectory prediction, which comprises the following steps: collecting the historical request log of a first terminal, inputting a trained trajectory prediction model, and obtaining the trajectory prediction data of a second terminal within a preset time period; obtaining the predicted request content of the second terminal, and constructing cache data; sending the cache data to a first edge cache node determined according to the original position information of the second terminal for caching through a cloud platform; when the second terminal initiates a request, judging whether the second terminal has migrated; if yes, matching the corresponding migration strategy in a preset cache migration mechanism according to the environmental characteristics, determining the target edge cache node and obtaining the actual request content; and if no, obtaining the actual request content from the first edge cache node. The present application realizes the adaptive selection of the cache migration strategy, so that the system can adapt to complex scenarios such as high mobility of satellite networks and dynamic changes of resources.

Description

Technical Field

[0001] This invention relates to the fields of satellite communication and edge computing technology, and in particular to an edge cache migration method based on cross-terminal trajectory prediction. Background Technology

[0002] Content caching and distribution mechanisms based on edge nodes can effectively reduce network latency and improve data access efficiency. In satellite IoT scenarios, due to the wide distribution and high-speed movement of terminals, how to achieve efficient caching and transmission under the constraints of limited link and storage resources has become an urgent technical problem to be solved.

[0003] Existing caching mechanisms typically deploy caches based on content access frequency or the current location of the terminal to improve cache hit rate or reduce access latency. However, these methods often rely on a single factor for decision-making, lacking the ability to predict changes in terminal behavior and adapt to dynamic changes in the network environment. Furthermore, while some technical solutions introduce user behavior prediction models to achieve prefetching caching, they still have the following shortcomings: 1. Lack of cross-terminal correlation modeling capability. Existing prediction methods are mostly based on single-terminal historical behavior modeling, lacking the ability to explore the correlation between terminals and failing to characterize the propagation characteristics of request behavior, thus limiting the accuracy and generalization ability of predictions. 2. Lack of environment-adaptive caching decision-making capability. Existing caching strategies mostly rely on static rules or single indicators, failing to comprehensively consider factors such as link quality fluctuations, node coordination capabilities, and dynamic changes in storage resources, resulting in insufficient cache adaptability in environments with high mobility, weak coordination, or complex links. 3. Monotonous cache migration strategies. Existing cache migration mechanisms mostly employ fixed strategies, such as migration to the nearest location or simple copying, lacking the ability to differentiate for different environments, easily leading to resource waste or increased terminal request response latency. 4. Lack of synergy between prediction results and cache migration mechanisms. Prediction models and cache migration mechanisms are usually independent of each other, lacking effective linkage, making it difficult to form a closed-loop optimization from prediction and decision-making to migration, limiting the role of prediction information in cache scheduling. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an edge cache migration method based on cross-terminal trajectory prediction. This method solves the problems of existing technologies failing to consider cross-terminal correlation modeling, adaptive cache decision-making, having a single cache migration strategy, and lacking coordination between prediction results and cache migration mechanisms. This invention enables predictive deployment and dynamic migration optimization of cache resources, thereby improving cache hit rate and reducing data access latency. The technical solution provided by this invention includes: The edge cache migration method based on cross-terminal trajectory prediction includes the following steps: S1. Collect the current request logs of the first terminal and the second terminal, input the trained trajectory prediction model, and obtain the trajectory prediction data of the second terminal within a preset time period; S2. Obtain the predicted request content of the second terminal and construct cache data; send the cache data to the first edge cache node determined according to the original location information of the second terminal through the cloud platform for caching; S3. When the second terminal initiates a request, determine whether the current location has migrated relative to the original location information; if yes, then match the corresponding migration strategy in the preset cache migration mechanism according to the environmental features extracted from the confidence level of the trajectory prediction model, the satellite link quality and the storage status of the edge node, determine the target edge cache node and obtain the actual request content; if no, then obtain the actual request content from the first edge cache node.

[0005] Preferably, the training method for the trajectory prediction model includes: Historical request logs from multiple terminals are collected, and individual features, cooperative movement features, and basic state features of the first and second terminals are extracted from the historical request logs as a training set to train the trajectory prediction model constructed through graph convolutional networks, temporal convolutional networks, and attention mechanisms.

[0006] Preferably, the method for determining whether the second terminal has migrated includes: calculating the distance between the current location and the original location; if the distance is less than a preset migration threshold, then the second terminal is determined not to have migrated; otherwise, the second terminal is determined to have migrated.

[0007] Preferably, a corresponding migration strategy is matched in a preset cache migration mechanism based on the environmental characteristics. Specific methods include: If the single beam switching time is less than the first preset threshold of the cached data prefetch period, the current environment is determined to meet the high mobility condition; if the inter-node signaling delivery success rate is less than the second preset threshold, the current environment is determined to meet the weak coordination condition; if the environment meets the determination conditions of high mobility and / or weak coordination, a multi-node redundancy distribution mechanism is triggered before the second terminal initiates a request.

[0008] Preferably, the multi-node redundancy distribution mechanism includes: determining multiple prediction cache nodes as target edge cache nodes based on the trajectory prediction data, and copying and distributing the cached data to the multiple target edge cache nodes for caching through a cloud platform.

[0009] Preferably, a corresponding migration strategy is matched in a preset cache migration mechanism based on the environmental characteristics. Specific methods include: If the collaborative access similarity is greater than the third preset threshold and / or the spatiotemporal trajectory overlap is greater than the fourth threshold, then the current environmental features are determined to meet the strong correlation condition; if the available bandwidth margin is greater than the fifth preset threshold and / or the channel signal-to-noise ratio is greater than the sixth preset threshold, then the current environmental features are determined to meet the bandwidth advantage condition; if the environmental features meet the determination conditions of strong correlation and / or bandwidth advantage, then the edge node P2P collaborative transmission mechanism is triggered.

[0010] Preferably, the edge node P2P cooperative transmission mechanism includes: The cached data of the first edge cache node is transmitted to the target edge cache node determined based on the current location information of the second terminal for caching.

[0011] Preferably, a corresponding migration strategy is matched in a preset cache migration mechanism based on the environmental characteristics. Specific methods include: If the remaining storage ratio of the first edge cache node is less than the seventh preset threshold, the current environmental characteristics are determined to meet the storage-limited condition; if the satellite link packet loss rate is greater than the eighth preset threshold, the current environmental characteristics are determined to meet the extreme environmental condition; if the environmental characteristics meet the determination conditions of storage-limited and / or extreme environment, the cloud platform resource routing pointer redirection mechanism is triggered.

[0012] Preferably, the cloud platform resource routing pointer redirection mechanism includes: The cloud platform is used to generate a metadata index pointing to the first edge cache node on the target edge cache node, and the cached data is obtained from the first edge cache node through the metadata index.

[0013] Preferably, a corresponding migration strategy is matched in a preset cache migration mechanism based on the environmental characteristics. Specific methods include: If the environmental characteristics meet the criteria for link interruption, then when the second terminal makes a request, the cached data is resent to the target edge cache node determined based on the current location information of the second terminal through the cloud platform for caching.

[0014] Beneficial effects: This invention enables advance perception of the target terminal's future trajectory and requested content, improving the accuracy of cache pre-deployment. By introducing environmental feature vectors based on prediction confidence, link quality, and node storage status, it achieves adaptive selection of cache migration strategies, enabling the system to adapt to complex scenarios such as high mobility, weak coordination, and dynamic resource changes in satellite networks. Furthermore, by constructing a multi-strategy collaborative cache migration mechanism, it links and optimizes the prediction results with the cache migration process, forming a complete closed-loop solution. This effectively improves cache hit rate, reduces data access latency, and reduces the transmission of invalid data and resource overhead. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating an edge cache migration method based on cross-terminal trajectory prediction provided in a preferred embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1

[0019] like Figure 1 As shown, this invention provides an edge cache migration method based on cross-terminal trajectory prediction.

[0020] Existing caching mechanisms typically deploy caches based on content access frequency or the current location of the terminal. By pre-caching frequently accessed data or storing data on edge nodes closer to the terminal, they aim to improve cache hit rates or reduce access latency. While some technical solutions incorporate user behavior prediction models for cache prefetching, they still primarily rely on historical behavior of a single terminal or static rules for caching decisions.

[0021] Existing prediction methods are mostly based on single-terminal modeling, lacking the ability to explore the relationships between terminals, making it difficult to accurately predict the behavior of target terminals. Their caching decisions lack comprehensive consideration of environmental factors such as link quality and node status, making them difficult to adapt to dynamically changing network environments. Caching migration strategies are relatively simplistic, lacking the ability to differentiate processing for different scenarios. Furthermore, the prediction results and the cache migration process are independent of each other, failing to form an effective linkage, resulting in a significant reduction in cache scheduling efficiency. To address these technical problems, in some preferred embodiments, the specific steps of the technical solution of this invention are as follows: S1. Collect the current request logs of the first terminal and the second terminal, input the trained trajectory prediction model, and obtain the trajectory prediction data of the second terminal within a preset time period.

[0022] A trajectory prediction model is a model used to predict the future location changes of a target terminal. Based on the prediction results, the association probability between a second terminal and each candidate edge cache node can also be calculated, providing a basis for subsequently determining the nodes for target cache migration. In existing technologies, trajectory prediction methods typically model based on the historical location of a single terminal, predicting the terminal's future movement trajectory through time series analysis or simple machine learning methods. This method only models a single terminal, lacking the ability to explore the relationships between different terminals, making it difficult to characterize the propagation characteristics of request behavior between terminals, resulting in low accuracy and generalization ability of the prediction results. To address these issues, in some preferred embodiments, this invention collects the current request logs of a first terminal and a second terminal, inputs them into a pre-trained trajectory prediction model, and obtains the trajectory prediction data of the second terminal within a preset time period, achieving cross-terminal trajectory prediction. The predicted trajectory data includes the trajectory prediction coordinates of the second terminal and a probability heatmap. In some preferred embodiments, the trajectory prediction model of this invention is obtained by collecting historical request logs from multiple terminals, extracting individual features, cooperative movement features, and basic state features of the first and second terminals as a training set, and constructing and training it using graph convolutional networks, temporal convolutional networks, and attention mechanisms.

[0023] S2. Obtain the predicted request content of the second terminal and construct cache data; send the cache data to the first edge cache node determined according to the original location information of the second terminal through the cloud platform for caching.

[0024] In some preferred embodiments, the method for obtaining the predicted request content of the second terminal according to the present invention includes: deploying an AI prediction model trained with the associated features of the first and second terminals on a cloud platform; using the historical behavior data of the first terminal as input to the AI ​​prediction model on the cloud platform; and obtaining the predicted request content of the second terminal within a preset time period through model feature mapping and prediction calculation. The method for obtaining the predicted request content can also be a reasonable design by those skilled in the art based on actual conditions or site requirements, and the present invention does not limit it. Then, a time-sensitive tag is attached to the predicted request content to construct cached data. In some preferred embodiments, the constructed cached data can be sent through the cloud platform to a first edge cache node determined based on the original location information of the second terminal for caching, thereby achieving advance deployment of cached resources.

[0025] S3. When the second terminal initiates a request, determine whether the current location has migrated relative to the original location information; if yes, then match the corresponding migration strategy in the preset cache migration mechanism according to the environmental features extracted from the confidence level of the trajectory prediction model, the satellite link quality and the storage status of the edge node, determine the target edge cache node and obtain the actual request content; if no, then obtain the actual request content from the first edge cache node.

[0026] Here, environmental features refer to a set of features extracted based on network status and node resource status to characterize the current system operating environment. In some preferred embodiments, the environmental features of the present invention include state features such as high mobility, weak coordination, strong correlation, excellent bandwidth, and limited storage; the environmental features are obtained by analyzing and calculating basic indicators such as the confidence level of the trajectory prediction model, satellite link quality, and storage status of edge cache nodes.

[0027] In existing technologies, when a terminal's location changes, the nearest edge node is typically selected for cache access based solely on the current location, or a fixed strategy is used for data migration, such as simple data copying or migration to the nearest node. However, such methods do not comprehensively consider factors such as link quality, node resource status, and prediction reliability, lacking differentiated decision-making capabilities for different network environments, which can easily lead to resource waste or increased access latency. To address these technical problems, this invention, when the second terminal initiates a request, first obtains the current location information of the second terminal and determines whether migration has occurred based on the relationship between the current location information and the original location information. If no migration has occurred, cached data corresponding to the actual request content is directly matched from the first edge cache node for response. If migration is determined to have occurred, basic indicators such as the confidence level of the trajectory prediction model, satellite link quality, and storage status of the edge cache node are further obtained, and current environmental features are extracted based on these basic indicators. Subsequently, in a preset cache migration mechanism, a corresponding migration strategy is matched according to the environmental features, and a target edge cache node is determined. The cached data is then migrated to the target edge cache node to achieve dynamic optimization and scheduling of cache resources.

[0028] This embodiment explores the behavioral propagation characteristics and correlations between different terminals, significantly improving the accuracy and generalization ability of second terminal trajectory prediction, providing a reliable decision-making basis for subsequent cache pre-deployment. Simultaneously, it realizes the transformation of cache prefetching from passive response to proactive deployment, effectively improving cache hit rate and reducing access latency for initial terminal requests. For dynamically changing network environments, this embodiment abandons the single, fixed cache migration strategy in existing technologies, realizing personalized migration decisions based on scenario differences, flexibly adapting to different network operating environments; it significantly improves cache scheduling efficiency and edge node resource utilization, and is particularly suitable for complex network scenarios with high-speed terminal movement, significant link fluctuations, and heterogeneous node resources, exhibiting strong robustness and environmental adaptability.

[0029] Example 2

[0030] Based on Example 1, this example provides a better training method for trajectory prediction models.

[0031] To address the issues of low prediction accuracy and poor generalization in existing trajectory prediction technologies that rely solely on single-terminal historical data and lack cross-terminal correlation feature mining, the training method for the trajectory prediction model of this invention includes, in some preferred embodiments, collecting historical request logs from multiple terminals and extracting individual features, cooperative movement features, and basic state features of a first terminal and a second terminal from the historical request logs as a training set. This training set is used to train the trajectory prediction model constructed using graph convolutional networks, temporal convolutional networks, and attention mechanisms. The historical request logs refer to the collection of raw data such as location, movement, access, and link status recorded by the terminal within a past time window, used to extract features for training the trajectory prediction model. Since the trajectory prediction model of this invention cannot only process time series but also spatial topological relationships, this invention extracts three-dimensional features from the historical request logs to construct the training set. Individual features refer to the terminal's latitude, longitude, altitude, speed, acceleration, etc.; cooperative movement features refer to the overlap and service correlation of the first and second terminals under the same satellite beam; and basic state refers to satellite ephemeris data, current link bandwidth, and remaining storage at edge nodes, etc.

[0032] In some preferred embodiments, after obtaining the above parameters, the present invention first extracts the spatial correlation across terminals using a Graph Convolutional Network (GCN). Each terminal is considered a node in a graph, and the relationships between terminals (such as spatial distance, whether they are within the same satellite beam coverage area) are considered edges in the graph. Feature information from neighboring terminals is aggregated through graph convolution operations to obtain the potential impact of neighboring terminals on the current target terminal. Then, a Temporal Convolutional Network (TCN) is used to extract the periodicity and trend of terminal movement. Through the structural design of causal convolution and dilated convolution, the network can process long-term series data, capturing the long-term dependencies and short-term fluctuations of terminal position, speed, etc., over time. Finally, an attention mechanism is used to learn which neighboring terminals have a greater predictive weight for the current target terminal. When aggregating feature information from multiple neighboring terminals, the attention mechanism assigns different attention coefficients to different neighboring nodes, enabling the model to automatically focus on neighboring terminals that are more important for predicting the current target terminal, suppressing the influence of irrelevant or interfering information.

[0033] The trajectory prediction model trained in this embodiment can perceive the potential influence of adjacent terminals on the current terminal's movement trajectory, overcoming the limitation of traditional methods that can only handle single terminal time series. Secondly, by extracting the periodic and trend features of terminal movement through a temporal convolutional network, it enhances the ability to model long-term dependencies. Through an attention mechanism, it automatically learns the influence weights of different adjacent terminals, enabling the model to focus on key related terminals, suppress irrelevant interference, and improve the accuracy of prediction. The trajectory prediction model in this embodiment can characterize the propagation characteristics of request behavior between terminals, further improving the accuracy and generalization ability of the prediction results, and providing a reliable foundation for triggering subsequent cache migration strategies.

[0034] Example 3

[0035] Based on Example 1, this example is used to determine whether the second terminal has migrated.

[0036] In order to determine whether the terminal has migrated, in some preferred embodiments, the method of the present invention for determining whether the second terminal has migrated includes: calculating the distance between the current position and the original position; if the distance is less than a preset migration threshold, it is determined that the second terminal has not migrated; otherwise, it is determined that the second terminal has migrated. First, the system obtains the current location information of the second terminal when it initiates the request. In some preferred embodiments, the current location information can be obtained from the probability heatmap and corresponding trajectory prediction coordinates in the trajectory prediction data. It also obtains the pre-stored original location information of the second terminal, i.e., the location information of the second terminal when the cached data is deployed to the first edge cache node. Second, the system calculates the actual straight-line distance between the current location and the original location of the second terminal using a positioning algorithm, or combines this with the path distance of the terminal's movement trajectory. A preset migration threshold is then invoked. This preset migration threshold can be flexibly configured based on scenario parameters such as the edge node's service radius, the terminal's movement rate, and network transmission latency requirements. Finally, the calculated actual distance is compared with the preset migration threshold. If the actual distance is less than the preset migration threshold, it indicates that the movement range of the second terminal has not exceeded the effective service range of the first edge cache node, and the second terminal has not migrated. In this case, the system directly matches the cached data response request from the first edge cache node. If the actual distance is greater than or equal to the preset migration threshold, it indicates that the movement of the second terminal has exceeded the effective service range of the first edge cache node, and the second terminal has migrated. This triggers a subsequent migration strategy based on environmental features to determine the target edge cache node and complete the cache migration process. This embodiment can reduce data communication latency and ensure that terminal requests can be responded to quickly and efficiently.

[0037] Example 4

[0038] Based on any one of Embodiments 1 to 3, this embodiment is used to match the corresponding migration strategy in the preset cache migration mechanism.

[0039] To address the shortcomings of existing technologies in scenarios with high terminal mobility and weak inter-node coordination, where cache migration strategies lack advance prediction and targeted processing, and still rely on the terminal initiating a request before migration is executed, leading to untimely migration, increased request latency, and even cache access failures, the present invention proposes a preferred migration strategy. This strategy includes: first, extracting core fundamental indicators of the current network environment, including single beam switching time and inter-node signaling delivery success rate; second, comparing these two fundamental indicators with corresponding preset thresholds; if the single beam switching time is less than a first preset threshold of the cache data prefetching period (e.g., less than half the cache data prefetching period), the current environment is deemed to meet the high mobility condition; if the inter-node signaling delivery success rate is less than a second preset threshold (e.g., less than 75%), the current environment is deemed to meet the weak coordination condition; if the environment meets the criteria for both high mobility and / or weak coordination, a multi-node redundancy distribution mechanism is triggered before the second terminal initiates a request.

[0040] In some preferred embodiments, the multi-node redundancy distribution mechanism of the present invention includes: determining multiple predicted cache nodes as target edge cache nodes based on the trajectory prediction data, and copying and distributing the cached data to the multiple target edge cache nodes for caching via a cloud platform. For the aforementioned environment of high mobility and weak inter-node coordination, the present invention no longer waits for terminal requests before performing single-point migration, but instead triggers the multi-node redundancy distribution mechanism in advance; based on the future multi-location distribution of the terminal obtained from the trajectory prediction model, multiple edge nodes with high probability of being traversed are selected as target cache nodes, and the cloud platform actively copies and distributes the cached data in parallel to multiple nodes for pre-caching. Redundant deployment offsets the migration delay and link risks caused by high movement speed and poor node coordination, enabling the terminal to hit the nearest cache at any switching location, ensuring service continuity and low latency.

[0041] This embodiment transforms the traditional passive single-point migration after a request into proactive multi-node pre-deployment before a request, effectively avoiding problems such as cache migration lag, link failure, and excessive access latency caused by high-speed terminal movement and poor node coordination. By covering the terminal's future movement area with redundant cache, it ensures that the cache can be hit nearby when the location changes, greatly improving the cache hit rate and service continuity. At the same time, the differentiated matching of dynamic scenario-specific migration strategies avoids invalid migration and resource waste, enhancing the robustness and environmental adaptability of the caching system.

[0042] Example 5

[0043] Based on any one of Embodiments 1 to 3, this embodiment provides another migration strategy that matches the corresponding migration strategy in the preset cache migration mechanism.

[0044] To address the problem that existing technologies, in edge node data transmission scenarios, still employ a single-dimensional triggering method, failing to comprehensively consider the degree of service correlation between terminals and the real-time quality of the network link, resulting in inaccurate collaborative transmission triggering, low transmission efficiency, and wasted network resources, thus hindering efficient collaborative transmission of cached data, this invention provides a superior migration strategy. This strategy includes: if the collaborative access similarity is greater than a third preset threshold and / or the spatiotemporal trajectory overlap is greater than a fourth threshold, then the current environmental characteristics are determined to meet the strong correlation condition. Specifically, collaborative access similarity refers to the overlap ratio of the first terminal and the second terminal accessing the same content within a historical period; if this ratio is greater than the third preset threshold, it is considered a strong correlation. Spatiotemporal trajectory overlap refers to the estimated duration for which two terminals will be within the coverage area of ​​the same edge node (or the same satellite beam) within a future time window, calculated based on the probability heatmap output by the trajectory prediction model; if the overlap duration exceeds a preset threshold (the fourth threshold), then the strong correlation condition is met.

[0045] If the available bandwidth margin is greater than the fifth preset threshold and / or the channel signal-to-noise ratio (SNR) is greater than the sixth preset threshold, the current environmental characteristics are deemed to meet the bandwidth optimization condition. Here, available bandwidth margin refers to the remaining available bandwidth of the communication links between edge nodes (such as inter-satellite links (ISL) or ground station relay links) monitored in real time. If the instantaneous available bandwidth is greater than the minimum transmission rate requirement of the data to be migrated within a preset completion time, i.e., the fifth preset threshold, then it is determined to be bandwidth-optimal. Considering the dynamic nature of satellite links, this invention monitors the link's SNR or bit error rate. If the channel SNR is consistently higher than the demodulation threshold (i.e., the sixth preset threshold) and remains stable, it is considered that the link bandwidth is optimal. If the environmental characteristics meet the strong correlation and / or bandwidth optimization criteria, the edge node P2P cooperative transmission mechanism is triggered.

[0046] In some other preferred embodiments, the edge node P2P cooperative transmission mechanism of the present invention includes: The cached data from the first edge cache node is transmitted to the target edge cache node determined based on the current location information of the second terminal for caching. When the P2P collaborative transmission mechanism is triggered, the system determines the current location information of the second terminal, identifies the corresponding target edge cache node based on the current location information (such as the edge node corresponding to the satellite beam currently in which the second terminal is located, or the edge node it will enter within a future time window in the predicted trajectory), and then transmits the cached data stored in the first edge cache node (such as content that the second terminal may request or content related to collaborative access by the first terminal) to the target edge cache node for pre-caching. In some preferred embodiments, the P2P collaborative transmission mechanism of the present invention further includes: during transmission, prioritizing the selection of communication links that meet the bandwidth advantage conditions for data transmission, and dynamically adjusting the transmission path and transmission rate according to the real-time link status.

[0047] This embodiment triggers the edge node P2P collaborative transmission mechanism by comprehensively judging strong correlation conditions and bandwidth optimization conditions, overcoming the shortcomings of single-dimensional judgment in the prior art, avoiding invalid or inefficient transmission, and improving the accuracy of migration strategy triggering and network resource utilization efficiency. At the same time, after triggering, the cached data of the first edge cache node is transmitted to the target edge cache node corresponding to the current location of the second terminal for pre-caching. During the transmission process, communication links that meet the bandwidth optimization conditions are selected first for data transmission. The transmission path and rate are dynamically adjusted according to the real-time link status, effectively reducing data access latency, improving the transmission success rate in extreme satellite link environments, enhancing system robustness, and realizing intelligent collaborative transmission with cross-terminal service awareness.

[0048] Example 6

[0049] Based on any one of Embodiments 1 to 3, this embodiment provides another migration strategy that matches the corresponding migration strategy in the preset cache migration mechanism.

[0050] To address the problems in existing technologies where traditional full cache data migration methods are still used in scenarios with limited storage at edge nodes and extreme satellite link environments (high packet loss), resulting in significant consumption of node storage resources and link bandwidth during the migration process, and the high failure rate and increased latency under extreme link conditions, thus failing to achieve efficient access to cached data, the present invention provides a preferred migration strategy. This strategy includes: First, extracting core basic indicators of the current network and nodes, including the remaining storage ratio of the first edge cache node and the satellite link packet loss rate; second, comparing the two basic indicators with corresponding preset thresholds. If the remaining storage ratio of the first edge cache node is less than a seventh preset threshold (e.g., less than 15%), the current environment is deemed to meet the storage-limited condition, meaning the first edge cache node lacks sufficient storage resources to support cached data or complete data migration; if the satellite link packet loss rate is greater than an eighth preset threshold (e.g., greater than 10%), the current environment is deemed to meet the extreme environment condition, indicating extremely poor satellite link transmission stability and a high risk of data loss and stuttering during migration; if the environmental characteristics meet the storage-limited and / or extreme environment conditions, a cloud platform resource routing pointer redirection mechanism is triggered.

[0051] In some other preferred embodiments, the cloud platform resource routing pointer redirection mechanism of the present invention includes: using the cloud platform to generate a metadata index pointing to the first edge cache node at the target edge cache node. This metadata index only contains lightweight data such as the location information and identification information of the cached data, and will not occupy too much storage and link resources. When the second terminal initiates a request, the target edge cache node quickly locates the first edge cache node and obtains the corresponding cached data through the metadata index. The cached data is obtained from the first edge cache node through the metadata index without the need to migrate the complete cached data, thereby adapting to the needs of storage and extreme link scenarios.

[0052] This embodiment identifies storage-constrained and extreme environment scenarios, avoiding the resource waste and migration failure issues of traditional full data migration in such scenarios. It adopts a cloud platform resource routing pointer redirection mechanism, replacing full cached data migration with lightweight metadata indexing, which significantly reduces node storage occupation and satellite link transmission pressure, adapting to edge nodes with tight storage resources and extreme link environments with high packet loss. At the same time, it quickly locates cached data through metadata indexing, ensuring that requests from the second terminal can be responded to efficiently. While saving resources, it avoids access failures and high latency issues caused by extreme environments. This embodiment further improves the scenario coverage of the preset cache migration mechanism.

[0053] Example 7

[0054] Based on any one of Embodiments 1 to 3, this embodiment provides another migration strategy that matches the corresponding migration strategy in the preset cache migration mechanism.

[0055] To address the problem in existing technologies where conventional cache migration methods are still used when environmental characteristics meet the link interruption criteria, which cannot handle the inability to transmit cached data or migration failures caused by link interruptions, thus resulting in untimely responses and access interruptions to second terminal requests, the present invention employs the following migration strategy in some preferred embodiments: First, real-time monitoring of the current network link status is performed to extract link-related basic indicators. Based on preset judgment criteria, it is determined whether the environmental characteristics meet the link interruption criteria. In some preferred embodiments, the link interruption determination can be based on a comprehensive assessment of indicators such as link connection status and data transmission success rate to ensure accuracy. This can also be a reasonable design by those skilled in the art based on actual conditions or site requirements. If the environmental characteristics meet the link interruption criteria, when the second terminal makes a request, the traditional cache migration process is no longer executed, as the traditional process is prone to migration failure due to link interruption. Instead, the acquired cached data is re-sent to the target edge cache node determined based on the current location information of the second terminal via the cloud platform for caching, ensuring that the second terminal can successfully obtain the cached data to respond to the request.

[0056] This embodiment can identify link interruption scenarios, avoiding the problems of inability to migrate normally or migration failure when the link is interrupted. When the link is interrupted and the second terminal initiates a request, the cached data is redistributed to the target edge cache node corresponding to the current location through the cloud platform, bypassing the limitations of the interrupted link, ensuring that the cached data can be deployed smoothly, guaranteeing the timely response to the second terminal's request, and avoiding access interruption. At the same time, this embodiment does not require complex link repair or additional migration adaptation operations, the process is simple and efficient, further improving the scenario coverage capability of the preset cache migration mechanism, enhancing the robustness and reliability of the cache system under link abnormality scenarios, and ensuring that the terminal can still obtain stable cache services under link interruption scenarios.

[0057] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. An edge cache migration method based on cross-terminal trajectory prediction, characterized in that, Includes the following steps: S1. Collect the current request logs of the first terminal and the second terminal, input the trained trajectory prediction model, and obtain the trajectory prediction data of the second terminal within a preset time period; S2. Obtain the predicted request content from the second terminal and construct cached data; The cached data is sent via the cloud platform to the first edge cache node determined based on the original location information of the second terminal for caching; S3. When the second terminal initiates a request, determine whether the current location has shifted relative to the original location information; If yes, then based on the environmental features extracted from the confidence level of the trajectory prediction model, the satellite link quality, and the storage status of the edge node, the corresponding migration strategy is matched in the preset cache migration mechanism to determine the target edge cache node and obtain the actual requested content; if no, then the actual requested content is obtained from the first edge cache node.

2. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 1, characterized in that, The training method for the trajectory prediction model includes: Historical request logs from multiple terminals are collected, and individual features, cooperative movement features, and basic state features of the first and second terminals are extracted from the historical request logs as a training set to train the trajectory prediction model constructed through graph convolutional networks, temporal convolutional networks, and attention mechanisms.

3. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 1, characterized in that, The method for determining whether the second terminal has migrated includes: calculating the distance between the current location and the original location; if the distance is less than a preset migration threshold, then the second terminal is determined not to have migrated; otherwise, the second terminal is determined to have migrated.

4. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 1, characterized in that, Based on the environmental characteristics, a corresponding migration strategy is matched in the preset cache migration mechanism. Specific methods include: If the single beam switching time is less than the first preset threshold of the cached data prefetch period, the current environment is determined to meet the high mobility condition; if the inter-node signaling delivery success rate is less than the second preset threshold, the current environment is determined to meet the weak coordination condition; if the environment meets the determination conditions of high mobility and / or weak coordination, a multi-node redundancy distribution mechanism is triggered before the second terminal initiates a request.

5. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 4, characterized in that, The multi-node redundancy distribution mechanism includes: determining multiple prediction cache nodes as target edge cache nodes based on the trajectory prediction data, and copying and distributing the cached data to the multiple target edge cache nodes for caching through the cloud platform.

6. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 1, characterized in that, Based on the environmental characteristics, a corresponding migration strategy is matched in the preset cache migration mechanism. Specific methods include: If the collaborative access similarity is greater than the third preset threshold and / or the spatiotemporal trajectory overlap is greater than the fourth threshold, then the current environmental features are determined to meet the strong correlation condition; if the available bandwidth margin is greater than the fifth preset threshold and / or the channel signal-to-noise ratio is greater than the sixth preset threshold, then the current environmental features are determined to meet the bandwidth advantage condition; if the environmental features meet the determination conditions of strong correlation and / or bandwidth advantage, then the edge node P2P collaborative transmission mechanism is triggered.

7. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 6, characterized in that, The edge node P2P collaborative transmission mechanism includes: The cached data of the first edge cache node is transmitted to the target edge cache node determined based on the current location information of the second terminal for caching.

8. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 1, characterized in that, Based on the environmental characteristics, a corresponding migration strategy is matched in the preset cache migration mechanism. Specific methods include: If the remaining storage ratio of the first edge cache node is less than the seventh preset threshold, the current environmental characteristics are determined to meet the storage-limited condition; if the satellite link packet loss rate is greater than the eighth preset threshold, the current environmental characteristics are determined to meet the extreme environmental condition; if the environmental characteristics meet the determination conditions of storage-limited and / or extreme environment, the cloud platform resource routing pointer redirection mechanism is triggered.

9. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 8, characterized in that, The cloud platform resource routing pointer redirection mechanism includes: The cloud platform is used to generate a metadata index pointing to the first edge cache node on the target edge cache node, and the cached data is obtained from the first edge cache node through the metadata index.

10. The edge cache migration method based on cross-terminal trajectory prediction as described in claim 1, characterized in that, Based on the environmental characteristics, a corresponding migration strategy is matched in the preset cache migration mechanism. Specific methods include: If the environmental characteristics meet the criteria for link interruption, then when the second terminal makes a request, the cached data is resent to the target edge cache node determined based on the current location information of the second terminal through the cloud platform for caching.

Images