A social-aware caching method and system suitable for wireless networks

Abstract

The present disclosure provides a social-aware caching method and system suitable for a wireless network, relating to the technical field of wireless communication, the method comprising: in an edge user network area, dividing users and devices into different groups according to a physical graph and a social relationship graph and obtaining a social-aware communication graph; judging potential connection objects in a region where a terminal user is located through the social-aware communication graph, and screening a user set of trusted users through a trust delivery mechanism, taking the user set of trusted users as caching nodes to obtain caching content; after obtaining the caching content, predicting user content preferences based on a caching content recommendation model of federated meta-learning and deep learning, caching the prediction results on the selected users serving as the caching nodes, and distributing the caching content according to the preferences of the users for different content in the next moment, so as to minimize the transmission delay and transmission cost on the basis of optimizing energy consumption. The present disclosure realizes accurate caching content pushing.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of wireless communication, in particular to a social-aware caching method and system suitable for wireless networks. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute the prior art.

[0003] With the explosive growth of intelligent devices and the emergence of many new applications, leading to the explosive growth of mobile data flow, increasing the load of backhaul links, the traditional centralized cloud structure has been difficult to meet the low latency requirements of content access. Therefore, edge caching technology suitable for mobile devices is proposed, which is used to provide computing and caching functions at the edge of the mobile network, by storing memory closer to the end user, reducing the transmission pressure of the network link, and improving the delay of content delivery. Since the service is closer to the end user, it is more sensitive to problems such as delay and needs high quality of experience. In addition, some application services also rely on low-latency transmission and high-speed data services. Then, the fog radio access network (F-RAN) architecture is proposed, which uses edge devices to locally process radio signals, cooperatively manage radio resources, and distribute storage capabilities to reduce the heavy burden of the front-end network.

[0004] However, the inventors have found that since the user end is usually equipped with limited storage space, it is impossible to cache all popular content on the end user device; how to cache the content that the user wants to access most in the limited storage space is a challenge, and the preference degree of the future end user for different content needs to be predicted in order to formulate the corresponding caching strategy; however, the current caching strategy based on content popularity has the problem of inaccurate prediction of hot content, and the exploration of user preference content and potential preference willingness is shallow, and the accuracy of content pushing is not enough. In addition, the trust problem between social users and the user privacy problem have not been well guaranteed, that is, when using a mobile device, whether the user is willing to share his content information, and whether there is a risk of leakage of user personal data information in the uploading process. Most importantly, the problem of balancing the relationship between energy consumption and transmission delay, cost, has not been fully studied in the entire transmission process. SUMMARY

[0005] In order to solve the above problems, the present disclosure proposes a social-aware caching method and system suitable for wireless networks, which, on the premise of guaranteeing user privacy, adopts a federated meta-learning method, designs a social-aware edge caching method suitable for fog radio access networks involving trust degree judgment, optimizes energy consumption, and minimizes transmission delay and transmission cost.

[0006] According to some embodiments, the present disclosure adopts the technical solutions as follows:

[0007] A social-aware caching method suitable for a wireless network comprises:

[0008] An edge user network area is obtained; in the edge user network area, users and devices are divided into different groups according to a physical graph and a social relationship graph, and a social-aware communication graph is obtained;

[0009] Potential connection objects in a region where a terminal user is located are determined through the social-aware communication graph, a user set of trusted users is screened through a trust delivery mechanism, and the user set of trusted users is used as a caching node to obtain caching content;

[0010] After the caching content is obtained, a caching content recommendation model based on federated meta-learning and deep learning is used to predict user content preferences, the prediction results are cached on the selected users used as the caching node, and the allocation of caching content to the preferences of users for different content at the next moment is performed, so that the transmission delay and transmission cost are minimized on the basis of optimizing energy consumption.

[0011] According to some embodiments, the present disclosure adopts the technical solutions as follows:

[0012] A social-aware caching system suitable for a wireless network comprises:

[0013] A content obtaining module obtains an edge user network area; in the edge user network area, users and devices are divided into different groups according to a physical graph and a social relationship graph, and a social-aware communication graph is obtained;

[0014] Potential connection objects in a region where a terminal user is located are determined through the social-aware communication graph, a user set of trusted users is screened through a trust delivery mechanism, and the user set of trusted users is used as a caching node to obtain caching content;

[0015] A content recommendation caching module, after the caching content is obtained, uses a caching content recommendation model based on federated meta-learning and deep learning to predict user content preferences, caches the prediction results on the selected users used as the caching node, and performs the allocation of caching content to the preferences of users for different content at the next moment, so that the transmission delay and transmission cost are minimized on the basis of optimizing energy consumption.

[0016] According to some embodiments, the present disclosure adopts the technical solutions as follows:

[0017] A computer-readable storage medium, in which a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor of a terminal device, and a social-aware caching method suitable for a wireless network is performed.

[0018] According to some embodiments, the present disclosure adopts the technical solutions as follows:

[0019] A terminal device comprises a processor and a computer readable storage medium, the processor is used to implement instructions; the computer readable storage medium is used to store a plurality of instructions, the instructions are suitable for being loaded and executed by the processor to implement a social awareness caching method for a wireless network.

[0020] Compared with the prior art, the present disclosure has the beneficial effects that:

[0021] The social awareness caching method for a wireless network provided by the present disclosure is an edge caching method that integrates social awareness and federated meta-learning of trust delivery mechanism, and is used to select the optimal caching strategy and content recommendation decision. The problems of shallow exploration of user preference content and potential preference willingness and high delay and cost in the transmission process in the existing edge caching scheme of fog radio access network are solved. In the case of ensuring user privacy, a social awareness edge caching method suitable for wireless networks, especially fog radio access networks, involving trust degree judgment is designed.

[0022] Based on the special properties of edge user networks, the present disclosure adopts a social awareness communication sharing mode, divides the user communication situation into a physical graph and a social relationship graph, and obtains a social awareness communication graph model through the two. This is conducive to fully exploring the user communication situation and content preference, determining the most suitable potential content caching node, and realizing the precision of content pushing.

[0023] The present disclosure designs a user trust delivery mechanism for potential communication users who need to establish a connection. A delay and cost model considering energy consumption in the transmission process is constructed. The evaluation index is considered in the update of the caching strategy and the content recommendation decision, to construct a caching update strategy considering energy consumption and cost, and a caching content recommendation decision considering delay and energy consumption; the prediction model is trained by using the federated meta-learning method. Meta-learning can be used to quickly learn and quickly adapt in a multi-task scenario. By using the federated learning method, the security of user information data is effectively guaranteed, and the privacy of users is protected.

[0024] The present disclosure introduces a fog radio access network architecture, and uses edge devices to reduce the heavy burden of the front-end network and improve the transmission quality of experience by using local radio signal processing, cooperative radio resource management and distributed storage capabilities of the F-AP node. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings, which form a part of the present disclosure, are used to provide a further understanding of the present disclosure, and the schematic embodiments of the present disclosure and their descriptions are used to explain the present disclosure, and do not constitute an improper limitation on the present disclosure.

[0026] Figure 1 A system model architecture diagram provided by an embodiment of the present disclosure;

[0027] Figure 2 A workflow diagram when acquiring cache content provided by an embodiment of the present disclosure;

[0028] Figure 3 A federated meta-learning framework diagram disclosed by an embodiment of the present disclosure;

[0029] Figure 4 A construction schematic diagram of a social-aware communication graph based on a physical graph and a social relationship graph disclosed by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] The present disclosure will be further described below in conjunction with the accompanying drawings and embodiments.

[0031] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs.

[0032] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and furthermore, it should be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0033] Embodiment 1

[0034] In an embodiment of the present disclosure, a social-aware caching method suitable for a wireless network is provided, comprising:

[0035] Step 1: Obtain an edge user network area; in the edge user network area, divide users and devices into different groups according to a physical graph and a social relationship graph and obtain a social-aware communication graph;

[0036] Step 2: Determine potential connection objects in the area where the terminal user is located through the social-aware communication graph, and filter a trusted user set of trusted users through a trust delivery mechanism, and use the trusted user set of trusted users as cache nodes to acquire cache content;

[0037] Step three: after obtaining the cache content, the user content preference is predicted based on the cache content recommendation model of federated meta-learning and deep learning, the prediction result is cached on the selected user used as the cache node, and the allocation of the cache content to the user's preference for different content at the next moment is performed, so as to minimize the transmission delay and transmission cost on the basis of optimizing the energy consumption.

[0038] As an embodiment, the present disclosure adopts a social-aware communication sharing mode based on the special properties of edge user networks, divides the user communication situation into a physical graph and a social relationship graph, and obtains a social-aware communication graph model through the two graphs. This is conducive to fully mining the user communication situation and content preference, determining the most suitable potential content cache node, and realizing accurate content pushing.

[0039] Specifically, as shown in the figure, Figure 4 the construction of the social-aware communication graph is based on the construction of the physical graph and the social relationship graph, wherein:

[0040] (1) Physical graph:

[0041] Due to the influence of factors such as communication signal attenuation caused by different physical distances between terminal users. Only terminal users with close distance and detectable communication signal strength between each other will be selected as candidates for communication relays, and then a communication connection is established. Therefore, the concept of a physical graph is introduced, and specifically, G(U u ,E u ) is used to represent the physical connectivity properties between terminal users. Wherein U u represents the vertex set, E u represents the edge set, i.e. e iju ∈{0,1}. Wherein, e iju =1 only when two terminal users i, j establish a communication connection. Otherwise, it is zero. e is a binary variable, e iju =1 only when two terminal users (for example, i and j represent two terminal users) establish a communication connection, that is, one of the users is a feasible relay for the other user.

[0042] (2) Social relationship graph: From a realistic point of view, considering the selfishness and social trust phenomenon of terminal users, users with stronger social relationships are more willing to directly share their content, while users with weaker social relationships may not be willing to provide content to neighbors (other terminal users used to establish a connection) due to privacy issues and the like. Therefore, G(Us,Es) is introduced to represent the social connectivity properties between terminal users. Wherein Us represents the vertex set, Es represents the edge set, i.e. e ijs∈ {0, 1}. When two end users i, j have close social relationships, such as friends and family members, set e ijs = 1; otherwise, e ijs = 0.

[0043] (3) Socially-aware communication graph:

[0044] Based on the physical graph and social graph obtained by the above analysis, a socially-aware graph G(U g , E g ) can be further obtained. Wherein U g represents the vertex set, E g represents the edge set, i.e. e ijg ∈ {0, 1}. Through calculation, e ijg = e iju * e ijs know whether there is a communication path. In particular, when e ijg = 1, there is a communication path between the end user k and the end user j in the physical graph and the social graph.

[0045] When there is an available communication path between two end users (such as users i, j), the key to whether the absolute content is shared and transmitted lies in whether the two users pass the trust delivery mechanism. Here, according to the end users i and j, a trust delivery mechanism is specially designed, and the trust degree is represented by φ i,j , the formula is as follows:

[0046]

[0047] Wherein P i,j represents the physical transmission distance between the end users i and j, N i,j represents the historical communication times between the end users i and j, represents the communication times between the end users i and j and their common neighbors, represents the sum of the interest preferences of the end users i and j for a certain type of content to be transmitted. α, β, γ represent weight coefficients.

[0048] According to the above formula definition, when φ i,j is higher than the pre-set threshold value, the end users i and j will form a social relationship, and there is a communication path between them. At this time, e ijg = 1.

[0049] Further through the judgment of the trust delivery mechanism, the user set U U φ (i) representing the user index set trusted by the end user i, means that Uφ The user in (i) is willing to share their cached content with end user i. Furthermore, U φ (i) The set of all neighbor indexes U of terminal user i neb (i) Taking the intersection yields the user index set that may establish a communication connection with end user i and provide transmission content. The formula is as follows:

[0050]

[0051] Then, as an example, the potential connection objects in the area where the end user is located are determined by the socially aware communication graph, and the user set of trustworthy users is filtered by the trust delivery mechanism. The user set of trustworthy users is used as a cache node to obtain cached content.

[0052] Specifically, such as Figure 2 As shown, when an end user generates a content file request, the steps for retrieving the cached content are as follows:

[0053] Step 1: When the content to be requested is cached in its own storage space, the content can be directly retrieved from its cache list, and there is no need to establish an external communication link.

[0054] Step 2: If the user's local cache list cannot satisfy the content request, then the user can request the content from neighboring users in the communication area with whom a communication connection can be established.

[0055] Step 3: If neighboring users still cannot meet the content request requirements, the content request can be sent to the local F-AP node to obtain the content.

[0056] Step 4: In this patented method, the case of F-AP node association is considered. Therefore, when the content is not satisfied, a request can be sent to the adjacent F-AP node. If the content can be found in the adjacent F-AP node, the content is obtained.

[0057] Step 5: If the above methods fail to obtain the required content, then search the central cloud server's main file content library.

[0058] As one implementation, after obtaining cached content, a cached content recommendation model based on federated meta-learning and deep learning is proposed. This model combines the results of federated meta-learning and deep learning methods to further predict user content preferences, provide caching decisions, and ensure user privacy. The scheme uses federated meta-learning based on content popularity prediction to make cached content recommendation decisions. A deep learning model is trained, and the trained prediction model is used to allocate cached content based on user preferences for different content at the next time step, preparing for node caching. This approach aims to minimize transmission latency and cost while optimizing energy consumption.

[0059] The training method for the cached content recommendation model based on federated meta-learning and deep learning described in this disclosure includes:

[0060] S1: Each time a certain proportion of terminal users are selected to participate in the cache recommendation training process, the local F-AP node will download the initial network parameters of the neural network model to update the local model parameters.

[0061] S2: End users participating in training within the domain download the initial model parameters from the F-AP node, use the locally computed data containing the user's local historical information to train the neural network model, obtain the updated model parameters, and upload them to the F-AP node;

[0062] S3: The F-AP node uploads the model parameters uploaded by each user terminal to the central cloud server. The central cloud server aggregates all the model parameters and obtains the updated global model parameters.

[0063] S4: The central cloud server then distributes the updated global model parameters to each F-AP node for the next iteration; repeat steps S2 and S3 until the model converges.

[0064] F-AP nodes use a trained model to determine the optimal recommendation scheme for cached content. The optimal cache recommendation scheme is derived based on the minimum energy consumption required to complete the content caching task.

[0065] Based on the limited transmission and computing resources of wireless devices, energy consumption factors are introduced to derive a cache update strategy that considers cost and energy consumption, and a content recommendation strategy that considers latency and energy consumption. The optimization goal is to minimize the cost and latency of all computing tasks.

[0066] The minimum delay function is determined as follows:

[0067]

[0068] Where C = [c1, c2, ..., c N] is a vector of length N, representing the content cached on each F-AP node; It is the specified content c i The probability of being accessed on F-AP node i; Retrieving content from edge node i to the cloud c i The required delay time.

[0069] The minimum cost function is determined as follows:

[0070]

[0071] Among them, s i This is the total size of the content stored in F-AP node i; It is the specified content c i The cost of storage on F-AP node i Content is retrieved from the cloud. i Network costs to F-AP node i.

[0072] The energy consumption is defined as consisting of two parts: the energy consumption for storing content on the F-AP node and the energy consumption for transmitting content. The transmission energy consumption mainly comes from the wireless transmission energy consumption overhead from the F-AP node to the user, the wired assistance transmission overhead between F-AP nodes, and the wired transmission overhead from the F-AP node within the domain to the remote cloud service content center.

[0073] Specifically, the storage energy consumption and content transfer energy consumption on the F-AP node are defined as follows:

[0074] Storage energy consumption on F-AP nodes:

[0075]

[0076] in, It is the cost of storing the specified content ci on F-AP node i. It is the specified content c i Storage energy consumption when storing on F-AP node i.

[0077] Content transmission power consumption:

[0078]

[0079] in, This refers to the energy consumption of wireless transmission from F-AP node i to the user; This refers to the power consumption of wired assisted transmission from F-AP node i to other F-AP nodes; This refers to the power consumption of wired transmission from F-AP node i to the remote content center.

[0080] Taking into account both storage energy consumption and content transmission energy consumption, the energy consumption model can be defined as follows:

[0081] E = εE s +(1-ε)E t

[0082] Here, ε is the energy consumption coefficient, which is used to control the weight of storage energy consumption and content transmission energy consumption.

[0083] In minimizing total energy consumption, the caching strategy and content recommendation decisions need to be updated. For caching strategy updates, the minimum cost function can be modified to take storage energy consumption into account.

[0084]

[0085] Here, ce is the storage energy consumption coefficient, which is used to control the weight of storage energy consumption.

[0086] For content recommendation decisions, a pre-trained model will be used. Specifically, the probability distribution of each piece of content at each F-AP node will be calculated, and the content with the highest probability will be cached. This allows for cached content recommendation decisions to be made across multiple F-AP nodes, and the model is trained using a federated meta-learning model. The minimum latency function can be modified to consider a comprehensive metric that simultaneously takes into account latency and energy consumption.

[0087]

[0088] in, It's the energy consumption for content transmission. This refers to the distance from F-AP node i to content c. i The delay.

[0089] Finally, the federated meta-learning method is used to solve the problem. A global model is trained using local data on end users, and then the global model is sent to end users for updates until the model reaches the convergence condition (the model loss value no longer changes). This achieves a cached content recommendation decision that considers both latency and cost in the edge caching problem.

[0090] Example 2

[0091] One embodiment of this disclosure provides a socially aware caching system suitable for wireless networks, comprising:

[0092] The content acquisition module acquires the edge user network area; within the edge user network area, it divides users and devices into different groups based on the physical graph and social relationship graph and acquires the social-aware communication graph.

[0093] By using a socially-aware communication graph, potential connection objects within the area where the end user is located are identified, and a trust delivery mechanism is used to filter the user set of trustworthy users. The user set of trustworthy users is then used as a cache node to obtain cached content.

[0094] The content recommendation caching module, after acquiring cached content, predicts user content preferences based on a cached content recommendation model using federated meta-learning and deep learning. The prediction results are cached on the selected users who are used as cache nodes. The module also allocates cached content according to the user's preferences for different content in the next time step, minimizing transmission latency and transmission cost while optimizing energy consumption.

[0095] Specifically, the system model is as follows: Figure 1 As shown, consider establishing a fog wireless access network architecture consisting of M F-AP nodes, U end users, and one central cloud server. This includes three different types of cache nodes: end users, F-AP nodes (equipped with fog computing servers), and the central cloud server. Let the set of F-AP nodes be M = {1, ..., n, ..., M}, where one F-AP node serves a group of U users. The user index set served by each F-AP node is denoted by U(k) = {1, ..., U}. Each F-AP node and end user is equipped with a limited storage space C. f This is used for prefetching and storing partially cached content. Content accessible to all end users is represented by a content library F = {1, ..., F}, where F is the index of all content. Each content file is the same size, and their size is denoted by (D...). f ) 1×γ This method assumes that the central cloud server has a sufficiently large cache space to always store the entire file library F, and that the total cache of F-AP cannot exceed the storage budget limit C specified by the mobile network operator. f Each end user can obtain the requested content from the F-AP node via a cellular connection.

[0096] Example 3

[0097] One embodiment of this disclosure provides a computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device of the socially aware caching method for wireless networks.

[0098] Example 4

[0099] This disclosure provides a terminal device including a processor and a computer-readable storage medium. The processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions adapted for loading and execution by the processor of the socially aware caching method applicable to wireless networks.

[0100] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0101] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0102] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A social-aware caching method for wireless networks, characterized in that, Comprise: Obtain edge user network area; In the edge user network area, divide users and devices into different groups according to physical graph and social relationship graph and obtain social awareness communication graph; Determine potential connection objects in the area where the terminal user is located through the social awareness communication graph, and filter the user set of trusted users through the trust delivery mechanism, and obtain the cache content by taking the user set of trusted users as the cache node; When a terminal user generates a content file request, the steps for obtaining cache content are as follows: When the content to be requested is cached in the storage space, the content is obtained directly in the cache list, and no external communication link is established; If the user's local cache list cannot meet the content request, the neighbor user in the communication area that can establish a communication connection is requested; If the neighbor user still cannot meet the requirements of the content request, the content request is sent to the local F-AP node to obtain the content; After obtaining the cache content, the cache content recommendation model based on federated meta-learning and deep learning predicts the user content preference, caches the prediction result on the selected user used as the cache node, and allocates the cache content to the user's preference for different content in the next moment, so as to minimize the transmission delay and transmission cost on the basis of optimizing energy consumption; The training method of the cache content recommendation model based on federated meta-learning and deep learning comprises: S1: Select a certain proportion of terminal users to participate in the cache recommendation training process each time, and the local F-AP node downloads the initial network parameters of the neural network model to update the local model parameters; S2: The terminal user participating in the training in the domain downloads the initial model parameters from the F-AP node, trains the neural network model using the data containing the user's local historical information calculated locally to obtain updated model parameters, and uploads them to the F-AP node; S3: The F-AP node uploads the model parameters uploaded by each user terminal to the central cloud server, and the central cloud server aggregates all the model parameters to obtain updated global model parameters; S4: The central cloud server distributes the updated global model parameters to each F-AP node for the next iteration; repeat steps S2 and S3 until the model converges; The energy consumption is composed of storage energy consumption of cache content on the F-AP node and content transmission energy consumption, wherein the content transmission energy consumption is composed of wireless transmission energy consumption overhead of the F-AP node to the user, wired assisted transmission overhead between F-AP nodes, and wired transmission overhead of the F-AP node in the domain to the remote cloud service content center; define the storage energy consumption and content transmission energy consumption on the F-AP node, and define the energy consumption model by comprehensively considering the storage energy consumption and content transmission energy consumption; In the process of minimizing the total energy consumption, the cache strategy and the content recommendation decision are updated. For cache strategy update, the minimum cost function is modified to consider the storage energy consumption. For content recommendation decision, the trained content recommendation model is used to make the recommendation decision. The probability distribution of each cached content on each F-AP node is calculated, and the content with the highest probability is selected for caching. 2.The social-aware caching method for wireless networks of claim 1, wherein, Considering the joint case of F-AP nodes, when the content is not satisfied, a request can be sent to the adjacent F-AP node. If the content is found in the adjacent F-AP node, the content is obtained. If the required content cannot be obtained, the central cloud server's total file content library is searched. 3.The social-aware caching method for wireless networks of claim 1, wherein, The next moment's user preference for different content is allocated to the cached content. On the basis of optimizing energy consumption, the cache update method with the minimum transmission delay and transmission cost includes: introducing the energy consumption factor to determine the cache update strategy considering the cost and energy consumption, and the content recommendation strategy considering the delay and energy consumption. The optimization goal is to minimize the cost and delay of all computing tasks.

4. A socially aware caching system suitable for a wireless network, characterized in that, It includes: A content acquisition module acquires an edge user network area. In the edge user network area, users and devices are divided into different groups according to physical and social relationship graphs, and a social awareness communication graph is acquired. Through the social awareness communication graph, the potential connection objects in the area of the terminal user are determined, and a trusted user set is screened through a trust delivery mechanism. The trusted user set is used as a cache node to obtain cached content. When a terminal user generates a content file request, the step of obtaining cached content is as follows: When the requested content is cached in the storage space, the content is directly obtained in the cache list without establishing an external communication link. If the user's local cache list cannot meet the content request, the neighbor users in the communication area who can establish a communication connection are requested. If the neighbor users still cannot meet the requirements of the content request, the content request is sent to the local F-AP node to obtain the content. A content recommendation cache module, after obtaining the cached content, predicts the user's content preference based on the cache content recommendation model of federated meta-learning and deep learning. The prediction result is cached on the selected user used as a cache node, and the next moment's user preference for different content is allocated to the cached content to minimize the transmission delay and transmission cost on the basis of optimizing energy consumption. The training method of the cache content recommendation model of federated meta-learning and deep learning includes: S1: A certain proportion of terminal users are selected to participate in the cache recommendation training process each time. The local F-AP node downloads the initial network parameters of the neural network model to update the local model parameters. S2: The terminal users participating in the training in the domain download the initial model parameters from the F-AP node, train the neural network model using the data containing the local historical information calculated locally, obtain the updated model parameters, and upload them to the F-AP node. S3: The F-AP node uploads the model parameters uploaded by each user terminal to the central cloud server, the central cloud server aggregates all the model parameters, and obtains updated global model parameters; S4: The central cloud server distributes the updated global model parameters to each F-AP node for the next round of iteration; repeat the above steps S2 and S3 until the model converges; The energy consumption is composed of storage energy consumption of the cached content on the F-AP node and content transmission energy consumption, wherein the content transmission energy consumption is composed of wireless transmission energy consumption overhead of the F-AP node to the user, wired assisted transmission overhead between the F-AP nodes, and wired transmission overhead of the F-AP node in the domain to the remote cloud service content center; the storage energy consumption and the content transmission energy consumption on the F-AP node are defined, and the storage energy consumption and the content transmission energy consumption are comprehensively considered to define the energy consumption model; In the process of minimizing the total energy consumption, the cache strategy and the content recommendation decision are updated, for the cache strategy update, the storage energy consumption is considered by modifying the minimum cost function; for the content recommendation decision, the trained content recommendation model is used to make the recommendation decision, the probability distribution of each cached content on each F-AP node is calculated, and the content with the highest probability is selected for caching.

5. A computer readable storage medium, characterized in that, A plurality of instructions are stored therein, and the instructions are adapted to be loaded and executed by the processor of the terminal device to implement the social-aware caching method for wireless networks according to any one of claims 1-3.

6. A terminal device, characterized by comprising: The computer readable storage medium is used for storing a plurality of instructions, and the instructions are adapted to be loaded and executed by the processor to implement the social-aware caching method for wireless networks according to any one of claims 1-3.

Images