A method and system for 5G network task offloading and traffic splitting

Abstract

This invention provides a method and system for task offloading and traffic splitting in 5G networks, belonging to the field of wireless communication technology. The method includes: determining all available paths for each network slice; constructing a first matrix and multiple second matrices; optimizing all second matrices with the objective of maximizing the number of nodes included in the selected available paths, resulting in an optimized matrix population; obtaining a node load matrix based on the first matrix and the optimized matrix population; sorting all nodes based on the value of each node load matrix element in the node load matrix, generating an array; traversing each array element to allocate computing resources to the network slice passing through each node, generating a computing resource allocation scheme; and distributing the computing resource allocation scheme to the physical network through an automated interface. This invention optimizes multi-path selection and resource allocation by combining network slicing and intra-network computing, maximizing the participation of network nodes, thereby achieving efficient task offloading and traffic splitting.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a method and system for 5G network task offloading and traffic diversion. Background Technology

[0002] With the rapid development of fifth-generation mobile communication technology (5G), the demand for high speed, low latency, and high reliability in networks is increasing. These demands have driven the application of emerging technologies such as network slicing and intra-network computing, enabling 5G networks to provide customized network resources for different types of services and applications. However, in such a complex and diverse network environment, how to efficiently offload tasks and distribute load has become a pressing technical challenge.

[0003] In the current 5G network architecture, task offloading primarily relies on transferring computing tasks from mobile terminals or edge devices to cloud data centers or edge computing nodes to reduce the computational burden on terminal devices and improve the overall system response speed. However, this approach often focuses on single-path optimization, neglecting multi-path coordination and the full utilization of node resources. This not only limits the full utilization of network resources but may also lead to some nodes being overloaded while other nodes are idle, thus affecting the load balancing and system performance of the entire network.

[0004] Therefore, how to comprehensively consider multi-path selection, node resource utilization, and dynamic network conditions to achieve efficient task offloading and distribution has become an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides a 5G network task offloading and diversion method, system, electronic device, and storage medium to overcome the deficiencies in the prior art and achieve efficient task offloading and diversion by comprehensively considering multi-path selection, node resource utilization, and dynamic network conditions.

[0006] This invention provides a method for 5G network task offloading and traffic splitting, comprising the following steps:

[0007] Determine all available paths for each network slice; each available path contains at least one node.

[0008] Construct a first matrix and multiple second matrices. The first matrix is ​​used to represent the relationship between available paths and nodes. Each second matrix is ​​used to represent the relationship between each network slice and a selected available path. Each network slice can only select at most one available path.

[0009] With the goal of maximizing the number of nodes contained in the selected available paths, optimize all of the second matrix to obtain the optimized matrix population;

[0010] Based on the first matrix and the optimized matrix population, the node load matrix is ​​obtained;

[0011] Based on the value of each node load matrix element in the node load matrix, all nodes are sorted to generate an array;

[0012] Iterate through each element of the array, allocate computing resources to the network slice that passes through each node, and generate a computing resource allocation scheme, wherein each element of the array corresponds to a node;

[0013] The computing resource allocation scheme is distributed to the physical network via an automated interface.

[0014] According to the present invention, a 5G network task offloading and traffic splitting method is provided. In the first matrix, the first matrix element A(i,y)=1 indicates that the i-th available path will pass through node y, and A(i,y)=0 indicates that the i-th available path will not pass through node y. In the second matrix, the second matrix element B(x,i)=1 indicates that network slice x uses available path i, and B(x,i)=0 indicates that network slice x does not use available path i.

[0015] This invention provides a method for 5G network task offloading and traffic splitting, wherein optimizing all second matrices with the objective of maximizing the number of nodes contained in the selected available paths to obtain an optimized matrix population specifically includes:

[0016] The entire second matrix is ​​optimized using a genetic algorithm, and after multiple iterations, the optimized matrix population is obtained.

[0017] This invention provides a method for 5G network task offloading and traffic splitting, wherein the genetic algorithm specifically includes:

[0018] In each iteration, two second matrices are randomly selected from all the second matrices, and the same rows are randomly selected and swapped to generate new matrix individuals;

[0019] Randomly select a second matrix and randomly select two different rows. Swap the second matrix elements with a value of 1 in the two rows to generate a new matrix individual;

[0020] Using the number of nodes contained in the selected available path as the evaluation criterion, matrix individuals with fitness values ​​higher than a preset threshold are selected to enter the next iteration, wherein the number of nodes contained in the selected available path is directly proportional to the fitness value.

[0021] This invention provides a method for 5G network task offloading and traffic distribution, wherein the optimized matrix population includes multiple optimized second matrices; the step of obtaining a node load matrix based on the first matrix and the optimized matrix population specifically includes:

[0022] Multiple third matrices are generated, each of which is obtained by multiplying the first matrix by each of the optimized second matrices;

[0023] Summing each of the third matrices column-wise yields multiple fourth matrices;

[0024] The node load matrix is ​​determined from all the fourth matrices based on the value of each fourth matrix element in each of the fourth matrices.

[0025] According to the present invention, a method for 5G network task offloading and traffic distribution is provided, wherein determining the node load matrix from all the fourth matrices based on the value of each fourth matrix element in each of the fourth matrices specifically includes:

[0026] When the number of non-zero fourth matrix elements in each of the fourth matrices is inconsistent, the fourth matrix containing the most non-zero fourth matrix elements will be used as the node load matrix.

[0027] When the number of non-zero elements in any two fourth matrices is the same, calculate the variance of each fourth matrix;

[0028] The fourth matrix with the smallest variance is taken as the node load matrix.

[0029] According to the present invention, a method for 5G network task offloading and traffic splitting is provided, wherein each node load matrix element in the node load matrix corresponds to a node, and the value of each node load matrix element represents the number of network slices carried by each node.

[0030] This invention provides a method for 5G network task offloading and traffic distribution, wherein the method involves sorting all nodes based on the value of each node load matrix element in the node load matrix to generate an array, specifically including:

[0031] When the values ​​of each node load matrix element are inconsistent, the array is generated by arranging all the node load matrix elements from left to right in ascending order based on the magnitude of all the node load matrix elements' values;

[0032] When the values ​​of any two nodes' load matrix elements are the same, calculate the energy efficiency of the node corresponding to each node's load matrix element;

[0033] Based on the energy efficiency of all the nodes, the elements of the node load matrix are arranged from left to right in descending order to generate the array.

[0034] This invention provides a method for 5G network task offloading and traffic splitting, wherein traversing each array element in the array, allocating computing resources for network slices passing through each node, and generating a computing resource allocation scheme specifically includes:

[0035] When the value of an array element is zero, skip the node corresponding to that element.

[0036] When the value of an array element is 1, the minimum value between the remaining computing power of the node corresponding to the array element and the remaining computing requirements of the corresponding network slice is allocated to the network slice, and the remaining resources of the node and the network slice are updated.

[0037] When the value of an array element is greater than 1, the remaining computing power of the node corresponding to the array element is evenly distributed among the network slices it carries, and the remaining resources of the node and the network slice are updated.

[0038] According to the present invention, a method for 5G network task offloading and traffic splitting is provided, wherein before determining all available paths for each network slice, the method further includes:

[0039] Calculate all possible paths for each of the network slices;

[0040] Calculate the reliability value and the sum of delays for all possible paths; wherein the reliability value is calculated as the product of the reliability of all traversed nodes and the reliability of the communication link, and the sum of delays is calculated as the sum of the calculated delay of all traversed nodes and the link delay of the communication link;

[0041] Eliminate possible paths where the reliability value and / or the total delay does not meet the preset requirements to obtain all available paths.

[0042] This invention also provides a 5G network task offloading and diversion system, comprising the following modules:

[0043] The first processing module is used to determine all available paths for each network slice; each available path contains at least one node.

[0044] The second processing module is used to construct a first matrix and multiple second matrices. The first matrix is ​​used to represent the relationship between available paths and nodes. Each second matrix is ​​used to represent the relationship between each network slice and a selected available path. Each network slice can only select at most one available path.

[0045] The second processing module is also used to optimize the entire second matrix with the goal of maximizing the number of nodes contained in the selected available path, so as to obtain an optimized matrix population.

[0046] The second processing module is further configured to obtain a node load matrix based on the first matrix and the optimized matrix population;

[0047] The third processing module is used to sort all nodes based on the value of each node load matrix element in the node load matrix and generate an array;

[0048] The computing resource generation module is used to traverse each array element in the array, allocate computing resources to the network slice passing through each node, and generate a computing resource allocation scheme, wherein each array element corresponds to a node;

[0049] The resource distribution module is used to distribute the computing resource allocation scheme to the physical network through an automated interface.

[0050] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the 5G network task offloading and diversion methods described above.

[0051] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the 5G network task offloading and diversion method as described above.

[0052] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described 5G network task offloading and diversion methods.

[0053] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0054] By identifying all available paths for each network slice and ensuring that each available path contains at least one node, the comprehensiveness and flexibility of path selection are improved. This ensures that task offloading can be optimized among multiple paths that meet latency and reliability requirements, enhancing the network's fault tolerance and resource utilization. By constructing a first matrix and multiple second matrices—the first matrix representing the relationship between available paths and nodes, and each second matrix representing the relationship between each network slice and the selected available path, restricting each network slice to selecting at most one available path—a clear mapping of path-node relationships and standardization of network slice path selection are achieved. This ensures the orderly and efficient allocation of resources, facilitating subsequent optimization and management. By optimizing all second matrices with the goal of maximizing the number of nodes contained in the selected available paths, an optimized matrix population is obtained. This fully utilizes the computing resources of more nodes in the network, improving the overall network's load balancing capability and computational efficiency, and avoiding concentrated resource usage. Based on the first matrix and the optimized matrix population, a node load matrix is ​​obtained, accurately reflecting the load of each node during task offloading. This provides reliable data support for subsequent resource allocation and load balancing, ensuring efficient utilization of network resources. By sorting all nodes based on the values ​​of each element in the node load matrix and generating an array, orderly management of network node load is achieved. This facilitates the identification of high and low loads, optimizes resource allocation strategies, and prevents node overload or resource idleness. By traversing each element in the array, computing resources are allocated to network slices passing through each node, generating a computing resource allocation scheme. This ensures the rational allocation and dynamic adjustment of computing resources, optimizing resource usage based on node load, improving task processing efficiency and overall system performance. The computing resource allocation scheme is automatically distributed to the physical network via an interface, enabling rapid deployment and real-time updates of resource allocation. This reduces manual intervention, improves system response speed and management efficiency, and achieves efficient task offloading and distribution. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0056] Figure 1 This is one of the flowcharts illustrating the 5G network task offloading and diversion method provided by the present invention.

[0057] Figure 2 This is the second flowchart of the 5G network task offloading and diversion method provided by the present invention.

[0058] Figure 3 This is the third flowchart of the 5G network task offloading and diversion method provided by the present invention.

[0059] Figure 4 This is the fourth flowchart of the 5G network task offloading and diversion method provided by the present invention.

[0060] Figure 5 This is the fifth flowchart of the 5G network task offloading and diversion method provided by the present invention.

[0061] Figure 6 This is a schematic diagram of the structure of the 5G network task offloading and diversion system provided by the present invention.

[0062] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0064] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0065] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0066] The following is combined with Figures 1-7 This invention describes the 5G network task offloading and diversion method, system, electronic device, and storage medium provided by the present invention.

[0067] Figure 1 This is one of the flowcharts illustrating the 5G network task offloading and traffic diversion method provided by the present invention, such as... Figure 1 As shown, including but not limited to the following steps:

[0068] Step 11: Determine all available paths for each network slice; each available path contains at least one node.

[0069] In step 11, determining all available paths for each network slice is fundamental to the task offloading and load balancing process. In complex 5G network environments, network slices often have different Quality of Service (QoS) requirements, such as latency and reliability. To ensure that each network slice can obtain the optimal transmission path under different conditions, it is first necessary to calculate and filter all available paths that meet these requirements, thus laying the foundation for subsequent path selection and resource allocation.

[0070] The following examples will detail how to determine all available paths for each network slice.

[0071] In one possible implementation, prior to step 11, the method further includes the following steps:

[0072] Calculate all possible paths for each network slice;

[0073] Calculate the reliability value and total delay of all possible paths; the reliability value is calculated as the product of the reliability of all traversed nodes and the reliability of the communication link, and the total delay is calculated as the sum of the calculated delay of all traversed nodes and the link delay of the communication link.

[0074] Eliminate possible paths where the reliability value and / or total latency do not meet the preset requirements to obtain all available paths.

[0075] Specifically, the first step is to establish diverse options for task transmission by calculating all possible paths for each network slice. Specifically, using Dijkstra's algorithm or other suitable path calculation algorithms, such as the A* algorithm, the set of all possible paths from the access point of each network slice to the target cloud data center is calculated. The purpose of this process is to comprehensively explore possible paths in the network to provide sufficient selection space for subsequent filtering. Due to the complexity of the network topology, this step needs to consider path availability and potential network nodes to ensure that different network slices have enough paths to choose from during transmission.

[0076] Next, a detailed reliability and latency assessment is performed on each calculated possible path. Specifically, first, the reliability value of each path is calculated by multiplying the reliability of all traversed nodes and communication links. This effectively assesses the stability and reliability of the path during task transmission. Then, the total latency of the path is calculated by adding the calculated latency of all traversed nodes to the link latency of each communication link. By assessing reliability and latency, a quantitative metric for the quality of service (QoS) can be provided for each path, helping to identify which paths can meet the requirements of a specific network slice.

[0077] Finally, all calculated possible paths are filtered out, eliminating those that do not meet reliability and / or latency requirements, thus obtaining the final set of available paths. The filtering process is based on preset quality of service requirements, aiming to ensure that each available path can meet the reliability and low latency requirements of task transmission. After eliminating unqualified paths, the remaining paths are all available paths that each network slice can use to transmit tasks.

[0078] The effects of this approach are significant. On the one hand, it ensures high reliability and low latency in task transmission, guaranteeing the quality of service for diverse tasks in the 5G network. On the other hand, by eliminating paths that do not meet the requirements, it reduces the complexity of subsequent optimization and resource allocation processes, thereby improving computational efficiency.

[0079] Step 12: Construct a first matrix and multiple second matrices. The first matrix represents the relationship between available paths and nodes, and each second matrix represents the relationship between each network slice and the selected available path. Each network slice can only select at most one available path.

[0080] In step 12, the construction of the first matrix and multiple second matrices is to visually represent the relationship between network slices, nodes, and paths, laying a data foundation for subsequent path optimization and computational resource allocation. This matrix representation clearly establishes the mapping relationship between paths and nodes, and between network slices and paths, thereby achieving more efficient and reasonable task offloading and load balancing.

[0081] In one possible implementation, the first matrix element A(i,y)=1 in the first matrix indicates that the i-th available path will pass through node y, and A(i,y)=0 indicates that the i-th available path will not pass through node y; the second matrix element B(x,i)=1 in the second matrix indicates that network slice x uses available path i, and B(x,i)=0 indicates that network slice x does not use available path i.

[0082] More specifically, a first matrix is ​​constructed to represent the relationship between available paths and nodes. Specifically, the element of the first matrix is ​​defined as A(i,y): if the i-th available path passes through node y, then A(i,y)=1; otherwise, A(i,y)=0. By constructing the first matrix A, the relationship between paths and nodes is clearly represented. This allows for a direct reflection of the distribution of network nodes traversed by each path during subsequent path selection and allocation, thus facilitating path optimization and balanced resource allocation.

[0083] Multiple second matrices are constructed, each representing the relationship between each network slice and the selected available path. Specifically, the element of the second matrix is ​​defined as B(x,i), where B(x,i)=1 indicates that network slice x has selected available path i, and B(x,i)=0 indicates that the network slice has not selected path i. Since each network slice can only select at most one path, this constraint ensures that each row of each second matrix B has at most one element of 1. By constructing multiple second matrices B, the selection relationship between each network slice and available path can be clearly defined, which plays a crucial role in subsequent task unloading and resource allocation based on network slices.

[0084] By constructing a first matrix and multiple second matrices, the complex network slicing and path selection problem can be transformed into a matrix representation, making the relationship between network slices, paths, and nodes clear at a glance. This facilitates further optimization and computation, and allows for direct path selection and network node load analysis during matrix operations, which helps to achieve efficient management of network resources through optimization algorithms.

[0085] Step 13: Optimize all second matrices with the goal of maximizing the number of nodes contained in the selected available paths, and obtain the optimized matrix population.

[0086] In step 13, optimizing the entire second matrix with the goal of maximizing the number of nodes included in the selected available paths aims to improve resource utilization efficiency in the 5G network, enabling more network nodes to participate in task offloading and computation, maximizing network resource usage, and avoiding node resource idleness. Optimizing the second matrix effectively improves the overall network load balancing performance, ultimately enhancing task processing efficiency and reliability.

[0087] In one possible implementation, step 13 specifically includes the following steps:

[0088] The genetic algorithm was used to optimize all second matrices, and after multiple iterations, the population of optimized matrices was obtained.

[0089] Specifically, step 13 uses a genetic algorithm to optimize the entire second matrix, ultimately generating an optimized matrix population. A genetic algorithm is an optimization algorithm based on natural selection and genetic mechanisms, suitable for solving complex combinatorial optimization problems. Its core idea is to simulate the biological evolution process, continuously optimizing individuals in the population through operations such as selection, crossover, and mutation to find the globally optimal solution. By simulating the biological evolution process, the genetic algorithm performs selection, crossover, and mutation to gradually optimize the second matrix, finding the optimal second matrix, thereby maximizing the number of nodes included in the selected path and enabling more network nodes to participate in task computation and offloading. This process provides an efficient path optimization method for task offloading in 5G networks, ensuring maximum resource utilization and achieving load balancing.

[0090] In one possible implementation, refer to Figure 2 , Figure 2 This is the second flowchart of the 5G network task offloading and traffic splitting method provided by the present invention. The genetic algorithm specifically includes steps 21-23:

[0091] Step 21: In each iteration, randomly select two second matrices from all the second matrices, and randomly select the same rows to swap them, generating new matrix individuals.

[0092] Step 22: Randomly select a second matrix and randomly select two different rows. Swap the second matrix elements with a value of 1 in the two rows to generate a new matrix individual.

[0093] Step 23: Using the number of nodes contained in the selected available path as the evaluation criterion, select matrix individuals with fitness values ​​higher than a preset threshold to enter the next iteration. The number of nodes contained in the selected available path is directly proportional to the fitness value.

[0094] Specifically, step 21 first performs a crossover operation. In each iteration, two second matrices are randomly selected from all the second matrices, for example, second matrix Bp and second matrix Bq, and the same rows are randomly selected and swapped to generate new matrix individuals. The crossover operation forms new solution schemes through gene exchange between matrices, increasing the diversity of the population. The new matrix individuals after crossover can contain path selection combinations from different solutions, thereby exploring a wider solution space and avoiding premature entrapment in local optima.

[0095] Next, step 22 performs a mutation operation to introduce greater population diversity. The mutation operation involves randomly selecting two rows from the second matrix and swapping the matrix elements with a value of 1 in those two rows, generating new matrix individuals. This mutation process is equivalent to redistributing paths between network slices, but mutation only occurs if the original path is available, to maintain the validity of the solution. Through mutation, the genetic algorithm can avoid getting stuck in a homogeneous state after convergence, effectively expanding the search range and creating possibilities for exploring the optimal solution.

[0096] Finally, step 23 evaluates fitness based on the number of nodes contained in the selected available paths. Fitness values ​​are directly proportional to the number of nodes; paths containing more nodes indicate that more network resources are activated to participate in task unloading, thus increasing overall resource utilization efficiency. In each iteration, individuals with fitness values ​​higher than a preset threshold are selected for the next round. This selection strategy ensures continuous optimization of high-quality solutions, eliminates individuals with low fitness, and thus gradually improves the population quality.

[0097] After multiple iterations, the optimal second matrix generated by the genetic algorithm, along with other second matrices with high fitness, together form the optimized matrix population.

[0098] Step 14: Based on the first matrix and the optimized matrix population, obtain the node load matrix.

[0099] In step 14, generating the node load matrix based on the first matrix A and the optimized matrix population is a crucial step in realizing 5G network task offloading and load balancing. This is done to accurately reflect the load status of each node during task offloading, thereby providing reliable data support for subsequent resource allocation and load balancing, and ensuring efficient utilization of network resources.

[0100] In one possible implementation, refer to Figure 3 , Figure 3 This is the third flowchart of the 5G network task offloading and diversion method provided by the present invention. Step 14 specifically includes steps 141-143:

[0101] Step 141: Generate multiple third matrices, each of which is obtained by multiplying the first matrix by each optimized second matrix.

[0102] Step 142: Sum each third matrix column by column to obtain multiple fourth matrices.

[0103] Step 143: Determine the node load matrix from all fourth matrices based on the value of each fourth matrix element in each fourth matrix.

[0104] In the specific implementation process, firstly, in step 141, multiple third matrices are generated by performing matrix multiplication operations between the first matrix A and each optimized second matrix B. The first matrix A represents the relationship between available paths and nodes, while the optimized second matrix B represents the relationship between network slices and selected paths. The result of the matrix multiplication, i.e., the third matrix, specifically shows the number of network slices carried by each node under a specific path selection scheme. The purpose of this step is to closely integrate path selection with node load, forming the node load distribution under each path scheme.

[0105] Next, in step 142, each generated third matrix is ​​summed column-wise to obtain multiple fourth matrices. Each element in the fourth matrix represents the number of network slices carried by the corresponding node. By summing the columns of the third matrices, the load of each node under a specific path selection scheme can be quantified. This process transforms the results of complex matrix operations into specific node load data, providing a clear numerical basis for subsequent load analysis and optimization.

[0106] Then, in step 143, the final node load matrix is ​​determined based on the values ​​of the elements in each fourth matrix.

[0107] In one possible implementation, step 143 specifically includes the following steps:

[0108] When the number of non-zero fourth matrix elements in each fourth matrix is ​​inconsistent, the fourth matrix with the largest number of non-zero fourth matrix elements will be used as the node load matrix.

[0109] When the number of non-zero elements in any two fourth matrices is the same, calculate the variance of each fourth matrix;

[0110] The fourth matrix with the smallest variance is used as the node load matrix.

[0111] Specifically, to determine the final node load matrix, the generated fourth matrices need to be filtered and optimized to ensure the efficiency and balance of task offloading and load balancing. First, by comparing the number of non-zero elements in each fourth matrix, the fourth matrix containing the most non-zero elements is selected as the node load matrix. The purpose of this step is to ensure that as many nodes as possible participate in task offloading, thereby maximizing network resource utilization and avoiding potential bottlenecks caused by resources being concentrated on a few nodes.

[0112] If multiple fourth matrices have the same number of non-zero elements, their load distribution balance is further evaluated by calculating the variance of each fourth matrix. A matrix with a smaller variance indicates a more balanced distribution of tasks across nodes, reducing the risk of some nodes being overloaded while others are idle.

[0113] Ultimately, the fourth matrix with the smallest variance was selected as the final node load matrix. This selection ensured the balance of load distribution during task unloading, improving the overall network stability and system performance.

[0114] By implementing the above steps, the generated node load matrix not only covers more nodes participating in task offloading, but also optimizes the utilization efficiency of network resources through balanced load distribution, significantly improving the task processing capability and service quality of 5G networks in complex environments.

[0115] Step 15: Sort all nodes based on the value of each node load matrix element in the node load matrix and generate an array.

[0116] In step 15, all nodes are sorted and an array is generated based on the value of each element in the node load matrix. This is to optimize the subsequent computing resource allocation process and ensure the rationality and efficiency of resource allocation. By sorting the node load, the load level of different nodes in the network can be clearly identified, thereby formulating more targeted resource allocation strategies and improving the overall network's load balancing capability and computing resource utilization efficiency.

[0117] In one possible implementation, each node load matrix element in the node load matrix corresponds to a node, and the value of each node load matrix element represents the number of network slices carried by each node.

[0118] In one possible implementation, refer to Figure 4 , Figure 4 This is the fourth flowchart of the 5G network task offloading and diversion method provided by the present invention. Step 15 specifically includes steps 151-153:

[0119] Step 151: When the values ​​of the load matrix elements of each node are inconsistent, based on the size of the values ​​of all node load matrix elements, arrange all node load matrix elements from left to right in ascending order to generate an array.

[0120] Step 152: When the values ​​of any two node load matrix elements are the same, calculate the energy efficiency of the node corresponding to each node load matrix element.

[0121] Step 153: Based on the total energy efficiency, arrange all node load matrix elements from left to right in descending order to generate an array.

[0122] In the specific implementation process, the first step is to analyze each element of the node load matrix. Each element in the node load matrix corresponds to a network node, and its value represents the number of network slices carried by that node. To effectively manage and allocate computing resources, it is essential to understand the load status of each node. If the load values ​​of all nodes are different, they are sorted according to the values ​​of the node load matrix elements from smallest to largest, generating an ordered array. This sorting process ensures that nodes with lower loads receive priority in computing resource allocation, thereby preventing high-load nodes from becoming performance bottlenecks due to insufficient resources.

[0123] In some cases, multiple nodes may have the same load value. In such situations, to further optimize resource allocation, it's necessary to introduce a node energy efficiency metric. Specifically, when any two nodes have identical load matrix element values, their energy efficiency is calculated. Nodes with higher energy efficiency should have higher priority in resource allocation because they can complete more computational tasks with lower energy consumption, improving the overall network's energy efficiency. Based on the energy efficiency calculation results, these nodes with the same load are arranged in descending order of energy efficiency, generating the final sorted array.

[0124] The sorting process described above generates an array that reflects not only the load on each node but also its energy efficiency. This sorting method is highly effective. First, it ensures that computing resources are preferentially allocated to nodes with lower loads and higher energy efficiency, thereby optimizing resource utilization and preventing performance degradation in some nodes due to overload. Second, the balanced resource allocation strategy improves the overall network's load balancing capability, reduces the risk of single-point overload, and enhances network stability and reliability.

[0125] Step 16: Traverse each array element in the array, allocate computing resources to the network slices that pass through each node, and generate a computing resource allocation scheme, where each array element corresponds to a node.

[0126] In step 16, traversing each element of the sorted array and allocating computing resources to the network slice passing through each node to generate a computing resource allocation scheme is a crucial step in realizing task offloading and load balancing in the 5G network. This approach ensures the rational allocation and dynamic adjustment of computing resources, thereby optimizing the overall network performance and resource utilization efficiency.

[0127] In one possible implementation, refer to Figure 5 , Figure 5 This is the fifth flowchart of the 5G network task offloading and diversion method provided by the present invention. Step 16 specifically includes steps 161-163:

[0128] Step 161: When the value of an array element is zero, skip the node corresponding to that array element.

[0129] Step 162: When the value of an array element is 1, allocate the minimum value between the remaining computing power of the node corresponding to the array element and the remaining computing requirements of the corresponding network slice to the network slice, and update the remaining resources of the node and the network slice.

[0130] Step 163: When the value of an array element is greater than 1, the remaining computing power of the node corresponding to the array element is evenly distributed among the network slices it carries, and the remaining resources of the node and the network slice are updated.

[0131] In the specific implementation process, firstly, in step 161, when array element D(j) = 0, it indicates that the node currently does not carry any network slices. Therefore, the system skips this node and does not perform any resource allocation operations. The purpose of this step is to avoid wasting computing resources on unnecessary nodes and to concentrate resources on allocating them to nodes with actual needs, thereby improving resource utilization efficiency.

[0132] Next, in step 162, when array element D(j) = 1, it indicates that the node only supports one network slice. At this point, the system calculates t = min{node's remaining computing power, network slice's remaining computing requirements}, which is the smaller of the node's available computing resources and the network slice's actual required computing resources. This value t is then allocated to the corresponding network slice, while deducting the node's remaining computing power and the network slice's remaining computing requirements. If all computing requirements of a network slice are met, the slice is considered fully allocated, removed from the allocation list, and the candidate matrix D is refreshed to ensure more efficient subsequent resource allocation. This process ensures that a single slice receives sufficient computing resources to meet its needs while avoiding over-allocation of resources.

[0133] In step 163, when array element D(j) > 1, it indicates that the node simultaneously hosts multiple network slices. To achieve fair and efficient resource allocation, the system distributes the node's remaining computing power evenly among the network slices it hosts. Specifically, t is defined as the remaining computing demand of a network slice divided by D(j), meaning the node's remaining computing power is evenly distributed to each slice. Then, the calculated t value is assigned to each corresponding network slice, and the remaining resources of the node and the network slice are deducted accordingly. This even distribution method ensures the fairness of resource allocation among multiple slices, avoids performance bottlenecks caused by uneven resource allocation in some slices, and improves the overall system's load balancing capability.

[0134] By implementing steps 161 to 163, the system can dynamically and rationally allocate computing resources, ensuring that each network slice receives the corresponding computing power according to its actual needs. This not only improves the utilization efficiency of computing resources, avoiding resource waste and node overload, but also enhances the network's load balancing capability and overall performance through a fair resource allocation strategy. Furthermore, dynamically adjusting resource allocation can quickly respond to changes in network load, enhancing the system's adaptability and stability.

[0135] Step 17: Distribute the computing resource allocation plan to the physical network via an automated interface.

[0136] In step 17, the computing resource allocation scheme is distributed to the physical network via an automated interface. This is the final step in implementing the 5G network task offloading and load balancing method, ensuring that the optimized resource allocation strategy can be practically applied in the network environment.

[0137] Specifically, the first step is to design and deploy an efficient automation interface that seamlessly connects to the network management system and physical network devices. In practice, this interface communicates with physical network devices via standardized protocols (such as RESTful APIs and NETCONF) to ensure the accurate transmission and execution of resource allocation plans. The interface not only needs efficient data transmission capabilities but also error detection and correction mechanisms to ensure the allocation plan is not tampered with or lost during transmission. Furthermore, the interface should have real-time monitoring and feedback functions, enabling it to promptly obtain the execution status and feedback information from network devices after the resource allocation plan is issued, ensuring the effective implementation of the plan.

[0138] Through the implementation of this automated interface, computing resource allocation schemes can be rapidly and accurately applied to the physical network, significantly reducing the need for manual intervention and improving the efficiency and accuracy of resource allocation. This not only ensures the timeliness and reliability of task offloading and load balancing processes but also enhances the response speed and service quality of the entire 5G network. The automated deployment mechanism also supports dynamic adjustment and real-time optimization. When the network environment changes, the interface can quickly update and redeploy resource allocation schemes, maintaining optimal utilization of network resources and further enhancing network adaptability and resilience.

[0139] Reference Figure 6 , Figure 6 This is a schematic diagram of the structure of the 5G network task offloading and diversion system provided by the present invention, including:

[0140] The first processing module is used to determine all available paths for each network slice; each available path contains at least one node.

[0141] The second processing module is used to construct a first matrix and multiple second matrices. The first matrix is ​​used to represent the relationship between available paths and nodes, and each second matrix is ​​used to represent the relationship between each network slice and the selected available path. Each network slice can only select at most one available path.

[0142] The second processing module is also used to optimize the entire second matrix with the goal of maximizing the number of nodes contained in the selected available paths, so as to obtain the optimized matrix population.

[0143] The second processing module is also used to obtain the node load matrix based on the first matrix and the optimized matrix population;

[0144] The third processing module is used to sort all nodes based on the value of each node load matrix element in the node load matrix and generate an array;

[0145] The computing resource generation module is used to traverse each array element in the array, allocate computing resources to the network slice that passes through each node, and generate a computing resource allocation scheme, where each array element corresponds to a node;

[0146] The resource allocation module is used to distribute computing resource allocation schemes to the physical network through an automated interface.

[0147] In one possible implementation, the second processing module is further configured to optimize all second matrices using a genetic algorithm and, after multiple iterations, obtain an optimized matrix population.

[0148] In one possible implementation, the second processing module is further configured to:

[0149] In each iteration, two second matrices are randomly selected from all the second matrices, and the same rows are randomly selected and swapped to generate new matrix individuals;

[0150] Randomly select a second matrix and randomly select two different rows. Swap the second matrix elements with a value of 1 in the two rows to generate a new matrix individual;

[0151] Using the number of nodes contained in the selected available path as the evaluation criterion, matrix individuals with fitness values ​​higher than a preset threshold are selected to enter the next iteration. The number of nodes contained in the selected available path is directly proportional to the fitness value.

[0152] In one possible implementation, the second processing module is further configured to:

[0153] Multiple third matrices are generated, each of which is obtained by multiplying the first matrix by each optimized second matrix;

[0154] Summing each third matrix column by column yields multiple fourth matrices;

[0155] The node load matrix is ​​determined from all fourth matrices based on the value of each fourth matrix element in each fourth matrix.

[0156] In one possible implementation, the second processing module is further configured to:

[0157] When the number of non-zero fourth matrix elements in each fourth matrix is ​​inconsistent, the fourth matrix with the largest number of non-zero fourth matrix elements will be used as the node load matrix.

[0158] When the number of non-zero elements in any two fourth matrices is the same, calculate the variance of each fourth matrix;

[0159] The fourth matrix with the smallest variance is used as the node load matrix.

[0160] In one possible implementation, the third processing module is further configured to:

[0161] When the values ​​of the load matrix elements of each node are inconsistent, the load matrix elements of all nodes are arranged from left to right in ascending order based on the size of the values ​​of all node load matrix elements, and an array is generated.

[0162] When the values ​​of any two nodes' load matrix elements are the same, calculate the energy efficiency of the node corresponding to each node's load matrix element;

[0163] Based on the overall energy efficiency, all node load matrix elements are arranged from left to right in descending order to generate an array.

[0164] In one possible implementation, the computing resource generation module is further configured to:

[0165] When the value of an array element is zero, skip the node corresponding to that element.

[0166] When the value of an array element is 1, the minimum value between the remaining computing power of the node corresponding to the array element and the remaining computing requirements of the corresponding network slice is allocated to the network slice, and the remaining resources of the node and the network slice are updated.

[0167] When the value of an array element is greater than 1, the remaining computing power of the node corresponding to the array element is evenly distributed among the network slices it carries, and the remaining resources of the node and the network slice are updated.

[0168] In one possible implementation, the first processing module is further configured to:

[0169] Calculate all possible paths for each network slice;

[0170] Calculate the reliability value and total delay of all possible paths; the reliability value is calculated as the product of the reliability of all traversed nodes and the reliability of the communication link, and the total delay is calculated as the sum of the calculated delay of all traversed nodes and the link delay of the communication link.

[0171] Eliminate possible paths where the reliability value and / or total latency do not meet the preset requirements to obtain all available paths.

[0172] It should be noted that the 5G network task offloading and diversion system provided by the present invention can execute the 5G network task offloading and diversion method of any of the above embodiments during specific operation, which will not be described in detail in this embodiment.

[0173] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a 5G network task offloading and traffic splitting method. This method includes: determining all available paths for each network slice; each available path containing at least one node; constructing a first matrix and multiple second matrices, where the first matrix represents the relationship between available paths and nodes, and each second matrix represents the relationship between each network slice and a selected available path, with each network slice selecting at most one available path; optimizing all second matrices with the objective of maximizing the number of nodes contained in the selected available paths, resulting in an optimized matrix population; obtaining a node load matrix based on the first matrix and the optimized matrix population; sorting all nodes based on the value of each node load matrix element in the node load matrix, generating an array; traversing each array element, allocating computing resources to the network slice passing through each node, generating a computing resource allocation scheme, where each array element corresponds to a node; and distributing the computing resource allocation scheme to the physical network via an automated interface.

[0174] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0175] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to execute the 5G network task offloading and traffic splitting method provided in the above embodiments. The method includes: determining all available paths for each network slice; each available path contains at least one node; constructing a first matrix and a plurality of second matrices, wherein the first matrix is ​​used to represent the relationship between available paths and nodes, each second matrix is ​​used to represent the relationship between each network slice and a selected available path, and each network slice can only select at most one available path; optimizing all second matrices with the objective of maximizing the number of nodes contained in the selected available paths, to obtain an optimized matrix population; obtaining a node load matrix based on the first matrix and the optimized matrix population; sorting all nodes based on the value of each node load matrix element in the node load matrix, to generate an array; traversing each array element in the array, allocating computing resources to the network slice passing through each node, to generate a computing resource allocation scheme, wherein each array element corresponds to a node; and distributing the computing resource allocation scheme to the physical network through an automated interface.

[0176] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by processor 710, is implemented to perform the 5G network task offloading and traffic splitting method provided in the above embodiments. The method includes: determining all available paths for each network slice; each available path containing at least one node; constructing a first matrix and a plurality of second matrices, wherein the first matrix is ​​used to represent the relationship between available paths and nodes, each second matrix is ​​used to represent the relationship between each network slice and a selected available path, and each network slice can only select at most one available path; optimizing all second matrices with the objective of maximizing the number of nodes contained in the selected available paths, to obtain an optimized matrix population; obtaining a node load matrix based on the first matrix and the optimized matrix population; sorting all nodes based on the value of each node load matrix element in the node load matrix, to generate an array; traversing each array element in the array, allocating computing resources to the network slice passing through each node, to generate a computing resource allocation scheme, wherein each array element corresponds to a node; and distributing the computing resource allocation scheme to the physical network through an automated interface.

[0177] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0178] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for task offloading and traffic diversion in a 5G network, characterized in that, include: Determine all available paths for each network slice; Each of the available paths contains at least one node; Construct a first matrix and multiple second matrices. The first matrix is ​​used to represent the relationship between available paths and nodes. Each second matrix is ​​used to represent the relationship between each network slice and a selected available path. Each network slice can only select at most one available path. With the goal of maximizing the number of nodes contained in the selected available paths, optimize all of the second matrix to obtain the optimized matrix population; Based on the first matrix and the optimized matrix population, the node load matrix is ​​obtained; The optimized matrix population includes multiple optimized second matrices; The process of obtaining the node load matrix based on the first matrix and the optimized matrix population specifically includes: Multiple third matrices are generated, each of which is obtained by multiplying the first matrix by each of the optimized second matrices; Summing each of the third matrices column-wise yields multiple fourth matrices; The node load matrix is ​​determined from all the fourth matrices based on the value of each element in each of the fourth matrices; Based on the value of each node load matrix element in the node load matrix, all nodes are sorted to generate an array; Traversing each element of the array, allocating computing resources to the network slice passing through each node, and generating a computing resource allocation scheme, wherein each element of the array corresponds to a node; the process of traversing each element of the array, allocating computing resources to the network slice passing through each node, and generating a computing resource allocation scheme specifically includes: When the value of an array element is zero, skip the node corresponding to that element. When the value of an array element is 1, the minimum value between the remaining computing power of the node corresponding to the array element and the remaining computing requirements of the corresponding network slice is allocated to the network slice, and the remaining resources of the node and the network slice are updated. When the value of an array element is greater than 1, the remaining computing power of the node corresponding to the array element is evenly distributed among the network slices it carries, and the remaining resources of the node and the network slice are updated. The computing resource allocation scheme is distributed to the physical network via an automated interface.

2. The 5G network task offloading and diversion method according to claim 1, characterized in that, In the first matrix, the first matrix element A(i,y) = 1 indicates that the i-th available path will pass through node y, and A(i,y) = 0 indicates that the i-th available path will not pass through node y; in the second matrix, the second matrix element B(x,i) = 1 indicates that network slice x uses available path i, and B(x,i) = 0 indicates that network slice x does not use available path i.

3. The 5G network task offloading and diversion method according to claim 1, characterized in that, The process of optimizing the entire second matrix to maximize the number of nodes contained in the selected available paths, resulting in an optimized matrix population, specifically includes: The entire second matrix is ​​optimized using a genetic algorithm, and after multiple iterations, the optimized matrix population is obtained.

4. The 5G network task offloading and diversion method according to claim 3, characterized in that, The genetic algorithm specifically includes: In each iteration, two second matrices are randomly selected from all the second matrices, and the same rows are randomly selected and swapped to generate new matrix individuals; Randomly select a second matrix and randomly select two different rows. Swap the second matrix elements with a value of 1 in the two rows to generate a new matrix individual; Using the number of nodes contained in the selected available path as the evaluation criterion, matrix individuals with fitness values ​​higher than a preset threshold are selected to enter the next iteration, wherein the number of nodes contained in the selected available path is directly proportional to the fitness value.

5. The 5G network task offloading and diversion method according to claim 1, characterized in that, The step of determining the node load matrix from all the fourth matrices based on the value of each fourth matrix element in each of the fourth matrices specifically includes: When the number of non-zero fourth matrix elements in each of the fourth matrices is inconsistent, the fourth matrix containing the most non-zero fourth matrix elements will be used as the node load matrix. When the number of non-zero elements in any two of the fourth matrices is the same, calculate the variance of each of the fourth matrices; The fourth matrix with the smallest variance is taken as the node load matrix.

6. The 5G network task offloading and diversion method according to claim 1, characterized in that, Each element in the node load matrix corresponds to a node, and the value of each element represents the number of network slices carried by each node.

7. The 5G network task offloading and diversion method according to claim 6, characterized in that, The step of sorting all nodes and generating an array based on the value of each node load matrix element in the node load matrix specifically includes: When the values ​​of each node load matrix element are inconsistent, the array is generated by arranging all the node load matrix elements from left to right in ascending order based on the magnitude of all the node load matrix elements' values; When the values ​​of any two nodes' load matrix elements are the same, calculate the energy efficiency of the node corresponding to each node's load matrix element; Based on the energy efficiency of all the nodes, the elements of the node load matrix are arranged from left to right in descending order to generate the array.

8. The 5G network task offloading and diversion method according to claim 1, characterized in that, Before determining all available paths for each network slice, the method further includes: Calculate all possible paths for each of the network slices; Calculate the reliability value and the sum of delays for all possible paths; wherein the reliability value is calculated as the product of the reliability of all traversed nodes and the reliability of the communication link, and the sum of delays is calculated as the sum of the calculated delay of all traversed nodes and the link delay of the communication link; Eliminate possible paths where the reliability value and / or the total delay does not meet the preset requirements to obtain all available paths.

9. A 5G network task offloading and diversion system, characterized in that, include: The first processing module is used to determine all available paths for each network slice; Each of the available paths contains at least one node; The second processing module is used to construct a first matrix and multiple second matrices. The first matrix is ​​used to represent the relationship between available paths and nodes. Each second matrix is ​​used to represent the relationship between each network slice and a selected available path. Each network slice can only select at most one available path. The second processing module is also used to optimize the entire second matrix with the goal of maximizing the number of nodes contained in the selected available path, so as to obtain an optimized matrix population. The second processing module is further configured to obtain a node load matrix based on the first matrix and the optimized matrix population; the optimized matrix population includes multiple optimized second matrices. The step of obtaining the node load matrix based on the first matrix and the optimized matrix population specifically includes: generating multiple third matrices, each of which is obtained by multiplying the first matrix by each of the optimized second matrices; summing each of the third matrices column-wise to obtain multiple fourth matrices; and determining the node load matrix from all the fourth matrices based on the value of each fourth matrix element in each of the fourth matrices. The third processing module is used to sort all nodes based on the value of each node load matrix element in the node load matrix and generate an array; A computing resource generation module is used to traverse each array element in the array, allocate computing resources to network slices passing through each node, and generate a computing resource allocation scheme, wherein each array element corresponds to a node; the step of traversing each array element in the array, allocating computing resources to network slices passing through each node, and generating a computing resource allocation scheme specifically includes: when the value of the array element is zero, skipping the node corresponding to the array element; when the value of the array element is 1, allocating the minimum value between the remaining computing power of the node corresponding to the array element and the remaining computing demand of the corresponding network slice to the network slice, and updating the remaining resources of the node and the network slice; when the value of the array element is greater than 1, distributing the remaining computing power of the node corresponding to the array element evenly among the network slices it carries, and updating the remaining resources of the node and the network slice. The resource distribution module is used to distribute the computing resource allocation scheme to the physical network through an automated interface.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the 5G network task offloading and diversion method as described in any one of claims 1-8.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the 5G network task offloading and diversion method as described in any one of claims 1-8.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the 5G network task offloading and diversion method as described in any one of claims 1-8.

Images