A self-organizing edge user allocation system and method based on reconfigurable smart surfaces

Abstract

This invention discloses a self-organizing edge user allocation system based on a reconfigurable smart surface. The system includes a user layer and a service layer; it also includes a reconfigurable smart surface layer. The user layer sends task requests from devices in the network. The reconfigurable smart surface layer processes the requested tasks through a fast response channel to obtain a queue of request tasks with optimized processing times. The service layer uses a maximum independent set user allocation algorithm to process the task with the most requests in the queue with optimized processing times and outputs an optimal task selection and processing matching strategy. This invention significantly improves the latency of intelligent network data transmission and processing by reconfiguring the computing network topology based on the needs of different users and business scenarios through the reconfigurable smart surface layer.

Description

Technical Field

[0001] This invention belongs to the field of improved intelligent communication network technology, and particularly relates to a self-organizing edge user allocation system and method based on reconfigurable intelligent surfaces. Background technology:

[0002] With the rapid development of the Internet of Things (IoT), big data, and cloud computing, future network environments will face increased device connections and data traffic. Driven by this trend, 6G networks have become a focal point, with key requirements including low latency, high throughput, and wide coverage. Computational networks, as a new network framework, aim to provide users with efficient and adaptable services by integrating different resources. However, in intelligent communication networks, the path loss issues in millimeter wave and Asia-Pacific Hertz bands make resource optimization in computing networks more difficult, and these problems have gradually become bottlenecks for the development of many technologies, especially in some service-intensive application scenarios.

[0003] To address the aforementioned issues, reconfigurable smart surfaces, as a candidate technology for optimizing the topology of next-generation communication networks, can provide an effective solution.

[0004] Disadvantages of existing technology

[0005] 1. Severe path loss in wireless signal frequency bands leads to unstable communication links. Although multiple resources are integrated within the computing network framework to provide better services, resource allocation and topology optimization still face challenges due to the uncertainty of wireless signal quality. This uncertainty can lead to network performance degradation because signal quality may fluctuate at different times and locations, resulting in network connectivity instability and service quality fluctuations.

[0006] 2. The issue of user allocation effectiveness. While reconfigurable smart surfaces offer a novel approach to improving network topology, how to most effectively utilize them for user allocation and optimization to meet the specific needs of different services and users remains an unresolved challenge when dealing with business-intensive demands. Therefore, the diversity and dynamism of services may limit the performance of existing technologies in highly flexible user allocation management. Summary of the Invention

[0007] To address the technical problems existing in the prior art, this invention provides a self-organizing edge user allocation system and method based on a reconfigurable smart surface. This invention uses a reconfigurable smart surface layer to reconstruct the topology of the computing network based on the needs of different users and business scenarios, thereby significantly improving the latency of data transmission and processing in the intelligent network.

[0008] To address the problems of the existing technology, the present invention adopts the following technical solution:

[0009] A self-organizing edge user allocation system based on a reconfigurable smart surface, comprising a user layer and a service layer; the self-organizing edge user allocation system further comprises a reconfigurable smart surface layer; wherein:

[0010] The user layer is used to send device task requests in the network;

[0011] The reconfigurable smart surface layer obtains a request task queue with optimized processing time through a fast response channel;

[0012] The service layer uses the maximum independent set user allocation algorithm to output the optimal task selection and processing matching strategy for the task with the most requests in the time-optimized task queue.

[0013] Furthermore, the reconfigurable smart surface layer obtains an optimized time for the request task queue process through a fast response channel, including:

[0014] The direct channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula:

[0015]

[0016] In the formula, B represents the bandwidth between the two edge devices, and σ 2 It is the variance of Gaussian noise, P n h is the transmit power of the nth device. n，m It is the channel coefficient from the device to the server;

[0017] The indirect channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula, including:

[0018]

[0019] In the formula, θ i To adjust the reflection amplitude and phase shift at the reconfigurable smart surface of the i-th RIS, the channel coefficients from the n-th device to the associated m-th server, from the n-th device to the i-th RIS, and from the i-th RIS to the associated m-th server are respectively represented by h. n，m h i，n ∈C K×1 and h i，m ∈C K×1 Indicates; σ 2 P represents the variance of Gaussian noise. n It is the transmit power of the nth device;

[0020] The response time channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula:

[0021]

[0022] In the formula, b m ={0,1} represents the reflection link index between the server and the scheduling user through the reconfigurable smart surface. When it is 1, it means that the communication link is assisted by the reconfigurable smart surface, and when it is 0, it means that the communication link is assisted by the reconfigurable smart surface.

[0023] The edge execution time channel between the m-th server in the service layer and the associated n-th device is constructed according to the following formula:

[0024]

[0025] In the formula, a m = {0, 1} represents the edge execution metric, where 1 indicates that the task is being processed on the master server, and 0 indicates otherwise; Q s L represents the number of CPU cycles required to process each unit byte. n，s Indicates the size of the data uploaded by the business, f m，1 f m，0 Specify the dominant and non-dominant servers.

[0026] Furthermore, the service layer employs a maximum independent set user allocation algorithm to output the optimal task selection and processing matching strategy for the task with the highest number of requests in the time-optimized request task queue. This process includes:

[0027] 301. Construct an undirected graph G(V, E) of the request task response based on the user-layer allocated environment output optimization time request task queue, where each vertex V = {v1, v2, v3}. n , ..., v n} is a set of n network nodes, where each node represents a user service;

[0028] 302. Establish a subset of optimal resource matching strategies according to the following formula:

[0029]

[0030] In the formula: β is a weighting parameter used to balance the trade-off between the two objectives of minimum task processing time and maximum task completion amount;

[0031] 303. Determine whether the corresponding undirected graph node of the request task falls into the subset of the optimal resource matching strategy; if it does, output the optimal task selection and processing matching strategy, otherwise return 301.

[0032] To address the problems of the prior art, the present invention also adopts the following technical solution:

[0033] A self-organizing edge user allocation method based on reconfigurable smart surfaces includes the following steps:

[0034] The user layer sends device task requests in the network to the reconfigurable smart surface layer;

[0035] The reconfigurable smart surface layer obtains an optimized time queue of request tasks through a fast response channel, following these steps:

[0036] The direct channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula:

[0037]

[0038] In the formula, B represents the bandwidth between the two edge devices, and σ 2 It is the variance of Gaussian noise, P n h is the transmit power of the nth device. n，m It is the channel coefficient from the device to the server;

[0039] The indirect channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula, including:

[0040]

[0041] In the formula, θ i To adjust the reflection amplitude and phase shift at the reconfigurable smart surface of the i-th RIS, the channel coefficients from the n-th device to the associated m-th server, from the n-th device to the i-th RIS, and from the i-th RIS to the associated m-th server are respectively represented by h. n，m h i，n ∈C K×1 and h i，m ∈C K×1 Indicates; σ 2 P represents the variance of Gaussian noise. n It is the transmit power of the nth device;

[0042] The response time channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula:

[0043]

[0044] In the formula, b m ={0,1} represents the reflection link index between the server and the scheduling user through the reconfigurable smart surface. When it is 1, it means that the communication link is assisted by the reconfigurable smart surface, and when it is 0, it means that the communication link is assisted by the reconfigurable smart surface.

[0045] The edge execution time channel between the m-th server in the service layer and the associated n-th device is constructed according to the following formula:

[0046]

[0047] In the formula, a m = {0, 1} represents the edge execution metric, where 1 indicates that the task is being processed on the master server, and 0 indicates otherwise; Q s L represents the number of CPU cycles required to process each unit byte. n，s Indicates the size of the data uploaded by the business, f m，1 f m，0 Specify the dominant and non-dominant servers;

[0048] The optimal task selection and processing matching strategy is output based on the user allocation algorithm of the maximum independent set, which selects the task with the most requests in the time-optimized task queue.

[0049] 301. Construct an undirected graph G(V, E) of request task responses based on the user-layer allocated environment output optimization time request task queue, where each vertex V = {v1, v2, v3}. n , ..., v n} is a set of n network nodes, where each node represents a user service;

[0050] 302. Establish a subset of optimal resource matching strategies according to the following formula:

[0051]

[0052] In the formula: β is a weighting parameter used to balance the trade-off between the two objectives of minimum task processing time and maximum task completion amount;

[0053] 303. Determine whether the corresponding undirected graph node of the request task falls into the subset of the optimal resource matching strategy; if it does, output the optimal task selection and processing matching strategy, otherwise return 301.

[0054] Beneficial effects

[0055] 1. By reconfiguring the radio propagation environment using reconfigurable smart surface technology, new access options are provided between edge servers and end users during computing task transmission. This invention proposes a self-organizing edge technology solution based on reconfigurable smart surfaces. In general, reconfigurable smart surfaces reconfigure the computing network topology according to the needs of different users and business scenarios, thereby significantly reducing data transmission and processing latency.

[0056] 2. To ensure adaptability to user service needs, this invention proposes a novel user access allocation solution for self-organizing edge scenarios assisted by reconfigurable smart surfaces. Existing network resource allocation methods often cannot adequately adapt to special service requirements such as service-intensive scenarios. The technology of this invention can adjust the matching strategy between users and servers according to the needs of different services through the link transmission of reconfigurable smart surfaces, thereby achieving a high degree of adaptability to user allocation based on service requirements.

[0057] 3. Furthermore, this invention takes user allocation in demand-intensive scenarios as an example, using graph theory algorithms to match users and resources. This helps to seamlessly integrate new users and resources into the system, thereby achieving optimization across a wide range of scenarios. Compared with traditional user allocation techniques, the technical effectiveness and advantages of this solution are evident.

[0058] 4. This invention evaluates different methods, including random allocation, greedy allocation, graph-based dynamic programming allocation, and RS. 3 The methods were described, and their impact on system throughput and user allocation distribution was analyzed. Figure 3 The system completion rate of user services is demonstrated under different time thresholds. When the time threshold is small, RIS-assisted AHE can still effectively allocate user services. When the system time threshold is approximately 200 seconds, the allocation efficiency of the method in this invention is approximately 1.5 times that of greedy and random methods. With the help of RIS, the system throughput significantly improves as the time threshold gradually increases. The server's completion rate for user services remains around 80%, reaching a maximum of 94%. This result can be attributed to the fact that RIS provides more user allocation access options in the AHE scenario. This indicates that with the help of RIS, the server handles more dominant services, reducing the time wasted on cache loading. This greatly improves the system throughput. Attached Figure Description

[0059] Figure 1 This invention relates to a reconfigurable smart surface-assisted user allocation architecture in computing power networks.

[0060] Figure 2 This is an example of a graph-based scheduling service on a server according to the present invention.

[0061] Figure 3 It is a comparison of user service completion status.

[0062] Figure 4 The actual user's location information and the types of their needs.

[0063] Figure 5 The impact of different intervals on user allocation. Detailed Implementation

[0064] The following is in conjunction with the appendix Figure 1-5 The present invention is described as follows:

[0065] 1. System Model

[0066] like Figure 1 As shown, a self-organizing edge user allocation system based on a reconfigurable smart surface is disclosed. The self-organizing edge user allocation system includes a user layer and a service layer; the system also includes a reconfigurable smart surface layer (RISS). Specifically: the user layer sends task requests from devices in the network; the reconfigurable smart surface layer obtains a queue of request tasks with optimized processing time through a fast response channel; the service layer uses a maximum independent set user allocation algorithm to process the task with the highest number of requests in the optimized time request task queue and outputs an optimal task selection and processing matching strategy.

[0067] This invention proposes a RAHE scenario assisted by multiple RISs. This user allocation architecture consists of a RIS-assisted communication layer and a graph topology-supported edge computing layer. The server's location in the system is arbitrary, with each RIS assisting only one server. This invention specifies the server, device, and RISs as M = {1, 2, ..., M}, N = {1, 2, ..., N}, and I = {1, 2, ..., l}, respectively. Each user will randomly generate a service request, with the service type determined by the set S = {1, 2, ..., S}. Assume a user has one computing service, and each service is handled by a specific master server. Let J... n，s =(L n，s Q n，s L represents the service of user n, where L n，s Q represents the size of the data uploaded by the business. n，s This represents the total number of CPU cycles required to complete the task. Here, the invention assumes that all servers and devices are equipped with an antenna, and the RIS has K reflecting elements.

[0068] Communication channels are based on block fading, meaning that the channel coefficient remains constant during the current block of data transmission but may change in different time slots. Therefore, the channel coefficients from the nth device to the associated mth server, from the nth device to the i-th RIS, and from the i-th RIS to the associated m-th server are respectively represented by h. n，m h i，n ∈C K×1 and h i，m ∈C K×1 This means that the server fully understands the channel state information between all communication links to execute resource management strategies in order to maximize computing and communication efficiency.

[0069] Typically, system servers only cache programs required for specific service types. Processing new services can lead to significant delays due to loading. Therefore, servers have defined characteristics specific to certain services. In such cases, it's best to forward user services to the primary server. Although this process introduces additional communication latency due to forwarding, RIS can significantly reduce the communication burden.

[0070] 2. Calculation Model

[0071] In the reconstructed intelligent surface-assisted self-organizing edge RAHE resource integration layer, for the direct channel between the m-th server and the associated n-th device, the offloading rate is:

[0072]

[0073] In the above formula, B represents the bandwidth between the two edge devices, σ 2 It is the variance of Gaussian noise, P n h is the transmit power of the nth device. n，m This is the path coefficient from the device to the server. If the direct link between the server and the device is too weak due to long propagation distances (i.e., severe path loss in millimeter waves or Asia-Pacific Hertz bands) or obstacles such as surrounding vehicles and buildings, a RIS (Indirect Link Assist) is dispatched to enhance end-to-end signal quality. In this case, the indirect path between the server and the device assisted by the RIS is:

[0074]

[0075] In the formula, θ i To adjust the reflection amplitude and phase shift at the reconfigurable smart surface of the i-th RIS, the channel coefficients from the n-th device to the associated m-th server, from the n-th device to the i-th RIS, and from the i-th RIS to the associated m-th server are respectively represented by h. n，m h i，n ∈C K×1 and h i，m ∈C K×1 Indicates. σ 2 P represents the variance of Gaussian noise. n This is the transmit power of the nth device. Furthermore, for simplicity, the RIS hardware here supports ideal continuous phase shift (i.e., precise phase shift adjustment of the signal reflected towards the intended user). Therefore, the upload time calculation formula for the user computing task is as follows, where b m ={0, 1} represents the reflection link indicator between the server and the scheduling user via the reconfigurable smart surface. A value of 1 indicates communication via the reconfigurable smart surface-assisted communication link, while a value of 0 indicates otherwise.

[0076] The response time channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula:

[0077]

[0078] The edge execution time channel between the m-th server in the service layer and the associated n-th device is constructed according to the following formula:

[0079]

[0080] In this invention, neighboring reconfigurable smart surfaces act as relays for the corresponding servers, helping to improve their communication performance and providing an alternative access option for users at designated locations within the network. Furthermore, each user's computational tasks can be delegated to either the dominant server or the nearest non-dominant server. Because each server has a basic program to perform its functions, the non-dominant server should have a configured latency, such as a caching latency or a waiting period, which is achieved in this invention by setting different processing rates f. m，1 f m，0 Specify the dominant and non-dominant servers, where a m ={0, 1} represents the edge execution metric; a value of 1 indicates that the task is being processed on the master server, and a value of 0 indicates otherwise. Q s L represents the number of CPU cycles required to process each unit byte. n，s This indicates the size of the data uploaded by the business.

[0081] This invention investigates the user allocation problem in a RAHE scenario. The services allocated to each server are divided into two parts: one part for dominant services and the other for non-dominant services. The objective of this invention is to maximize the total number of services provided by the CPN while satisfying service latency requirements. The total delay for the server to complete the service is T. m，n，s (j)=∑(t s (j)+t tr (j)), the service completion time m for each server can be represented as T. t (j)=∑(x n *T m，n，s (j)). Where x n The value is 0 or 1, indicating whether the service is complete, thus determining the total number of transactions U processed by server m. m，n =∑(x n |T t <T max ), T max This represents the maximum time limit that the system can accept.

[0082] 3. User allocation algorithm based on maximum independent set

[0083] like Figure 2 As shown, the service layer uses the maximum independent set user allocation algorithm to optimize the processing of the task queue with the most requests and output the optimal task selection and processing matching strategy.

[0084] (1) Combinatorial optimization

[0085] Based on the server's attributes, this invention transforms the problem into optimal interval scheduling, considering users who need to use the service of a specific machine (server) within a certain period. The latency tolerance threshold T for each user is defined. m，n，s Change to the start time ST specified by the business m，n，s and end time ET m，n，s At the same time, these services should be processed without interruption. Therefore, the user allocation problem becomes a problem of finding a solution to select a subset of servers that ensures (1) there are no time conflicts between user services; (2) all services are processed as much as possible; and (3) the total user latency is kept as low as possible.

[0086] Generally, user scheduling can be viewed as the process of selecting a suitable server for each user. Under hard deadlines, the service provided by each user should satisfy the following equation: Suppose the server wants to serve set J n，s The allocation of users in the middle means Each user can only be assigned to one server, representing

[0087] User assignment in the RAHE scenario involves finding the largest possible set of non-overlapping services. By transforming this problem into a graph model, the user assignment objective becomes finding the largest subset of vertices without edges between them, which is equivalent to the maximum independent set problem.

[0088] Figure 2 Conflict graphs are used to illustrate conflicts between services. This problem can be represented by an undirected graph G(V, E), called an assignment graph. Each vertex V = {v1, v2, v3} n , ..., v n Let be a set of n network nodes, where each node represents a user service. If a temporary conflict exists between two services, an edge will connect the affected nodes. Therefore, the maximum amount of business that a server can handle corresponds to the maximum number of independent edges in the corresponding undirected graph. According to the definition of a maximum independent set, the optimal number of nodes (tasks) is selected under given constraints.

[0089] (2) Interval scheduling strategy based on integer linear programming

[0090] This invention utilizes Integer Linear Programming (ILP) to solve the weighted maximum independent set problem, combined with a combinatorial optimization objective function. The ILP framework readily accepts different constraints and objective functions within a graph structure. The mathematical expression of the problem is as follows:

[0091]

[0092] binary variable x i Indicates whether service i is included in the optimal subset U, and x i +x j <1 indicates that two adjacent nodes cannot appear in an independent set at the same time. β is a weighting parameter proposed in this invention to balance the trade-off between the two objectives.

[0093] The initial steps of Algorithm 1 are designed to be based on Figure 2 The system model configuration environment is given in the text. First, the first line of the input represents the initial number of users and service types in the system environment. Then, the second and third lines express the problem to be solved using a weighted average of the service subset size and total completion time as the objective function. The fourth line defines whether the nodes of the graph structure are in the maximum independent set to be solved. The fifth line indicates that the set of tasks completed by the system is the maximum independent set required by the algorithm, and the sixth line defines x... i +x j A value less than 1 indicates that two adjacent nodes cannot both appear in the same independent set, corresponding to a constraint of no service overlap in the system. Then, the ILP problem is formulated using an objective function. After determining the upper and lower bounds of L, this invention performs linear programming relaxation on the objective function value L through lines seven and eight to identify the largest independent set. Finally, the optimal solution is continuously updated. Therefore, for each check value of L, this invention executes the procedure at most N(N-1) times.

[0094]

[0095] This invention evaluates different methods, including random allocation, greedy allocation, graph-based dynamic programming allocation, and RS3 method, and analyzes their impact on system throughput and user allocation distribution. Figure 3The system completion rate of user services is demonstrated under different time thresholds. When the time threshold is small, RIS-assisted AHE can still effectively allocate user services. When the system time threshold is approximately 200 seconds, the allocation efficiency of the method in this invention is approximately 1.5 times that of greedy and random methods. With the help of RIS, the system throughput significantly improves as the time threshold gradually increases. The server's completion rate for user services remains around 80%, reaching a maximum of 94%. This result can be attributed to RIS providing more user allocation access options in the AHE scenario. This indicates that with the help of RIS, the server handles more dominant services, reducing the time wasted on cache loading. This greatly improves the system throughput. Regarding the improvement of user demand distribution, this invention compares the distribution of users on each server under different methods. Figure 4 This describes the user location information of the actual dataset. Because user locations and their service attributes are distributed extremely unevenly across the dataset, servers in certain regions may experience a surge of service requests. Figure 5 The left side shows that Edge Server 3 is under greater load than other servers. This server handles two to three times the number of service requests as the other servers. Figure 5 This invention demonstrates that the RS3 scheme can significantly impact service load balancing. In the RAHE scenario, the server with the heaviest workload experiences at least 30% load relief. While the GraphDP algorithm can also improve system throughput, its load balancing capabilities are significantly insufficient. The method of this invention can improve system service completion while managing the chaotic topology at the user layer. This enhancement significantly reduces server load imbalance caused by the service distribution density in the real-world dataset. This proves the superiority of RS3 in alleviating server load.

[0096] Although the present invention has been described above, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many modifications under the guidance of the present invention without departing from the spirit of the present invention, and these modifications are all within the protection scope of the present invention.

Claims

1. A self-organizing edge user allocation system based on a reconfigurable smart surface, the self-organizing edge user allocation system comprising a user layer and a service layer; characterized in that: The self-organizing edge user allocation system further includes a reconfigurable smart surface layer; wherein: The user layer is used to send device task requests in the network; The reconfigurable smart surface layer obtains an optimized time queue for request task processing through a fast response channel; including: The direct channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula: In the formula, This represents the bandwidth between two edge devices. It is Gaussian noise variance. It is the transmit power of the nth device; The indirect channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula, including: In the formula, To adjust the reflection amplitude and phase shift at the reconfigurable smart surface of the i-th RIS, the channel coefficients from the n-th device to the associated m-th server, from the n-th device to the i-th RIS, and from the i-th RIS to the associated m-th server are respectively represented by... , and express; The response time channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula: In the formula, This represents the reflection link between the server and the scheduling user via the reconfigurable smart surface. A value of 1 indicates communication via the reconfigurable smart surface, while a value of 0 indicates communication via the auxiliary communication link. The edge execution time channel between the m-th server in the service layer and the associated n-th device is constructed according to the following formula: In the formula, This indicates the edge execution metric; a value of 1 means that tasks are being processed on the master server, and a value of 0 means otherwise. This indicates the number of CPU cycles required to process each unit byte. Indicates the size of the data uploaded by the business. , Specify the dominant and non-dominant servers; The service layer uses a maximum independent set user allocation algorithm to output the optimal task selection and processing matching strategy for the task with the most requests in the time-optimized request task queue, including:

301. Construct an undirected graph of request task responses based on the request task queue that optimizes the output time of the user-level allocated environment. G (V, E) Each vertex V = { , , …, }yes n A set of network nodes, where each node represents a user service, and if there is a temporary conflict between two services, the edge will connect the affected node; 302、 The optimal resource matching strategy subset is established according to the following formula: In the formula: β is a weighting parameter used to balance the trade-off between the two objectives of minimum task processing time and maximum task completion amount; This represents the latency tolerance threshold for each user; Serving binary variable representation i Is it included in the optimal subset? U middle; 303. Determine whether the corresponding undirected graph node of the request task falls into the subset of the optimal resource matching strategy; if it does, output the optimal task selection and processing matching strategy, otherwise return 301.

2. A self-organizing edge user allocation method based on reconfigurable smart surfaces, characterized in that, The self-organizing edge user allocation method adopts the self-organizing edge user allocation system based on reconfigurable smart surfaces as described in claim 1, and includes the following steps: The user layer sends device task requests in the network to the reconfigurable smart surface layer; The reconfigurable smart surface layer obtains an optimized time queue of request tasks through a fast response channel, following these steps: The direct channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula: In the formula, This represents the bandwidth between two edge devices. It is Gaussian noise variance. It is the transmit power of the nth device; The indirect channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula, including: In the formula, To adjust the reflection amplitude and phase shift at the reconfigurable smart surface of the i-th RIS, the channel coefficients from the n-th device to the associated m-th server, from the n-th device to the i-th RIS, and from the i-th RIS to the associated m-th server are respectively represented by... , and express; The response time channel between the m-th server and the associated n-th device in the service layer is constructed according to the following formula: In the formula, This represents the reflection link between the server and the scheduling user via the reconfigurable smart surface. A value of 1 indicates communication via the reconfigurable smart surface, while a value of 0 indicates communication via the auxiliary communication link. The edge execution time channel between the m-th server in the service layer and the associated n-th device is constructed according to the following formula: In the formula, This indicates the edge execution metric; a value of 1 means that tasks are being processed on the master server, and a value of 0 means otherwise. This indicates the number of CPU cycles required to process each unit byte. Indicates the size of the data uploaded by the business. , Specify the dominant and non-dominant servers; The optimal task selection and matching strategy, based on the user allocation algorithm with the largest number of requests in the task queue, includes:

301. Construct an undirected graph of request task responses based on the request task queue that optimizes the output time of the user-level allocated environment. G (V, E) Each vertex V = { , , …, }yes n A set of network nodes, where each node represents a user service. If there is a temporary conflict between two services, the edge will connect the affected node; 302. Establish a subset of optimal resource matching strategies according to the following formula: In the formula: β is a weighting parameter used to balance the trade-off between the two objectives of minimum task processing time and maximum task completion amount; This represents the latency tolerance threshold for each user; Serving binary variable representation i Is it included in the optimal subset? U middle; 303. Determine whether the corresponding undirected graph node of the request task falls into the subset of the optimal resource matching strategy; if it does, output the optimal task selection and processing matching strategy, otherwise return 301.

Images