Cloud-side collaborative content caching and user association method and system

Abstract

The invention provides a cloud-side collaborative content caching and user association method and system, and relates to the technical field of wireless communication networks, and the method comprises the steps: building a weighted delay minimization problem with minimization of user content acquisition delay and association base station switching delay as a target; and dividing the constructed weighted delay minimization problem into a long-term user association and content caching problem under a long-term time scale and a short-term user association problem under a short-term time scale to obtain a content caching strategy and a user association strategy so as to minimize the content transmission delay between the user and the base station. According to the invention, the content transmission delay can be reduced on the basis of reducing the backhaul link burden, and the network performance is improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication network technology, and in particular to a cloud-edge collaborative content caching and user association method and system. Background Art

[0002] With the rapid development of the Internet of Things and 5G technology, the surge in mobile traffic has put significant pressure on the backhaul link. This pressure stems from the growing demand of users for high-bandwidth applications and real-time services. The existing network architecture faces many challenges in dealing with these demands, including the surge in device connections, the guarantee of low latency and high reliability, and the ability to flexibly adapt to changes in the network environment. Mobile edge caching technology has been proposed to alleviate this problem. By caching content at the base station, repeated content transmission is avoided, thereby effectively reducing the burden on the backhaul link. However, user mobility and interest changes make the optimization of content caching strategies and user association strategies complicated. The dynamic needs of users in a high-mobility environment and the rapid changes in their interests make it difficult for edge cached content to meet user needs, thereby affecting network performance. Summary of the invention

[0003] In order to address the deficiencies of the prior art, the purpose of the present invention is to provide a cloud-edge collaborative content caching and user association method and system, so as to reduce content transmission delay and improve network performance on the basis of alleviating the burden on the return link.

[0004] In order to achieve the above object, according to some embodiments, a first aspect of the present invention provides a method for associating cloud-edge collaborative content caching with users, including: The weighted delay minimization problem is constructed with the goal of minimizing the delay in user content acquisition and the delay in associated base station switching. The constructed weighted delay minimization problem is divided into a long-term user association and content caching problem in a long-term time scale and a short-term user association problem in a short-term time scale, and the content caching strategy and user association strategy are obtained to minimize the content transmission delay between the user and the base station. The content caching strategy includes predicting the content required by the user and caching it based on the bidirectional long short-term memory network according to the user's content request history; The user association strategy includes an online user association strategy and an inert user re-association strategy. The online user association strategy includes, according to a content caching strategy, associating a user with a base station corresponding to a sub-channel with a minimum download delay; the inert user re-association strategy includes, for each pair of users with different associated base stations after association, calculating the delay after the base station is exchanged, and if the delay is reduced after the exchange, exchanging the base stations associated with the two users.

[0005] A second aspect of the present invention provides a cloud-edge collaborative content caching and user association system, including: A problem construction module is configured to construct a weighted delay minimization problem with the goal of minimizing a delay in user obtaining content and a delay in switching an associated base station; The content caching and user association module is configured to divide the constructed weighted delay minimization problem into a long-term user association and content caching problem under a long-term time scale and a short-term user association problem under a short-term time scale, and obtain a content caching strategy and a user association strategy to minimize the content transmission delay between the user and the base station; the content caching strategy includes predicting the user's required content based on a bidirectional long short-term memory network and caching it according to the user's content request history; the user association strategy includes an online user association strategy and an inert user re-association strategy, and the online user association strategy includes associating the user with the base station corresponding to the sub-channel with the minimum download delay according to the content caching strategy; the inert user re-association strategy includes, for each pair of users with different associated base stations after association, calculating the delay after the base station exchange, and if the delay is reduced after the exchange, then exchanging the base stations associated with the two users.

[0006] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a cloud-edge collaborative content caching and user association method and system, which predicts the user's required content based on a bidirectional long short-term memory network according to user association information and the user's historical content requests, realizes active caching of content by the base station, improves the association between cached content and users, ensures that users can quickly and efficiently obtain the required content in future time windows, and effectively reduces content access delays. Through online user association strategies and lazy user re-association strategies, the association between users and base stations is optimized when user behavior changes, while paying attention to the needs of individual users, taking into account global network performance, thereby reducing overall content download delays and further improving network efficiency and user experience.

[0007] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0009] Figure 1 Schematic diagram of the method in Embodiment 1 of the present invention; Figure 2 Schematic diagram of a network architecture corresponding to the method in Embodiment 1 of the present invention; Figure 3 A schematic diagram of a content cache policy and a user association policy update process; Figure 4 A schematic diagram of predicting user-required content based on a bidirectional long short-term memory network; Figure 5 Schematic diagram of the active edge caching strategy. DETAILED DESCRIPTION

[0010] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0011] Embodiment 1 Embodiment 1 of the present invention provides a method for associating cloud-edge collaborative content caching with users, such as Figure 1-Figure 5 As shown, including: The weighted delay minimization problem is constructed with the goal of minimizing the delay in user content acquisition and the delay in associated base station switching. The constructed weighted delay minimization problem is divided into a long-term user association and content caching problem in a long-term time scale and a short-term user association problem in a short-term time scale, and the content caching strategy and user association strategy are obtained to minimize the content transmission delay between the user and the base station. The content caching strategy includes predicting the content required by the user and caching it based on the bidirectional long short-term memory network according to the user's content request history; The user association strategy includes an online user association strategy and an inert user re-association strategy. The online user association strategy includes, according to a content caching strategy, associating a user with a base station corresponding to a sub-channel with a minimum download delay; the inert user re-association strategy includes, for each pair of users with different associated base stations after association, calculating the delay after the base station is exchanged, and if the delay is reduced after the exchange, exchanging the base stations associated with the two users.

[0012] The method provided in this embodiment implements active caching of content based on a bidirectional long short-term memory (BiLSTM) network to predict content requests and optimize caching strategies. At the same time, a dynamic user association strategy is introduced, including an online user association strategy and an inert user reassociation strategy, in order to optimize the association between users and base stations when user behavior changes. The simulation results verify that the solution significantly improves network performance and user experience, and provides an effective solution to cope with the growing traffic and diverse user needs.

[0013] First, if Figure 1As shown, the scenario set in this embodiment includes a central server (cloud server), densely deployed edge nodes and a user layer. The edge node EN can request content from the server and cache it locally. By deploying multiple types of edge nodes, such as base stations BS, roadside units and / or wireless access points with wireless interfaces and edge caching functions, an efficient edge caching system can be implemented. The user layer includes all users equipped with limited cache capacity. In the scenario considered, the service range of the edge node is overlapping. It is assumed that the frequency subchannels assigned to different edge nodes are orthogonal, so there is no need to consider the interference between edge nodes. The available bandwidth of the edge node is Bm, which is divided into P subchannels, and the bandwidth of each subchannel is bm=Bm / P. It is assumed that the sub-bandwidth required to transmit user data between edge nodes is the same as the size of the user data. The central server CS and the edge node EN together form a stable edge cloud cache network. When the cache of EN is lost, the user can directly obtain the required content from the central server. In this embodiment, it is assumed that the user's location movement is known.

[0014] Time is evenly divided into a series of equal-length time slots, denoted as Γ = {1, 2, …, t, …, T}. Each time window (the set of time windows is denoted as Γ′ = {1, 2, …, j, …, J}) has the same number of time slots. The edge node will implement content caching decisions based on the content request in the first time slot of each time window. Figure 2 The network has an indivisible , which is the set of all data or files that a user may request in the network (denoted as ). Assume that the storage capacity of the central server is unlimited, but the edge nodes can only cache a small part of the content. Let the binary variable is the mth edge node EN m The content cache variable of Indicates EN m Content f has been cached in time window j. Each edge node has a capacity limit Sm. Some users can access content by connecting to adjacent edge nodes. express.

[0015] In addition, the content acquisition delay is calculated based on cache hits and misses. A cache hit means that the content requested by the user is cached in the edge node associated with the user, while a cache miss means that the content requested by the user is not cached in the associated edge node and needs to be obtained from the central server. Therefore, user u obtains EN on subchannel p m The delay of the content's time slot t (the smallest unit in time division, used to update the system status or perform some short-cycle operations) for: ; The frequency subchannels allocated to different base stations are orthogonal, and the available bandwidth is divided into P subchannels. is user u in EN m The delay time of local acquisition of content f, where L represents local transmission, is the size of content f; user u receives data from EN via subchannel p in time slot t m The rate of acquiring content f is User u from EN m The rate at which content f is acquired can be calculated using Shannon's formula: ; in, represents the bandwidth of subchannel p, Indicates the channel ratio of the subchannel.

[0016] represents the user's transmission from the central server CS to EN within time window j (a larger time interval that can contain multiple time slots). m The delay time to obtain the required content f, where Represents the transmission from the central server to the edge node. In the expression, the first part on the right side of the equal sign represents the m The delay in getting the content, the second part indicates that the content requested by the user is not in EN m In the process, the request is forwarded to the central server CS to obtain the file.

[0017] In addition, assume that the user sends a request in each time window. Let the content request matrix of all user requests be ,in indicates that user u requests content f in time window j, Indicates that there is no request for content f in time window j. In addition, the delay for all users to obtain content in time window j is: ; In addition, in order to obtain the required content, each user needs to associate with the appropriate edge node in each time slot. An effective user association strategy can not only reduce the delay of users obtaining content, but also improve the content caching strategy. It is used to indicate whether user u transmits with EN in time slot t. m EN on subchannel p m association.

[0018] In dynamic scenarios, user reassociation is inevitable due to the diversity of user request content and user mobility. This reassociation may cause service interruption or degradation of quality of service (QoS). Assume that a user is at the mth edge node EN m and the nth edge node EN n The switching delay between , switching delay It can be considered as part of the transmission time (describing the opportunity cost of switching, emphasizing that the switching process will take up time that could have been used to transmit data, thereby increasing the user's total delay). Definition is the corresponding associated EN set, where the elements Let t denote the base station associated with user u in time window j. Let t denote the last time slot of the previous time window j-1. Therefore, the total handover delay in time window j can be expressed as the product of the handover delay of the associated base station in the previous time window j-1 and the associated base station in the current time window, expressed as: ; In this scenario, to improve network performance and satisfy users’ content requests, the goal is to minimize the combined impact of the acquisition delay when downloading content from edge nodes or central servers and the switching delay when users reconfigure the network.

[0019] Optimizing content caching strategies through federation Associating policies with users The corresponding weighted delay minimization problem Problem1 is given by: ; ; ; ; ; ; in, is the balance coefficient; constraint C1 means that the size of the content cached by the edge node does not exceed its own capacity. represents the size of content f; constraint C2 represents whether the edge node caches content f; constraint C3 represents whether user u communicates with EN through subchannel p m association; constraint C4 indicates that a user is associated with at most one edge node; constraint C5 indicates that EN m Associated with at most |P| users (EN m It occupies at most P sub-channels and is associated with different users, that is, it is associated with at most |P| users).

[0020] In this embodiment, the content caching strategy refers to pre-storing or dynamically storing certain content at the edge of the network (such as a base station, edge server, or edge node of a content distribution network) to reduce the transmission overhead and latency of obtaining content from the core network; the user association strategy refers to deciding which base station or access point each user should connect to, and using sub-channels in a cellular network or a heterogeneous network, in order to optimize network performance and user experience.

[0021] This problem is classified as an NP-hard problem, so this embodiment also proposes a low-complexity content caching strategy and user association framework to find a suboptimal solution. The problem is decomposed into two low-complexity sub-problems: content caching strategy and user association strategy.

[0022] Content caching strategy and user association strategy can be implemented at different time scales. In practice, the user association strategy affects the content caching strategy of the next time window, while the long-term content caching strategy affects the user association strategy within each time window. In addition, within time window j, there is a coupling between the content caching strategy C and the user association strategy A, and there is an intrinsic relationship between the user association strategies of two consecutive time windows in the objective function, which reflects the impact of past decisions or states on the current state. Traditional methods rarely solve the problem effectively. To solve this problem, the problem is decomposed into two different time scales: Long-term user association and content caching at long-term time scales: Content requests are usually associated with long-term time scales. Therefore, content caching strategies can be formulated by analyzing historical request information. Generally speaking, at the beginning of each time window j, the content caching strategy is determined based on the user association information and content requests in the previous time window j-1. In this way, the caching strategy can be dynamically adjusted to ensure that users can quickly and efficiently obtain the required content in future time windows. In addition, this strategy can effectively reduce the latency of content access, improve the overall performance of the system, and enhance the user experience. At the same time, using historical data for prediction and optimization can better adapt to changes in user behavior and fluctuations in demand, thereby improving the accuracy and reliability of content distribution. The switching delay is related to the user association strategy of two consecutive time windows and is affected by the EN content caching strategy of the time window to which the time window belongs. In each time window j, the user is associated with the EN in the previous time window j-1, and the switching delay is minimized by switching the association strategy between different users.

[0023] Short-term user association at short time scales: User association strategies are crucial in MEC (Mobile Edge Computing), especially in highly dynamic environments where user locations and demands are constantly changing. Short-term user association strategies are used to play an important role in quickly responding to user requests and optimizing instant network performance by selecting the best subchannel.

[0024] Based on this, this embodiment proposes a BiLSTM-based active caching framework for predicting content popularity and learning historical request information. On this basis, by obtaining the content caching strategy to determine the association of users in each time window, an online user association strategy is designed. In order to prevent the online user association strategy from falling into the local optimum, this embodiment also introduces an inert user re-association strategy based on the time window to better balance the load distribution between base stations. The specific process is as follows: Figure 3 shown. Figure 3 A dynamic optimization framework combining content caching strategy and user association strategy is described. Arrows indicate the direction of data flow and strategy association. In the blue module on the left, through data preprocessing, BiLSTM model training and feature extraction, future user content needs are predicted and content caching strategies within the time window are formulated. The arrow connects to the orange module on the right, and the prediction results are used for user pre-association and dynamic association adjustment to achieve long-term user association and content caching on a long-term time scale. At the same time, combined with the real-time monitoring of content cache status, user request status and sub-channel status, an online algorithm is used to optimize the association strategy between users and base stations to achieve short-term user association on a short-term time scale. The whole process is regularly updated within the time window, and a loop is formed through arrow feedback to dynamically adjust the strategy to improve system performance.

[0025] for Figure 3 The long-term user association and content caching strategy on the long-term time scale with the time window as the scale on the left, including the content caching strategy and the lazy user re-association strategy. For the content caching strategy, data preprocessing is performed first: before starting modeling, relevant data in the network (such as user requests, content access frequency, network status, etc.) is collected and preprocessed. This step usually includes operations such as data cleaning and standardization. The output of this step is a prediction of future user content demand, indicating which content needs to be cached in the future time window.

[0026] Feature extraction is then performed: by analyzing historical data and model results, BiLSTM is used to extract features that have an important impact on caching decisions, such as content access frequency and period fluctuations, to form a predicted request probability matrix. Finally, based on the predicted request probability matrix, the content that needs to be cached is determined according to the user's request probability for the content. For example, the content with the highest request probability can be cached.

[0027] Lazy user reassociation strategy, including pre-association and lazy re-association. Pre-association: With the support of content caching strategy, predict user needs in advance and determine the user's best possible associated base station or sub-channel. For example, if the content required by the user has been cached to a certain edge node, then the user is associated with this node first.

[0028] Lazy reassociation: Dynamically adjust the user's association based on real-time network status (such as base station load, user demand changes, etc.). If the current association cannot meet user needs (such as insufficient bandwidth or high latency), reselect a better association path. "Lazy" here refers to reducing unnecessary frequent switching to avoid increasing system overhead or latency.

[0029] for Figure 3 The short-term user association strategy on the right side at the short-term time scale based on time slots includes real-time monitoring of system status and online user association based on online algorithms.

[0030] Real-time monitoring is implemented based on the real-time monitoring module. The real-time monitoring module is used to describe the distribution and status of the stored content in the edge node, such as which content is cached, the hit rate, whether there is expired data, etc. It provides real-time information of cache resources for online algorithms and assists in optimizing the association between users and cache nodes.

[0031] The online algorithm is implemented based on the online algorithm module, which is used to receive information from the content cache status, user request status and sub-channel status, comprehensively consider these real-time data, and dynamically optimize the association between users and the network.

[0032] BiLSTM model training: Use a bidirectional long short-term memory network (BiLSTM) to train historical data, especially time series data. The BiLSTM network can simultaneously use past and future information to capture the dependencies between user requests.

[0033] A user-centric approach is adopted, which considers the request time pattern and uses a BiLSTM network to capture the user's content requests that change over time. The BiLSTM is used to extract time series features, which are then further processed by a fully connected network (FCN) to generate the final output. Figure 4 As shown in Figure 1, the entire process can be summarized as follows: collect the user's content request history and convert it into a content request matrix R. Use BiLSTM to extract time series features; apply a fully connected network to further process the extracted time series features; generate a predicted request probability matrix to guide the optimization of the cache strategy.

[0034] The input data of the model is the user's content request history, represented in the form of time series. These data contain information such as request timestamps and content identifiers, which are used to capture the pattern of user requests. The content request matrix R is a representation matrix of the user's content request history. The rows of the matrix represent different content and the columns represent time windows. BiLSTM (Bidirectional Long Short-Term Memory Network) is used to process time series data and can capture the time dependency of user requests. The bidirectional feature enables it to learn past and future contextual information at the same time, thereby more comprehensively modeling the user's request pattern. The feature vector output by BiLSTM is input into the fully connected network (FCN) for further feature combination and nonlinear mapping. The fully connected network can generate higher-level feature representations to output more accurate prediction results. The final output is a predicted request probability matrix, which represents the probability of users requesting different content in a specific time window in the future.

[0035] The content caching strategy and the user association strategy achieve collaborative optimization through information interaction. First, the content caching strategy allocates frequently accessed content to appropriate edge nodes based on the user's content request history and content popularity prediction to generate a "content cache state". Based on this, the user association strategy dynamically optimizes the connection relationship between the user and the base station or cache node in combination with the real-time user request status and sub-channel status to minimize access delays and network loads. At the same time, the results of user association will be fed back to the cache strategy, such as cache hit rate and content popularity distribution, as an important basis for the next time window cache decision. This closed-loop dynamic adjustment mechanism ensures an efficient match between cache distribution and user needs, thereby continuously improving system performance in a changing network environment.

[0036] User association determines the user group and content requirements of edge node services, while the cache strategy improves hit rate and reduces latency by caching user-common content. The two interact with each other, changes in user association affect the distribution of cached content, and cache strategy optimization can improve user experience and association quality, achieving a balance between resource utilization and service performance through collaborative optimization.

[0037] In this regard, this embodiment further proposes an active edge caching strategy based on user access probability, which allows offline content to be placed in the cache and replace online content, thereby achieving caching of popular content and replacement of outdated content. The specific process is as follows: Figure 5 shown.

[0038] When the edge node EN receives a user's request for a certain content f, it determines whether the requested content f has been cached in EN. If so, it directly returns the content f to the user and updates the access probability of the content in EN. The access probability represents the change in the popularity of the content. If the content f does not exist in EN, EN obtains the content f from the central server CS, returns it to the user, and calculates the access probability of the content f. Subsequently, it checks whether the cache space of EN is full. If the cache space is not full, the content f is directly put into the cache; if the cache space is full, some content in the cache needs to be replaced. At this time, the content with the lowest access probability in the cache is searched for replacement to ensure that the new content can be cached. Finally, the EN cache space status is updated. Whether it is directly put into the cache or the old content is replaced, the cache status in EN will be updated in the end to ensure that the latest content storage is recorded.

[0039] In an MEC network with multiple ENs, users need to select the most appropriate EN for association within the coverage of multiple ENs. Therefore, this embodiment proposes a dynamic user association strategy, including an online user association strategy and an inert user re-association strategy.

[0040] The online user association policy includes the following steps: Step S101: Screening sub-channels that meet the user's minimum data rate requirement.

[0041] In step S101, according to the minimum data rate requirement of each user, a sub-channel that meets the conditions is screened out from a group of candidate sub-channels. The specific steps are as follows: S1011: Initialize the screening set. Create an empty set to represent all sub-channels that meet the user's minimum data rate requirement.

[0042] S1012: Traverse all available sub-channels. For each available sub-channel, calculate the transmission rate of the sub-channel. The rate is usually affected by multiple factors, including channel quality, network load, interference and other factors.

[0043] S1013: Rate condition determination. If the transmission rate of a subchannel meets or exceeds the user's minimum rate requirement, the subchannel is considered to meet the condition. These subchannels that meet the condition are added to the screening set as candidate subchannels.

[0044] Through this step, we screen out some sub-channels that meet the user rate requirements from all sub-channels, laying the foundation for the subsequent processing steps.

[0045] Step S102: Calculate the download delay of each sub-channel.

[0046] In step S102, according to the delay performance of each screened sub-channel, the sub-channel that best suits the current user needs is selected. The specific steps are as follows: S1021: Initialize delay calculation. Set an initial value to represent the currently known minimum delay, and set the initial value to infinity. Also set an identifier to represent the current optimal subchannel.

[0047] S1022: Traverse the selected sub-channels. For each selected sub-channel, calculate the download delay of the sub-channel. The calculation of the download delay takes into account various network factors, such as transmission distance, channel quality, network load, etc.

[0048] S1023: Delay comparison. Compare the delay of the current sub-channel with the recorded minimum delay. If the delay of the current sub-channel is lower, update the recorded minimum delay and mark the sub-channel as the current optimal sub-channel.

[0049] S1024: Selection that meets the conditions. Only when the delay meets the user's tolerance range, the sub-channel is considered as an available candidate sub-channel, and the subsequent comparison and selection are continued.

[0050] Through this step, the algorithm can select the most suitable sub-channel according to the delay performance, thereby improving the user experience and avoiding the impact of excessive delay on service quality.

[0051] Step S103: Associating the user with the optimal sub-channel.

[0052] In step S103, the user is associated with the optimal sub-channel according to the calculation results of steps S101 and S102. The specific steps are as follows: S1031: Check whether a suitable sub-channel is found. If a sub-channel with the smallest delay and meeting the user's requirements is successfully found in step S102, then the sub-channel is regarded as the optimal sub-channel.

[0053] S1032: Complete user association. Associate the current user with the optimal sub-channel. At this point, the user will be assigned to the sub-channel for communication.

[0054] S1033: Update status and end. After the association between the user and the sub-channel is completed, the association status is recorded, indicating that the user has been successfully assigned to the appropriate sub-channel. At the same time, if multiple users request association, the process will continue until all users are assigned to the appropriate sub-channel.

[0055] Through this step, the algorithm ensures that each user can obtain service on the most appropriate sub-channel, thereby maximizing the utilization efficiency of network resources and user experience.

[0056] In the steps of the above-mentioned online user association strategy, if the EN can meet the user's minimum data rate requirement, the user will be immediately associated with the EN, thereby quickly realizing the association of the mobile user with the EN. However, this method of online user association only focuses on the association of a single user. Once the system makes an association decision, it may cause the average content download delay of the entire network to increase. In order to avoid local optimality, this embodiment further proposes an inert user re-association strategy. This strategy aims to achieve more efficient user association by re-evaluating and adjusting the user-EN association after the initial decision, so as to optimize the performance of the entire network. This method not only focuses on the needs of a single user, but also takes into account the global network performance, thereby reducing the overall average content download delay and achieving better user experience and network efficiency.

[0057] The lazy user reassociation strategy optimizes overall resource allocation by delaying association decisions and evaluating global network performance. This approach not only considers the needs of individual users, but also balances the load between multiple users, reduces potential network bottlenecks, and improves overall network efficiency and user experience. By re-evaluating and adjusting user association status within the appropriate time window, better network performance and resource utilization can be achieved.

[0058] Specifically, for each user u, if the EN associated in the previous time window can satisfy its request, the same association is maintained. This step reduces unnecessary re-associations and reduces switching delays. (Prevent the algorithm from falling into an infinite loop or wasting time unnecessarily when computing resources are limited, which is equivalent to tolerating errors so that the algorithm terminates early when the accuracy requirements are met). For each pair of different mobile devices with different associated base stations, the post-switch delay is calculated and the user association strategy is updated. The resources of each base station and subchannel are limited. By exchanging the association of user equipment with base stations, these resources can be better utilized to reduce the overload of specific base stations, and the performance of the entire system will be improved.

[0059] The lazy user reassociation strategy includes the following steps: Step S201: Initialization and setting of initial user association policy.

[0060] Step S202: Iteratively optimize the user association strategy.

[0061] In step S201, the system is initialized and an initial user association strategy is set based on the current network status and user needs. The specific steps are as follows: S2011: Initialize association policy parameters. Set the initial delay indicators used to measure the current network performance, including core network delay and access network delay, which are initialized to zero. Also initialize the user's association status and historical association records.

[0062] S2012: Initially associate users. According to the user set, base station set, sub-channel set, and content request matrix, an initial associated base station and sub-channel are assigned to each user. The specific association principles are as follows: If the user's current transmission rate meets the minimum rate requirement and the user's service status has not changed within the current time window (ie, its associated base station is the same as the previous base station), the user's association strategy remains unchanged.

[0063] If the user's transmission rate cannot meet the minimum rate requirement, or its service status changes, it is necessary to reallocate the user's association strategy based on a comprehensive evaluation of rate and latency.

[0064] Through this process, the algorithm can quickly set an initial association strategy for each user, providing a basis for subsequent optimization.

[0065] Step S202 is the core part of the algorithm, which further reduces network delay and ensures the user's service quality by iteratively optimizing the user association strategy. The specific steps are as follows: S2021: Iteration process when the delay has not converged. During the optimization process, the loop is controlled according to the delay convergence state and the maximum number of iterations. When the network delay does not reach the convergence condition and the number of iterations does not exceed the maximum set value, the algorithm will continue to execute step S2022.

[0066] S2022: Adjust the association strategy for each pair of users. Perform the following evaluation on each pair of users in the user set (e.g., user u and user v): Communication range determination: Check whether the two users are associated with different base stations, and determine whether the two base stations are within the communication range. If so, proceed to the next step; Delay evaluation of exchange policy: Calculate the network delay caused by the exchange of association policies between user u and user v, and compare it with the current delay; If the delay after switching is less than the current delay, it indicates that the switching strategy is beneficial to optimizing network performance, and the association strategy between user u and user v is adjusted. Otherwise, the existing strategy remains unchanged.

[0067] Update iteration status: After completing the current iteration, record the updated user association policy, recalculate the latency indicators of the core network and access network, and increase the iteration count.

[0068] Iteration termination condition: When the delay reaches the convergence condition or the number of iterations exceeds the maximum set value, the iteration process ends.

[0069] S2023: Output the user association strategy. After the iterative optimization is completed, the user association strategy for the current time window is generated based on the final calculated delay result, and the result is output for network system application. The final user association strategy can ensure the following goals: minimize the user's re-association operation to avoid service interruption caused by frequent switching; optimize network delay performance to improve user experience; dynamically adapt to changes in user needs and network conditions to improve network resource utilization.

[0070] Embodiment 2 This embodiment provides a cloud-edge collaborative content caching and user association system, including: A problem construction module is configured to construct a weighted delay minimization problem with the goal of minimizing a delay in user obtaining content and a delay in switching an associated base station; The content caching and user association module is configured to divide the constructed weighted delay minimization problem into a long-term user association and content caching problem under a long-term time scale and a short-term user association problem under a short-term time scale, and obtain a content caching strategy and a user association strategy to minimize the content transmission delay between the user and the base station; the content caching strategy includes predicting the user's required content based on a bidirectional long short-term memory network and caching it according to the user's content request history; the user association strategy includes an online user association strategy and an inert user re-association strategy, and the online user association strategy includes associating the user with the base station corresponding to the sub-channel with the minimum download delay according to the content caching strategy; the inert user re-association strategy includes, for each pair of users with different associated base stations after association, calculating the delay after the base station exchange, and if the delay is reduced after the exchange, then exchanging the base stations associated with the two users.

[0071] It should be noted here that the various modules in this embodiment correspond one-to-one to the steps of the method in Example 1, and the specific implementation process is the same, which will not be repeated here.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for associating cloud-edge collaborative content caching with users, characterized in that: include: The weighted delay minimization problem is constructed with the goal of minimizing the delay in user content acquisition and the delay in associated base station switching. The constructed weighted delay minimization problem is divided into a long-term user association and content caching problem in a long-term time scale and a short-term user association problem in a short-term time scale, and the content caching strategy and user association strategy are obtained to minimize the content transmission delay between the user and the base station. The content caching strategy includes predicting the content required by the user and caching it based on the bidirectional long short-term memory network according to the user's content request history; The user association strategy includes an online user association strategy and an inert user re-association strategy, wherein the online user association strategy includes associating the user with a base station corresponding to a sub-channel having a minimum download delay according to a content caching strategy; The lazy user re-association strategy includes, for each pair of users with different associated base stations after association, calculating the delay after the base station exchange, and if the delay is reduced after the exchange, then exchanging the base stations associated with the two users.

2. A cloud-edge collaborative content caching and user association method as described in claim 1, characterized in that: The online user association strategy specifically includes: Screening the sub-channels that meet the user's minimum data rate requirement to obtain candidate sub-channels; Calculate the download delay of each candidate subchannel, and mark the subchannel with the smallest download delay as the optimal subchannel; Associate the user with the optimal sub-channel.

3. A cloud-edge collaborative content caching and user association method as described in claim 1, characterized in that: The lazy user re-association strategy specifically includes: Determine whether any two users are associated with different base stations and whether both base stations are within the communication range; If so, calculate the delay after the base stations associated with the two users are exchanged. If the delay after the exchange is less than the current delay, exchange the base stations associated with the two users.

4. A method for associating cloud-edge collaborative content caching with users as described in claim 3, characterized in that: The lazy user re-association strategy further includes: Pre-associate the user. If the content required by the user has been cached in the edge node, associate the user with the edge node.

5. A cloud-edge collaborative content caching and user association method as described in claim 1, characterized in that: The content caching strategy specifically includes: Obtain the user's content request history, and convert the content request history into a content request matrix; The content request matrix is ​​input into the trained bidirectional long short-term memory network and the fully connected network, and the predicted request probability matrix is ​​output to obtain the user's request probability for different contents in the future time window; The cache content is determined based on the obtained request probability.

6. A method for associating cloud-edge collaborative content caching with users as described in claim 1, characterized in that: The content caching strategy also includes determining the access probability of each content according to the user's request for the content, and determining the cached content based on the access probability.

7. A method for associating cloud-edge collaborative content caching with users as described in claim 6, characterized in that: The step of determining the access probability of each content according to the user's request for the content, and determining the cached content based on the access probability, includes: receiving user requests for content; Determine whether the requested content is cached in the edge node. If so, return the requested content to the user and update the access probability of the requested content. If not, the edge node obtains the corresponding requested content from the central server and returns it to the user, and calculates the access probability of the requested content; determines whether the cache space of the edge node is full. If the cache space is not full, caches the requested content into the cache space. If the cache space is full, finds the content with the lowest access probability in the cache space and replaces the content with the lowest access probability with the requested content.

8. A method for associating cloud-edge collaborative content caching with users as described in claim 1, characterized in that: The formula for the weighted delay minimization problem is expressed as: ； in, is the balance coefficient, represents the delay of all users obtaining content in time window j, represents the total switching delay in time window j, represents a time window set, A represents a user association policy, and C represents a content cache policy.

9. A method for associating cloud-edge collaborative content caching with users as claimed in claim 8, characterized in that: The constraints of the weighted delay minimization problem include: the size of the content cached by the edge node does not exceed its own capacity, whether the edge node caches content, whether the user is associated with the edge node through a sub-channel, one user is associated with at most one edge node, and each edge node is associated with at most a set number of users.

10. A cloud-edge collaborative content caching and user association system, characterized in that: include: A problem construction module is configured to construct a weighted delay minimization problem with the goal of minimizing a delay in user obtaining content and a delay in switching an associated base station; The content caching and user association module is configured to divide the constructed weighted delay minimization problem into a long-term user association and content caching problem under a long-term time scale and a short-term user association problem under a short-term time scale, and obtain a content caching strategy and a user association strategy to minimize the content transmission delay between the user and the base station; The content caching strategy includes predicting the content required by the user and caching it based on the bidirectional long short-term memory network according to the user's content request history; The user association strategy includes an online user association strategy and an inert user re-association strategy, wherein the online user association strategy includes associating the user with a base station corresponding to a sub-channel having a minimum download delay according to a content caching strategy; The lazy user re-association strategy includes, for each pair of users with different associated base stations after association, calculating the delay after the base station exchange, and if the delay is reduced after the exchange, then exchanging the base stations associated with the two users.

Images