Mobile social network resource allocation method based on cloud side-end cooperation

Abstract

The invention discloses a mobile social network resource allocation method based on cloud side-end cooperation, and the method comprises the steps: screening out a malicious ES through building a trust evaluation mechanism, guaranteeing the safety of a task unloading process, improving the data transmission efficiency between an MU and the ES through semantic transmission, constructing a two-layer task unloading frame and an incentive mechanism, and achieving the resource allocation of the MU and the ES. And optimizing an unloading strategy of the MU and the ES and a pricing strategy between the ES and the cloud. In the aspect of system composition, a mobile user, an edge server and a cloud server form a cloud edge-end cooperative mobile social network, and a trust evaluation machine comprehensively considers direct trust degree, indirect trust degree, intimacy degree, contribution history and the like, so that trust evaluation based on historical interaction records of the mobile user and the edge server is realized; for task unloading and excitation, an excitation-oriented two-layer task unloading scheme is designed, a Stackelberg game model is constructed to optimize strategies of all parties, and the problems of limited computing resources, high data transmission delay, low communication efficiency, security threat and unreasonable resource allocation of mobile equipment are solved.

Description

Technical Field

[0001] This invention belongs to the field of mobile computing and communication technology, specifically relating to a cloud-edge-device collaborative mobile social network resource allocation method, which aims to solve problems such as limited computing resources of mobile devices, security threats from task offloading, and resource allocation efficiency. Background Technology

[0002] With the booming development of mobile internet, applications such as augmented reality and real-time data analytics are increasingly demanding computing resources. However, limited by physical size and battery capacity, mobile devices typically have limited computing resources, making it difficult to meet the needs of these applications. While cloud computing allows mobile users (MUs) to migrate computationally intensive tasks to cloud centers, improving computing performance and reducing device power consumption, the simultaneous uploading of a large number of tasks can easily cause network congestion and processing delays, impacting user experience. Mobile edge computing, as an emerging computing paradigm, deploys computing and storage resources on edge servers closer to mobile users, enabling mobile devices to offload tasks to nearby edge servers (ESs) for processing, effectively shortening data transmission distances and reducing task processing latency.

[0003] In urban hotspots, numerous mobile devices simultaneously connect to edge servers, leading to a surge in communication pressure. Traditional communication technologies suffer from low transmission efficiency and high system overhead when dealing with such dense communication demands. In recent years, semantic communication technology has emerged, transmitting semantic information rather than raw data. It can transmit more information with the same bandwidth, improving communication efficiency and reducing transmission latency and system overhead. However, security risks exist between mobile users and edge servers. Malicious servers may steal user privacy data, tamper with task processing results, or provide low-quality services, seriously threatening mobile user security. Simultaneously, edge servers and cloud servers increase their own computational burden by processing tasks. Therefore, effective incentive mechanisms are needed to encourage their active participation in resource allocation and task processing to improve the computing service quality of cloud-edge-device collaborative mobile social networks. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that existing technologies suffer from limited computing resources, high data transmission latency, low communication efficiency, security threats, and unreasonable resource allocation in mobile devices.

[0005] To address the aforementioned technical problems, the present invention provides a method for allocating resources in a mobile social network through cloud-edge-device collaboration, comprising the following steps: Considering the historical interaction records between MU and ES, the social trust score is calculated by measuring direct trust, indirect trust, intimacy, and contribution history indicators. When MU unloads a task, it selects a trusted ES based on the social trust score. The semantic transmission rate is determined based on the signal-to-interference-plus-noise ratio and semantic similarity, combined with bandwidth and average semantic information content. Based on local computing time, transmission duration calculated through semantic transfer rate, offloading time, upload and computing time, and the overall latency of MU in completing tasks, accurately assess the latency of MU under different task offloading strategies. Based on necessary transmission power and transmission energy consumption, consider transmission power and transmission time to evaluate transmission energy consumption. Based on MU's local computing energy consumption, ES 𝑘's total workload, local computing latency, local computing energy consumption, cloud's local computing latency, and cloud computing energy consumption, comprehensively consider the energy consumption of different devices during the computing process. Construct utility functions for MU, ES, and cloud, and formulate corresponding optimization problems and constraints. We use a two-layer Stackelberg game theory framework and backward induction to solve the constructed optimization problem. In the first layer of the game, the edge server, as the leader, sets the price, and the mobile user MU, as the follower, maximizes the utility function of MU and obtains the optimal unloading strategy for MU 𝑖 by combining different conditions; In the second layer of the game, the cloud, as the leader, sets the price and obtains the optimal offloading strategy of MU. The edge server, as the follower, substitutes the optimal offloading strategy of MU into the utility function of ES and solves the optimal pricing strategy and the optimal offloading strategy of ES according to the constraints. Substituting the optimal offloading strategy of ES into the utility function of the cloud, the optimal pricing strategy of the cloud is obtained. The equilibrium solution of the two-layer game is obtained by back induction, and the optimal strategies of mobile users, edge servers and the cloud are obtained. Based on the corresponding optimal strategies, the perception tasks of MU are offloaded to the trusted ES. When the ES has insufficient computing resources, it requests the assistance of the cloud server.

[0006] Preferably, the direct trust level is calculated based on the historical interaction scores of MU to ES.

[0007] Preferably, the indirect trust level is calculated based on recommendations from other MUs and social relationships.

[0008] Preferably, the contact intimacy index is calculated based on the strength of the interaction between MU and ES, measured from the dimensions of time and frequency.

[0009] Preferably, the utility function of the MU The formula is as follows:

[0010] in, Adjust parameters to increase satisfaction. For maximum delay, Calculate the time locally. For upload and calculation time, The price paid to ES. For semantic transmission cost parameters, For locally calculated cost parameters, For transmission energy consumption, This represents the total workload.

[0011] Preferably, the utility function of the ES The formula is as follows:

[0012] in, For the revenue obtained from MU, The percentage of content that is offloaded to the cloud. Total workload For locally calculated cost parameters, For local calculation of energy consumption, The price paid to the cloud.

[0013] Preferably, the utility function of the cloud The formula is as follows:

[0014] in, For the revenue obtained from ES, The percentage of Elasticsearch (ES) offloaded to the cloud. Total workload For locally calculated cost parameters, This refers to the computing energy consumption of the cloud.

[0015] This invention discloses a cloud-edge-device collaborative mobile social network resource allocation method. It establishes a trust assessment mechanism to filter out malicious Elastic Search Engines (ES), ensuring the security of the task offloading process. Semantic transmission technology is used to improve data transmission efficiency between Mobile Users (MU) and ES. A two-layer task offloading framework and incentive mechanism are constructed, optimizing the offloading strategies for MU and ES, as well as the pricing strategy between ES and the cloud. In terms of system composition, the cloud-edge-device collaborative mobile social network consists of mobile users, edge servers, and cloud servers. Regarding the trust assessment mechanism, multiple indicators such as direct trust, indirect trust, intimacy, and contribution history are comprehensively considered to achieve trust assessment based on the historical interaction records between mobile users and edge servers. For the task offloading and incentive mechanism, an incentive-oriented two-layer task offloading scheme is designed, and a Stackelberg game model is constructed to optimize the strategies of each party, addressing the problems of limited computing resources on mobile devices, high data transmission latency, low communication efficiency, security threats, and unreasonable resource allocation. Attached Figure Description

[0016] Figure 1 A flowchart for optimizing the strategies of each party based on the Stackelberg game model provided in this embodiment of the invention; Figure 2 A three-layer system structure diagram provided for embodiments of the present invention; Figure 3 This is an example diagram of a three-way game process in a two-layer mobile cloud edge-end social network provided in an embodiment of the present invention; Figure 4 A block diagram of a two-layer mobile social network offloading system based on semantic transmission provided in an embodiment of the present invention; Figure 5 The curves showing the change in mobile user utility under different computing tasks as the proportion of mobile users uninstalling the computer, provided in the embodiments of the present invention. Figure 6 The curves showing the variation of mobile user utility with computing power under different computing tasks, as provided in the embodiments of the present invention; Figure 7 The edge server utility provided by the embodiments of the present invention varies with edge server pricing under different offload ratios. Change curve; Figure 8 The cloud server utility provided in this embodiment of the invention varies with different user uninstallation rates and cloud server pricing. Change curve; Figure 9 This invention provides a comparison of different solutions for mobile user utility under different user uninstallation rates in its embodiments. Figure 10 The cloud server utility provided in this embodiment of the invention is priced for different edge servers. The following compares the different solutions. Detailed Implementation

[0017] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0018] like Figure 1 As shown, this embodiment of the invention provides a method for allocating resources in a cloud-edge-device collaborative mobile social network, including the following steps: like Figure 2 As shown, a system model is constructed: For cloud-edge-device collaborative mobile social network scenarios, a two-layer task offloading system model is constructed, which includes mobile users (MU), edge servers (ES), and cloud servers, clarifying the computing resources and task processing relationships of each part.

[0019] Among them, the perception tasks of MU can be offloaded to ES. When ES computing resources are insufficient, it can request assistance from cloud servers, and relevant parameters are represented by specific symbols, such as... Represents the set MU. This indicates the total workload of MU_i. MU_i can choose to offload some of its workload to a trusted Elasticsearch instance for processing. Represents the ES collection. This indicates the percentage of MU 𝑖 that is unloaded from ES 𝑘. After receiving the task, ES 𝑘 can choose to process it locally or partially unload it to the cloud. This indicates the percentage of ES files that are unloaded to the cloud.

[0020] Establish a trust assessment mechanism: Considering the historical interaction records between MU and ES, establish a social trust model, including calculating direct trust, indirect trust, intimacy, and historical contribution indicators, and synthesize these indicators to obtain the social trust score. This is used to measure the trust between MU and ES, and to filter malicious ES. Direct trust is calculated based on historical interaction scores, indirect trust is calculated based on recommendation information and social relationships from other MUs, intimacy index measures the strength of interaction relationship from the dimensions of time and frequency, and contribution history index reflects the service contribution of ES.

[0021] The direct trust level is calculated based on MU's historical interaction rating of ES. The formula is:

[0022] in, This is MU's rating of ES's nth service, with a range of values. , Let be the time decay function, where is the decay rate used to reflect the timeliness of historical ratings, and is the current service time. This is the nth historical service time.

[0023] Indirect trust level calculated based on recommendations from other MUs and social relationships. This includes the following steps: First, calculate the confidence level of MU_i on MU_i:

[0024] in, The degree of social relationship of relationship 𝑞 Indicate whether there is a relationship 𝑞 between MU 𝑖 and MU 𝑗. The similarity value between MU𝑖 and MU𝑗 is measured by the Pearson correlation coefficient, which is obtained by comparing their scores for the same ES. Similarity value between MU 𝑖 and MU 𝑗 The formula is as follows:

[0025] Then calculate the indirect trust level: .

[0026] By using the indirect trust formula, MU can more comprehensively assess the credibility of ES by leveraging other users' reviews and social relationships.

[0027] Based on intimacy, the strength of the interaction between MU and ES is measured from the dimensions of time and frequency, and the contact intimacy index is calculated. The formula is: when and hour, ; when and hour, ; in, For MU 𝑖 and ES 𝑘 past time units of service time, The total service time for MU and all ES reflects connection stability. This represents the number of interactions between MU_i and ES_i within a past time unit. The total number of interactions between MU and all ES reflects the level of interaction activity. 𝜇 (0 ≤ 𝜇 ≤ 1) is a weight parameter used to adjust the relative importance of time and frequency factors.

[0028] Historical contribution metrics reflecting the extent of ES's service contribution in historical collaborations The calculation method is as follows: when hour: ; when hour: ; in The computational workload provided by ES for MU , The total amount of computing resources that MU 𝑖 has obtained through interactions with ES in its contribution history.

[0029] Based on the above indicators, social trust The calculation method is as follows:

[0030] in, As a weighting factor, and .

[0031] When MU uninstalls a task, it selects a trusted Elasticsearch Provider (ES) based on social trust levels.

[0032] Design a semantic transmission model: Define the text semantic task of MU A deep learning-driven semantic communication (DeepSC) model is used to calculate the signal-to-interference-plus-noise ratio (SINR), a measure of the quality of the received signal. The formula is as follows:

[0033] in, For the transmission power from MU to ES, For channel gain, For average channel power, This represents noise power.

[0034] The semantic similarity is calculated based on the quality of the received signal to quantify the degree of recovery of the task at the receiving end. The formula is as follows:

[0035] in, and These are the initial sentence for MU and the relevant recovered sentence for ES, respectively. A bidirectional encoder based on the BERT model represents sentences. for The transpose of .

[0036] The range of semantic similarity values ​​is: A higher value indicates higher sentence similarity. In practice, semantic similarity is determined by the average number of semantic symbols per word. and ES A decision can be expressed as:

[0037] And it is approximated using logical functions as follows:

[0038] in, , , , It is a constant.

[0039] To ensure accurate recovery of the uninstallation task, the semantic representation must satisfy:

[0040] To achieve the minimum required semantic similarity accuracy.

[0041] Based on the signal-to-interference-plus-noise ratio (SINR) and semantic similarity, combined with bandwidth and average semantic information content, the semantic transmission rate is determined to improve the data transmission efficiency between the MU and ES. The calculation method is as follows:

[0042] in, Bandwidth provided for ES 𝑘 The average semantic information content, These parameters, representing the average word length, are determined by the source type and the DeepSC model.

[0043] Construct a task offloading model: Based on local computation time, transmission duration, offloading time, upload and computation time, and the overall latency of the MU in completing the task, accurately assess the latency of the MU under different task offloading strategies. Based on necessary transmission power and transmission energy consumption, consider transmission power and transmission time to evaluate transmission energy consumption. Based on the MU's local computation energy consumption, the total workload of ES 𝑘, local computation latency, local computation energy consumption, cloud local computation latency, and cloud computation energy consumption, comprehensively consider the energy consumption of different devices during the computation process.

[0044] Local time calculation formula:

[0045] in, The number of CPU cycles required for MU to process each bit of data. This represents the computational speed of MU_i.

[0046] Transmission duration formula:

[0047] Unloading time formula:

[0048] Upload and calculation time formulas:

[0049] The overall delay formula for MU task completion:

[0050] Formula for the necessary transfer power from MU to ES:

[0051]

[0052] Transmission energy consumption formula:

[0053] MU's local computing energy consumption formula:

[0054] in, This is the power consumption factor.

[0055] The formula for the total workload of ES is:

[0056] Local computation delay formula:

[0057] Local energy consumption calculation formula:

[0058] in, Let be the power consumption coefficient of ES 𝑘.

[0059] The formula for local computing latency in the cloud:

[0060] Cloud computing energy consumption formula:

[0061] in, This represents the power consumption coefficient of the cloud.

[0062] like Figure 3 As shown, a utility model and optimization problem are constructed: utility functions for MU, ES, and cloud are constructed respectively, and the benefits and costs of each party are comprehensively considered, providing a quantitative basis for optimizing resource allocation. The task unloading process is modeled as a Stackelberg game to optimize the unloading strategies of MU and ES, as well as the pricing strategy between ES and cloud, to maximize the utility of each party. At the same time, optimization problems and constraints for each participant are formulated. For example, the optimization problem of MU includes constraints on variables such as unloading ratio and satisfaction adjustment parameters, and the optimization problems of ES and cloud also have corresponding variables and constraints.

[0063] The utility function of the MU The calculation method is as follows:

[0064] in, Adjust parameters to increase satisfaction. For maximum delay, Calculate the time locally.

[0065] For upload and calculation time, The price paid to ES. For semantic transmission cost parameters, For locally calculated cost parameters, For transmission energy consumption, This represents the total workload.

[0066] The utility function of the ES The calculation method is as follows:

[0067] in, For the revenue obtained from MU, The percentage of content that is offloaded to the cloud. Total workload For locally calculated cost parameters, For local calculation of energy consumption, The price paid to the cloud.

[0068] The utility function of the cloud The calculation method is as follows:

[0069] in, For the revenue obtained from ES, The percentage of Elasticsearch (ES) offloaded to the cloud. Total workload For locally calculated cost parameters, This refers to the computing energy consumption of the cloud.

[0070] Optimization problem: For MU, the optimization problem is: The constraints include: Ensure the uninstallation rate is within a reasonable range. A value of 0 indicates that MU does not unload tasks to ES, while a value of 1 indicates that all tasks are unloaded. Values ​​within this range can be used to flexibly adjust the amount of tasks unloaded.

[0071] Limit the values ​​of the satisfaction adjustment parameters. The closer MU is to 1, the more sensitive it is to task processing time and the more it focuses on the timeliness of task processing; the closer it is to 0, the more it focuses on other factors such as task processing energy consumption.

[0072] By determining the range of weight parameters in the closeness metric, and by adjusting the value of 𝜇, the relative importance of the time dimension and frequency dimension in measuring the strength of the interaction between MU and ES can be changed.

[0073] This ensures that social trust remains within an effective range, directly reflecting the closeness of the trust relationship between MU and ES, and providing an important reference for task offloading decisions.

[0074] This controls the overall delay in task completion, ensuring that tasks can be completed within an acceptable timeframe for the user and guaranteeing a positive user experience.

[0075] The transmission power is limited to prevent excessive power consumption of the device or interference with other devices, and is also limited by the device's hardware capabilities.

[0076] This constrains transmission energy consumption, matching it with the equipment's energy supply and energy-saving requirements, and avoids excessive energy consumption affecting the equipment's battery life.

[0077] This ensures that the utility of the MU is positive, guaranteeing that the MU can gain actual benefits from participating in task unloading and resource allocation; otherwise, the MU may not choose this unloading strategy.

[0078] For Elasticsearch, the optimization problem is: The constraints include: This indicates that the proportion of tasks that ES is offloading to the cloud is reasonable. When the value is 0, Elasticsearch does not unload tasks to the cloud and processes them entirely locally; when the value is 1, all tasks are unloaded to the cloud, and the unloading ratio is adjusted within this range based on its own resources and task situation.

[0079] This limits the pricing range of ES for MU's computing resources. To ensure that the price of ES is not so low as to fail to cover costs, This prevents excessively high pricing from causing MU to choose other ES or not uninstall the task.

[0080] Ensure that the local computing power consumption of ES is within an acceptable range. Due to limitations such as ES hardware performance and heat dissipation, excessive power consumption may affect the stability and lifespan of the equipment.

[0081] Ensuring that the utility of ES is positive incentivizes ES to actively participate in task processing and resource allocation; if the utility is negative, ES will lack the motivation to participate.

[0082] For the cloud, the optimization problem is: The constraints include: Limit the pricing range for cloud computing resources for Elasticsearch. Too low a threshold will affect cloud profitability, while too high a threshold may cause Elasticsearch to reduce the offloading of tasks to the cloud. A balance needs to be found within this range to maximize cloud utility.

[0083] This indicates that the cloud's local computing energy consumption is within a reasonable range. However, cloud computing resources are also limited, and excessive energy consumption will affect its overall operating efficiency and cost control.

[0084] This ensures that the cloud can derive positive utility from task processing and resource allocation, enabling it to provide stable computing services to Elasticsearch (ES) and Virtual Machine (MU). By setting these optimization problems and constraints, the feasibility and rationality of resource allocation and task offloading strategies in real-world scenarios are ensured.

[0085] Solving the optimization problem: The optimization problem is solved using a two-layer Stackelberg game framework and backward induction.

[0086] like Figure 4 As shown, in the first layer of the game, the edge server, acting as the leader, sets the price. Mobile user MU, as a follower, maximizes its own utility function. about Find the first derivative ,when At that time, utility function There exists a maximum value. Let By combining different conditions, the optimal unloading strategy for MU 𝑖 can be obtained under three different scenarios. Considering transmission power, delay, and energy consumption constraints.

[0087] like ,illustrate exist The unloading strategy is monotonically decreasing, and the optimal unloading strategy is: That is, MU does not unload tasks to ES.

[0088] like ,illustrate exist The top is monotonically increasing, and the optimal unloading strategy is: This means that MU will unload all tasks to ES 𝑘.

[0089] like and By solving We can obtain:

[0090] in:

[0091] ,

[0092]

[0093] This categorized discussion approach can accurately determine the optimal unloading strategy for MU based on different system parameters and conditions.

[0094] In the second layer of the game, the cloud, as the leader, sets the price. The optimal unloading strategy for MU is obtained. Then, the edge server acts as a follower and implements the optimal offloading strategy. Substitute into the utility function of ES:

[0095] right about Finding the first derivative, although it is difficult to express the optimal pricing strategy precisely with a mathematical formula, is still a process. However, according to This constraint can be solved using a linear search algorithm. The linear search algorithm works by continuously trying different... Value, calculate the corresponding And compare the sizes, and finally find the one that makes The optimal pricing strategy for the largest ES .

[0096] Sure Then, on about Find the first derivative ,make The optimal unloading strategy is obtained by solving the problem. Considering its own computing resources and cost constraints, the formula is:

[0097] in:

[0098]

[0099]

[0100] By solving this step-by-step method, we can obtain the optimal pricing and offloading strategies for ES under different conditions.

[0101] Optimal Unloading Strategy for ES Substitute the utility function of the cloud:

[0102] right about Find the first derivative ,make The optimal pricing strategy for cloud computing can be obtained by solving the problem. The formula is as follows:

[0103] in:

[0104]

[0105]

[0106] By solving for the optimal pricing strategy for the cloud, the utility of the cloud in the entire system can be maximized.

[0107] By using backward induction to find the equilibrium solution of the two-layer game, the optimal strategies for mobile users, edge servers, and the cloud are obtained.

[0108] The curves showing the change in mobile user utility with the mobile user offload ratio under different computing tasks are as follows: Figure 5 As shown; the curves of mobile user utility changing with computing power under different computing tasks are as follows. Figure 6 As shown; the utility of edge servers varies with edge server pricing under different offloading ratios. The change curve is as follows Figure 7 As shown; the effectiveness of cloud servers varies with cloud server pricing under different user uninstallation rates. The change curve is as follows Figure 8 As shown; a comparison of different solutions for mobile user utility under different user uninstallation rates, as follows: Figure 9 As shown; cloud server utility is priced differently for edge servers. The following compares different solutions as follows: Figure 10 As shown.

[0109] In terms of system composition, a cloud-edge-device collaborative mobile social network is adopted, consisting of mobile users, edge servers, and cloud servers. Mobile users generate computing tasks, but their computing resources are limited by physical size and battery capacity, making it difficult to meet the ever-increasing computing demands. Edge servers are deployed close to mobile users, possessing certain computing and storage resources, and can receive tasks offloaded by mobile users. Cloud servers have powerful computing capabilities and can provide auxiliary computing support to edge servers.

[0110] Regarding the trust assessment mechanism, to ensure the security of task unloading, a trust assessment mechanism based on the historical interaction records between mobile users and edge servers is proposed. This mechanism comprehensively considers multiple indicators such as direct trust, indirect trust, intimacy, and contribution history. Direct trust is calculated by weighted averaging based on users' ratings of the edge server's service quality for each instance, taking into account the timeliness of historical ratings and using an exponential decay function for weighting. Indirect trust is based on the social relationships and similarities among mobile users, measuring the reliability of recommendations from other users to calculate the indirect trust value. Intimacy comprehensively measures the strength of the interaction relationship between mobile users and edge servers from two dimensions: service time and interaction frequency. Contribution history reflects the service contribution of edge servers in historical collaborations by statistically analyzing the scale of computing resources provided to mobile users. Finally, these indicators are weighted and summed to obtain the social trust score. Mobile users select edge servers with high trustworthiness for task unloading based on the social trust score, effectively filtering malicious edge servers and ensuring network security and resource allocation security.

[0111] In terms of semantic transmission, semantic transmission technology is introduced to improve the data transmission efficiency between mobile users and edge servers. Traditional communication technologies suffer from low efficiency and high system overhead when handling intensive communication needs, while semantic communication technology transmits more information within the same communication bandwidth by transmitting semantic information rather than raw data. This invention defines a text semantic task for mobile users, uses a deep learning-driven semantic communication model to extract task semantics, and offloads it to the edge server. The semantic transmission rate is determined by calculating the signal-to-interference-plus-noise ratio (SINR) and combining it with semantic similarity indicators. Simultaneously, to ensure accurate recovery of the offloaded task, a minimum semantic similarity accuracy requirement is set for the semantic representation, thereby improving communication efficiency while reducing transmission latency and system overhead.

[0112] For the task offloading and incentive mechanism, an incentive-oriented two-layer task offloading scheme was designed, and a Stackelberg game model was constructed to optimize the strategies of each party. In the first layer, mobile users select trusted edge servers to offload part of their workload based on semantic transmission. In the second layer, when the edge server's computing resources are insufficient, part of the workload is offloaded to the cloud server. To maximize the interests of each party, utility functions for mobile users, edge servers, and cloud servers were constructed. The mobile user's utility function comprehensively considers the satisfaction of completing the workload, the price paid to the edge server, local computing costs, and the cost of offloading the workload. The edge server's utility function consists of the revenue from selling computing resources to mobile users minus its own computing energy consumption and the cost of purchasing resources from the cloud server. The cloud server's utility function is the revenue from selling computing resources to the edge server minus its local computing energy consumption. By modeling the task offloading process as a Stackelberg game, each party optimizes its own strategy based on the game outcome. Mobile users determine the optimal offloading ratio based on the pricing of edge servers. Edge servers determine their own offloading ratio and pricing to mobile users based on the offloading strategies of mobile users and the pricing of cloud servers. Cloud servers determine the optimal pricing strategy based on the offloading needs of edge servers, ultimately achieving efficient allocation of system resources and maximizing the utility of all parties.

Claims

1. A method for mobile social network resource allocation in cloud-edge collaboration, characterized in that, The method comprises the following steps: Considering the historical interaction records of the MU and the ES, the social trust degree is calculated by calculating the direct trust degree, the indirect trust degree, the intimacy degree and the contribution history index, and the MU selects the trusted ES according to the social trust degree when offloading tasks; According to the signal-to-interference-plus-noise ratio and the semantic similarity, the semantic transmission rate is determined in combination with the bandwidth and the average semantic information amount. According to the local calculation time, the transmission duration calculated by the semantic transmission rate, the offloading time, the uploading and calculation time, the overall delay of the MU in completing the task, the delay of the MU under different task offloading strategies is accurately evaluated, the transmission energy consumption is evaluated according to the necessary transmission power and the transmission energy consumption, and the transmission power and the transmission time are considered, and the energy consumption of different devices in the calculation process is comprehensively considered according to the local calculation energy consumption of the MU, the total workload of the ES k, the local calculation delay, the local calculation energy consumption, the local calculation delay of the cloud and the calculation energy consumption of the cloud. The utility function of the MU, the utility function of the ES and the utility function of the cloud are constructed, and the corresponding optimization problem and constraint condition are constructed. The constructed optimization problem is solved by using a two-layer Stackelberg game framework and a backward induction method. In the first layer game, the edge server sets the price as the leader, and the mobile user MU sets the price as the follower by maximizing the utility function of the MU, and the optimal offloading strategy of the MU i is obtained by combining different conditions. In the second layer game, the cloud sets the price as the leader after obtaining the optimal offloading strategy of the MU, and the edge server sets the price as the follower, and the optimal offloading strategy of the MU is substituted into the utility function of the ES, and the optimal pricing strategy and the optimal offloading strategy of the ES are obtained by solving the constraint condition. The optimal offloading strategy of the ES is substituted into the utility function of the cloud, and the optimal pricing strategy of the cloud is obtained by solving, and the equilibrium solution of the two-layer game is obtained by using the backward induction method, and the optimal strategy of the mobile user, the edge server and the cloud is obtained, and the perception task of the MU is offloaded to the trusted ES, and the cloud server is requested to assist when the edge server is short of computing resources.

2. The cloud-edge-end collaborative mobile social network resource allocation method of claim 1, wherein, The direct trust degree is calculated based on the historical interaction score of the MU to the ES.

3. The cloud-edge-end collaborative mobile social network resource allocation method of claim 1, wherein, The indirect trust degree is calculated based on the recommendation information and the social relationship of other MUs.

4. The cloud-edge-end collaborative mobile social network resource allocation method of claim 1, wherein, The contact intimacy degree index is calculated based on the intimacy degree from the time and frequency dimensions to measure the interaction relationship strength between the MU and the ES.

5. The cloud-edge-end collaborative mobile social network resource allocation method of claim 1, wherein, The utility function of the MUs The formula is as follows: wherein, is a satisfaction adjustment parameter, is a maximum delay, is a local computation time, is an upload and computation time, is a price paid to ES k, is a semantic transmission cost parameter, is a local computation cost parameter, is a transmission energy consumption, is a total workload.

6. The cloud-edge-end collaborative mobile social network resource allocation method of claim 1, wherein, The utility function of the ES The formula is as follows: wherein, is the benefit obtained from the MU, is the proportion offloaded to the cloud, is the total workload, is the local computing cost parameter, is the local computing energy consumption, is the price paid to the cloud.

7. The cloud-edge-end collaborative mobile social network resource allocation method of claim 1, wherein, The utility function of the cloud The formula is as follows: wherein, is the revenue obtained from the ES, is the proportion of the ES offloaded to the cloud, is the total workload, is the local computing cost parameter, is the computing energy consumption of the cloud.

Images