Task offloading method, device and medium based on timing graph and graph matching theory in 6G networks

Abstract

The present invention discloses a task offloading method, device and medium based on timing graph and graph matching theory in 6G network, and for the problem of offloading dependent tasks in a moving scenario, a timing graph model is used to represent a set of base stations that dynamically change in the user's moving process in a time dimension, and the one-to-many mapping relationship between offload nodes and subtasks to be solved in the task offloading problem is abstracted into a graph homomorphism problem, and the task offloading strategy is solved based on the A* algorithm in the graph matching theory. The task offloading method, device and medium based on timing graph and graph matching theory in 6G network provided by the present invention comprehensively considers the two limiting factors of dependency between subtasks and user mobility based on the timing graph model, models the task offloading problem into a graph homomorphism problem from the graph matching theory, and proposes a reasonable solution to the problem of offloading dependent tasks in a moving scenario based on the A* algorithm. This method significantly reduces the actual completion delay of the application.

Description

[Technical field]

[0001] The present invention relates to a task offloading method, device and medium based on timing graph and graph matching theory in 6G networks, which belongs to the wireless communication technical field. [Background technology]

[0002] With the smooth rollout of 5G networks, in-depth research on 6G is being conducted at home and abroad. The world has basically reached a consensus on five major scenarios: immersive communication / super wireless broadband, ultra-large-scale connection, extreme / critical communication / ultra-reliable communication, communication perception integration, and artificial intelligence and communication integration. From the perspective of business use cases, digital twins and immersive applications are becoming one of the core businesses of 6G.

[0003] Many applications in the 6G era, such as XR (Extended Reality) and AI, are computationally intensive and place increasing demands on terminals. However, there are limitations to the computing power and power of terminals, making it necessary to provide users with additional computing power on the network side through edge computing and other methods.

[0004] In edge computing, the network not only needs to provide real-time computing services, but also needs to guarantee the QoS (e.g., delay) of computing tasks. In this regard, existing Chinese domestic manufacturers have proposed that the functional architecture of future 6G networks will be user-centric. In a user-centric network architecture, the network dynamically provides computing power services to the network based on user behavior (e.g., service needs and user mobility).

[0005] Taking immersive XR applications as an example, how can the network side dynamically provide computing power services to users based on mobility characteristics and the computing resources required by the application? That is, how can the two computing offloading methods of local processing and edge offloading select appropriate offloading nodes for each subtask of a single application in the user's process to reduce the actual computation delay of the application? This is an issue that needs to be considered. Summary of the Invention [Problem to be solved by the invention]

[0006] In order to solve the shortcomings in the prior art, the present invention provides a task offloading method, device and medium based on timing graph and graph matching theory in 6G networks. [Means for solving the problem]

[0007] In order to solve the above technical problems, the present invention provides a task offloading method based on timing graph and graph matching theory in 6G networks, which includes the following steps: JPEG0007636847000001.jpg226165JPEG0007636847000002.jpg190165

[0008] JPEG0007636847000003.jpg16161

[0009] JPEG0007636847000004.jpg24161

[0010] JPEG0007636847000005.jpg16161

[0011] JPEG0007636847000006.jpg52161

[0012] JPEG0007636847000007.jpg49161

[0013] JPEG0007636847000008.jpg38161

[0014] JPEG0007636847000009.jpg49161

[0015] JPEG0007636847000010.jpg63161

[0016] According to a second aspect of the present invention, there is provided a computer-readable recording medium having a computer program stored thereon, the computer program being capable of implementing a task offloading method based on timing graph and graph matching theory in a 6G network as claimed in any one of claims 1 to 7 when executed by a processor.

[0017] According to a third aspect of the present invention, there is provided a computing device comprising: a memory for storing instructions; and a processor for executing the instructions such that the computing device performs operations of a task offloading method based on timing graph and graph matching theory in 6G networks as claimed in any one of claims 1 to 7. Effect of the Invention

[0018] The task offloading method, device and medium based on timing graph and graph matching theory in 6G networks provided by the present invention use a timing graph model to represent the set of base stations that dynamically change during the user's movement process in the time dimension for the problem of offloading dependent tasks in mobile scenes, abstract the one-to-many mapping relationship between the offload nodes and subtasks to be solved in the task offloading problem into a graph homomorphism problem, and solve the task offloading strategy based on the A* algorithm in graph matching theory.

[0019] In order to further optimize the utility of applications, the timing information in the timing graph model is integrated into the design process of task offloading strategy solution algorithm. Therefore, to solve the problem of offloading dependent tasks in mobile scenarios, the present invention is based on the timing graph model, takes into account two factors, the dependency between subtasks and the mobility of users, and models the task offloading problem into a graph homomorphism problem from graph matching theory, and proposes a reasonable solution method for the problem of offloading dependent tasks in mobile scenarios based on the A* algorithm. This method greatly reduces the actual completion delay of applications. [Brief description of the drawings]

[0020] [Figure 1] 1 is a conceptual diagram of a task offload system according to the present invention. [Diagram 2] FIG. 2 is a diagram of a base station based on a timing graph model according to the present invention. [Diagram 3] FIG. 2 is an application diagram illustrating dependencies between subtasks according to the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] The present invention will be described in detail below with reference to specific examples.

[0022] The task offloading method based on timing graph and graph matching theory in a 6G network in a first embodiment includes the following steps. Step 1: Define the relevant parameters of the algorithm, specifically:

[0023] JPEG0007636847000011.jpg15161That is, there is a method in which a subtask performs local offloading in a user terminal, and a method in which a subtask is offloaded to an edge server connected to a base station to perform edge offloading. For ease of explanation, the edge server connected to the base station is referred to as a base station node, and the user node and the base station node are collectively referred to as an offload node.

[0024] JPEG0007636847000012.jpg54161

[0025] JPEG0007636847000013.jpg46161

[0026] The base station nodes that cover the 24161 users also change dynamically, so when a subtask is offloaded at a different time, the base station nodes that the subtask can select also change.

[0027] JPEG0007636847000015.jpg51161

[0028] Step 2: Define the optimization objectives of the OASA algorithm, specifically:

[0029] When the two offloading methods, local processing and edge offloading, coexist, we try to design a reasonable task offloading strategy for each subtask in the application to minimize the actual completion delay of the application. Considering that the computing power of the edge server is much higher than that of the user terminal, the application delay will be smaller when the same subtask selects the edge offloading method than when the local processing method is selected. Therefore, in order to minimize the completion delay of the application, we consider the subtask utility We define JPEG0007636847000016.jpg44 to weigh how much boost a task offloading strategy gives to a subtask in latency. The formula for defining subtask utility is as follows: JPEG0007636847000017.jpg1219

[0030] JPEG0007636847000018.jpg42161

[0031] JPEG0007636847000019.jpg5128

[0032] The overall optimization objective of the task offloading problem is to maximize the utility of the application, i.e., to select an appropriate task offloading strategy that minimizes the completion delay of the application, and the optimization objective is formulated as follows: JPEG0007636847000020.jpg612

[0033] JPEG0007636847000021.jpg14161

[0034] In the process of solving this problem, we face two main challenges:

[0035] The first challenge is user mobility. Since a user is in a moving state and its geographic location also changes, the base stations covering the user at different times are also different. That is, the edge servers to which the user can offload change dynamically, and the computation sources that different edge servers can provide are also different. When a user performs task offloading during the moving process, the adverse effect of mobility is that as the user switches base stations, the serving base station cannot timely feed back the computation results of the subtasks being processed before the user leaves the coverage area to the user, which may result in the interruption or even failure of task offloading. Considering such unfavorable factors, the accessibility time of the base station node is one of the factors to be considered when designing a task offloading strategy.

[0036] The second issue is the dependency between subtasks. An application is composed of multiple subtasks, and there is a dependency between the subtasks. That is, since a subtask needs to use the calculation result of the previous subtask as the input for the execution process in its own offload node, the subtask starts offloading after the previous subtask is fully processed. This means that the completion delay of a subtask not only includes the execution delay of the operation in its own offload node, but also includes the transfer delay of the calculation result of the previous subtask being sent to the corresponding offload node. Considering such unfavorable factors, in order to reduce the completion delay of a subtask, it is necessary to select an offload node with many calculation sources to reduce the execution delay, and to consider whether to offload two subtasks with a dependency to the same offload node to reduce the transfer delay according to the dependency between the subtasks.

[0037] To solve this problem, the task offloading problem is converted into a graph homomorphism problem, so that solving the maximum utility of an application, i.e., solving the minimum completion delay of an application, is converted into solving the shortest path. Since there is a dependency between subtasks, the completion time of the previous subtask is the start time of the subtask immediately after it. Therefore, between subtasks having a dependency, appropriate offload nodes are selected at different times to perform offloading, and the set of candidate nodes at different times changes dynamically due to the mobility of the user. In this regard, the matching pairs formed by selecting appropriate offload nodes from different sets of candidate offload nodes at different times of the subtask are nodes in the shortest path, so that the shortest path finally output by the algorithm is each subtask in the application and its actual offload node.

[0038] Step 3: Determine the input data and output results of the OASA algorithm, as follows: JPEG0007636847000022.jpg13128

[0039] JPEG0007636847000023.jpg63161

[0040] Step 4: Design the OASA (Optimized A_star Algorithm) algorithm.

[0041] JPEG0007636847000024.jpg22161

[0042] JPEG0007636847000025.jpg41161

[0043] JPEG0007636847000026.jpg28161

[0044] An application usually has one entry subtask, which needs to receive the user-specified application type (e.g., XR, AI) and other related parameters, and is called the entry subtask as the entry of the entire application. The entry subtask is usually processed locally in the user terminal.

[0045] JPEG0007636847000027.jpg44161

[0046] JPEG0007636847000028.jpg628As the exit of the entire application, the computation result is the computation result of the entire application. Typically, the exit subtask is arranged for local processing at the user terminal to ensure that the user receives the computation result of the application in a timely manner.

[0047] JPEG0007636847000029.jpg13161

[0048] JPEG0007636847000030.jpg14161

[0049] Step S5-2: Each subtask in JPEG0007636847000031.jpg415 is represented, and a subtask that performs delayed offloading, a subtask that performs serial processing, and a subtask that performs normal offloading are determined.

[0050] JPEG0007636847000032.jpg55161

[0051] A subtask that performs delayed offloading has the opportunity to access a base station node with better offloading conditions (the offloading conditions include the computational source provided by the base station and the length of time available after the user accesses the base station), which fully utilizes the favorable effect of user mobility. Since the base station that covers the user dynamically changes due to the user's mobility, if the base station that currently covers the user has unfavorable offloading conditions (e.g., the length of time the user can access is relatively short and switching occurs immediately after the user selects access), the user can select to access the base station at the next time to access a base station node with better offloading conditions, so that a subtask that may be originally processed locally can perform edge offloading.

[0052] A subtask that performs serial processing may select the same offload node as other subsequent subtasks of its previous subtask to reduce the number of transfers and thereby reduce the transfer delay of the subtask.

[0053] A subtask that performs normal offloading is a subtask that does not perform either the delayed offloading or serial processing, and this type of subtask selects an offload node with optimal offloading conditions from the set of base stations that currently cover the user.

[0054] JPEG0007636847000033.jpg92161

[0055] JPEG0007636847000034.jpg30161

[0056] JPEG0007636847000035.jpg129161

[0057] JPEG0007636847000036.jpg21161

[0058] JPEG0007636847000037.jpg42161

[0059] JPEG0007636847000038.jpg47161

[0060] JPEG0007636847000039.jpg46161

[0061] A second embodiment relates to a computer-readable recording medium, which stores a computer program, and when the computer program is executed by a processor, realizes the task offloading method based on timing graph and graph matching theory in 6G network according to any one of the first embodiment.

[0062] A computing device of a third embodiment includes a memory for storing instructions and a processor for executing the instructions such that the computing device performs operations of a task offloading method based on timing graph and graph matching theory in a 6G network described in any of the first embodiments.

[0063] The fourth embodiment is as follows. The A* algorithm represents the solution of the problem as a state to describe the progress at a certain point in time. Considering that solving an offloading strategy is essentially solving a one-to-many mapping relationship between subtasks and offloading nodes, i.e., a graph homomorphism problem, the state in the A* algorithm can represent the task offloading strategy at a certain point in time. In addition, since there are dependencies between subtasks, for example, for a subtask of serial processing, its offload node may be associated with the offload node of its previous subtask, so the offloading methods between subtasks have a certain correlation, and the A* algorithm represents the entire solution of the problem as a sequence of states, and this feature can be used to represent the association of the offloading methods between subtasks. Considering the suitability of the A* algorithm in the above two important aspects, the problem of offloading dependent tasks in a moving scene is solved based on the A* algorithm.

[0064] JPEG0007636847000040.jpg63161

[0065] JPEG0007636847000041.jpg113161Therefore, it can be seen from the scene that with the change of the user's geographical location, the base station covering the user changes dynamically, that is, the switching frequency of the user is relatively higher than that of a static user, and when the user leaves the coverage range of a certain base station, the base station becomes unavailable to the user, so the length of time that the user can use after accessing a certain base station has a certain limit. This means that the task offloading needs to be performed within the available time slot to ensure that the offloading can be performed normally.

[0066] JPEG0007636847000042.jpg55161

[0067] JPEG0007636847000043.jpg177161

[0068] JPEG0007636847000044.jpg104161.

[0069] In order to solve the task offloading strategy in the face of the challenges of subtask dependency and user mobility, the task offloading problem is abstracted into a graph homomorphism problem, and a one-to-many mapping relationship between offloading nodes and subtasks is established. This allows the mapping of nodes and edges in the application diagram to be considered to realize simultaneous optimization of the execution computation time and data transfer time of subtasks. For solving the graph homomorphism problem, the graph edit distance (GED) is taken as the graph similarity index, a set of edit operations is determined, and a cost corresponding to each edit operation is defined, and then a similarity index is defined. Finally, the task offloading strategy, which is a graph edit sequence, is solved by the provided OASA algorithm, which is an improved A* algorithm.

[0070] JPEG0007636847000045.jpg69161

[0071] JPEG0007636847000046.jpg77161

[0072] The OASA algorithm for solving the graph homomorphism problem, i.e., the task offloading problem, is as follows. JPEG0007636847000047.jpg208167

[0073] It should be noted that the above is only a preferred embodiment of the present invention, and many improvements and modifications may be made by those skilled in the art without departing from the principle of the present invention, and these improvements and modifications should also be regarded as within the protection scope of the present invention.

Claims

1.

2.

3.

4.

5.

6.

7.

8. A computer-readable recording medium storing a computer program, A recording medium, which, when the computer program is executed by a processor, realizes a task offloading method based on timing graph and graph matching theory in a 6G network according to any one of claims 1 to 7.

9. 1. A computing device comprising: A memory for storing instructions; A computer device, comprising: a processor for executing the instructions to perform operations of the task offloading method based on timing graph and graph matching theory in 6G networks according to any one of claims 1 to 7.

Images