Method and apparatus for mobile edge computing task offloading

Abstract

The application provides a mobile edge computing task offloading method and device, and relates to the technical field of computers. The method comprises the following steps: obtaining task information of a mobile edge computing task of a mobile edge computing system, wherein the task information comprises computing power requirement information required for completing the task; determining a target computing power device based on the computing power requirement information in a computing power device trust circle comprising at least one computing power device, wherein the comprehensive trust value of each computing power device comprised in the computing power device trust circle is greater than or equal to a first trust threshold, and the comprehensive trust value of the computing power device is determined based on the hardware parameter, the running state parameter, the historical behavior parameter and the multi-modal analysis parameter of the computing power device; and sending the mobile edge computing task to the target computing power device. The method and device of the application can improve the data security and reliability of task offloading by screening the target computing power device based on the comprehensive trust value of the computing power device, and can also improve the overall computing power collaboration efficiency.

Description

Technical Field

[0001] This application relates to the field of computer technology, and more specifically, to methods and apparatus for offloading mobile edge computing tasks. Background Technology

[0002] With the development of industrial internet and artificial intelligence technologies, edge devices generate a large number of complex computing tasks in scenarios such as smart factories that require collaboration among multiple mobile devices. Due to the limitations of traditional cloud-based centralized processing models, such as high response latency and heavy pressure on core network bandwidth, mobile edge computing (MEC) technology is usually adopted to offload computing tasks to edge computing devices for distributed processing.

[0003] Typically, when offloading tasks in mobile edge computing, the system needs to allocate tasks to computing devices in the edge network. However, computing devices in edge network environments often exhibit strong heterogeneity, and the entire system is in a highly dynamic state. Generally, current task offloading methods often allocate tasks based solely on simple computing power availability, lacking a comprehensive and multi-dimensional consideration of the overall reliability of alternative computing devices.

[0004] This makes it difficult to effectively ensure the data security of computing tasks during edge computing task offloading, and it can easily lead to task processing failures or data leaks. Overall, the task offloading efficiency and reliability of current mobile edge computing systems are relatively insufficient, and the utilization rate of computing power is low, making it difficult to meet the security and performance requirements of actual business operations. Summary of the Invention

[0005] This application provides a method and apparatus for offloading mobile edge computing tasks, which can improve the data security and reliability of task offloading, and also improve the overall computing power collaboration efficiency.

[0006] In a first aspect, a method for offloading mobile edge computing tasks is provided. The method includes: acquiring task information of a mobile edge computing task in a mobile edge computing system, the task information including computing power requirement information for completing the task; determining a target computing power device within a computing power device trust circle including at least one computing power device, based on the computing power requirement information, wherein the comprehensive trust value of each computing power device included in the computing power device trust circle is greater than or equal to a first trust threshold, and the comprehensive trust value of each computing power device is determined based on the hardware parameters, operating status parameters, historical behavior parameters, and multimodal parsing parameters of the computing power device; and sending the mobile edge computing task to the target computing power device.

[0007] In this embodiment, the comprehensive trust value of a computing power device is determined based on four dimensions: hardware parameters, operating status parameters, historical behavior parameters, and multimodal analysis parameters. This not only examines the physical capabilities of the computing power device but also its historical security reputation and data parsing accuracy. Furthermore, by setting a "first trust threshold" to filter target computing power devices, computing tasks are ensured to circulate only among qualified and trusted computing power devices. This fundamentally avoids the risk of data leakage caused by blindly uninstalling computing power devices. It not only protects data privacy and processing security but also effectively reduces the task processing failure rate by matching the actual operating status and improves the overall computing power collaboration efficiency of the system.

[0008] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: collecting the operating status data and multimodal analysis data of the computing power device through the computing power device in the mobile edge computing system; performing deep analysis on the collected operating status data and multimodal analysis data of the computing power device through a multimodal large model to obtain the operating status parameters and multimodal analysis parameters of the computing power device.

[0009] In this embodiment of the application, a multimodal large model is used to perform in-depth analysis of the collected operating status and multimodal data. This can accurately deal with extremely complex heterogeneous data in the edge environment, accurately obtain and quantify the operating status parameters and multimodal analysis parameters of the computing power device, and thus provide high-precision and in-depth data feature input for trust assessment, which can effectively improve the accuracy and reliability of subsequent assessment parameters.

[0010] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: determining the comprehensive trust value of computing devices in the mobile edge computing system; defining the set of computing devices whose comprehensive trust value is greater than or equal to a first trust threshold as the trust circle of the computing devices; wherein, the comprehensive trust value of the computing devices in the mobile edge computing system... Determined by the following equation: , in, The hardware computing power trust value is determined based on the hardware parameters of the computing device. This refers to the operational status trust value of computing power equipment, determined based on operational status parameters. The historical behavior trust value of computing devices, determined based on historical behavior parameters. The multimodal analytical trust value for computing power devices, determined based on multimodal analytical parameters. The weighting coefficients are non-negative, and .

[0011] In this embodiment, a technical approach is adopted that combines the trust values ​​of four dimensions—hardware computing power, operating status, historical behavior, and multimodal analysis—with non-negative weight coefficients to calculate a weighted sum, thereby deriving a final comprehensive trust value and delineating a trust circle set accordingly. This approach can overcome the problems of difficulty in uniformly and quantitatively evaluating multidimensional parameters and the lack of a reliable hierarchical organization mechanism when facing complex edge heterogeneous devices. It can transform abstract multidimensional features into precise mathematical quantitative indicators, thereby enabling dynamic hierarchical and structured management of edge computing resources and ensuring the data security and reliability of task offloading.

[0012] In conjunction with the first aspect, in certain implementations of the first aspect, determining the overall trust value of the computing power devices in the mobile edge computing system includes: determining that the computing power devices in the mobile edge computing system are experiencing an anomaly; and determining the overall trust value of the computing power devices according to the following equation. , , in, This is the overall trust value before the computing equipment malfunctions. This is the overall trust value after a computing device malfunctions. It is a historical trust value weighting system, and .

[0013] In this embodiment, when a computing device malfunctions, a weighted coefficient of historical trust value is introduced to dynamically update the comprehensive trust value by weighting and iterating the trust values ​​before and after the malfunction. This overcomes the technical problem that static trust assessment cannot reflect sudden situations (such as sudden power drops or network fluctuations) that occur in dynamic environments. It can then combine the historical performance and latest status of the computing device to achieve keen detection of abnormal situations and closed-loop management of the trust circle of computing devices, thereby enhancing the stability of the credit assessment of computing devices.

[0014] In conjunction with the first aspect, in some implementations of the first aspect, the abnormal situation of the computing power device includes at least one of the following: abnormal operation of the computing power device, abnormal network of the computing power device, abnormal task of the computing power device, and abnormal trust level of the computing power device.

[0015] In this embodiment, by specifying the abnormal situations that trigger trust level updates into four dimensions of anomalies—device operation, network, task, and trust level—a comprehensive and multi-dimensional anomaly early warning system can be established. This effectively prevents the omission of risks caused by single-dimensional monitoring and maximizes the fault tolerance and stability of the system during long-term operation.

[0016] In conjunction with the first aspect, in some implementations of the first aspect, the task information further includes business requirement information, which includes at least one of the following: business priority information, business sensitivity information, business processing accuracy information, business data security requirement information, and business response latency requirement information; the method further includes: determining the task type to which the mobile edge computing task belongs based on the computing power requirement information and business requirement information of the mobile edge computing task, and the minimum comprehensive trust value of the computing power device required for each task type; wherein, within the computing power device trust circle including at least one computing power device, determining the target computing power device based on the computing power requirement information includes: determining the target computing power device within the target computing power device trust circle matching the task type to which the mobile edge computing task belongs, wherein the comprehensive trust value of each computing power device included in the target computing power device trust circle is greater than or equal to a second trust threshold, and the second trust threshold is the minimum comprehensive trust value of the computing power device required for the task type to which the mobile edge computing task belongs.

[0017] In this embodiment, task types are determined based on business requirements including priority, sensitivity, and latency, combined with computing power requirements. A minimum comprehensive trust value threshold is set for different task types, and the trust circle of the target computing power device is matched accordingly. This can overcome the resource mismatch problem that is easily caused by different business requirements for security, latency, sensitivity, or accuracy. In turn, it can ensure extremely low response latency and extremely high security for high-priority sensitive tasks, while preventing low-priority tasks from crowding out high-quality resources, thereby enabling fine-grained hierarchical allocation of computing power.

[0018] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: dividing the mobile edge computing task into at least two sub-tasks based on the computing power requirement information and business requirement information of the mobile edge computing task, wherein the at least two sub-tasks include a first sub-task and a second sub-task; determining the task types to which the first sub-task and the second sub-task belong; wherein determining the target computing power device within the target computing power device trust circle that matches the task type to which the mobile edge computing task belongs includes: determining a first target computing power device for completing the first sub-task within a first target computing power device trust circle that matches the task type to which the first sub-task belongs; and determining a second target computing power device for completing the second sub-task within a second target computing power device trust circle that matches the task type to which the second sub-task belongs.

[0019] In this embodiment of the application, by decomposing the mobile edge computing task into at least two sub-tasks based on computing power and business needs, and determining the target computing power device within the trust circle of one or more target computing power devices that match the type of each sub-task, the bottleneck problem of the upper limit of computing power of a single trusted computing power device can be overcome. This can improve the efficiency of parallel collaborative processing of multiple computing power devices in the system, thereby greatly improving the overall throughput of the task and significantly shortening the task response time.

[0020] In conjunction with the first aspect, in some implementations of the first aspect, the task type includes a first task type, a second task type, and a third task type, wherein the minimum comprehensive trust value of the computing power equipment required for a task of the first task type is greater than the minimum comprehensive trust value of the computing power equipment required for a task of the second task type, and the minimum comprehensive trust value of the computing power equipment required for a task of the second task type is greater than the minimum comprehensive trust value of the computing power equipment required for a task of the third task type; wherein splitting the mobile edge computing task into at least two sub-tasks includes: when the task type to which the mobile edge computing task belongs is the first task type or the second task type, splitting the mobile edge computing task into at least two sub-tasks, wherein the task types of the first sub-task and the second sub-task are different.

[0021] In this embodiment, by setting a tiered minimum comprehensive trust value threshold and forcibly breaking down higher-level computing tasks into sub-tasks with different types of requirements, the security and reliability of higher-level task processing can be better guaranteed. On the other hand, it can overcome the problem of excessive waste of high-level computing resources caused by the inconsistency of security and computing power requirements of each stage of the task. This enables accurate matching based on the actual characteristics of the sub-tasks, thereby maximizing the utilization of idle computing power in the trust circle of computing power devices at all levels and maximizing the utilization of global computing power resources of the system.

[0022] In conjunction with the first aspect, in some implementations of the first aspect, the target computing power device is determined by: constructing an objective function based on the task response latency of the mobile edge computing task, the network bandwidth usage, and the device computing power load balancing of the mobile edge computing system; and optimizing the objective function based on the constraints of the hardware parameters, operating status parameters, and comprehensive trust value of the computing power device of the mobile edge computing system to determine the target computing power device.

[0023] In this embodiment, by constructing an objective function using three core indicators—comprehensive task response latency, network bandwidth usage, and device computing power load balancing—and optimizing the solution under constraints of hardware, state, and trust values, the technical problem of finding the optimal resource allocation solution quickly by relying solely on simple rules in scenarios with massive heterogeneous devices and multiple concurrent tasks can be overcome. This achieves the technical effect of improving the solution accuracy of multi-objective combination optimization and quickly locking the optimal cross-device collaborative unloading strategy for tasks globally.

[0024] In conjunction with the first aspect, in some implementations of the first aspect, the objective function includes: ;in, To reduce the task response latency of mobile edge computing tasks, Network bandwidth usage for mobile edge computing tasks. To improve the load balancing of computing power in mobile edge computing systems, The weighting coefficients are non-negative, and The constraints include: , , ;in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

[0025] In this embodiment of the application, by defining a specific linear objective function formula and clearly defining the constraints of remaining computing power, power consumption, and minimum trust value, a solid mathematical and logical boundary can be provided for the offloading method. This ensures that the algorithm will not over-allocate tasks during optimization, fully protecting the safety baseline of the selected computing power device in terms of physical power consumption, computing power, and trust value, thereby improving the data security and reliability of task offloading.

[0026] In conjunction with the first aspect, in some implementations of the first aspect, the objective function includes: ;in, To reduce the task response latency of mobile edge computing tasks, Network bandwidth usage for mobile edge computing tasks. To improve the load balancing of computing power in mobile edge computing systems, For computing power equipment Overall trust value The weighting coefficients are non-negative, and The constraints include: , , ;in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

[0027] In this embodiment of the application, by introducing "" into the objective function "This trust value penalty item for computing power devices can directly impose penalty costs on low-trust computing power devices at the mathematical calculation level, thereby effectively overcoming the defect of allocating tasks to idle but unreliable computing power devices in order to unilaterally pursue low latency, and further improving the data security and reliability of task offloading."

[0028] In conjunction with the first aspect, in some implementations of the first aspect, the device computing power load balancing of the mobile edge computing system Determined by the following equation: , in, The number of computing devices in a mobile edge computing system. For computing power equipment The computing power already used For computing power equipment Total computing power This represents the average computing power utilization rate of computing devices in a mobile edge computing system.

[0029] In this embodiment of the application, by statistically analyzing the deviation between the computing power utilization ratio of each computing power device and the overall average computing power utilization rate of the system, the computing power load balance of the computing power devices can be specifically calculated and quantified. This can overcome the problem that some computing power devices are easily overloaded while others are idle during dynamic operation of the system. In this way, it can accurately measure and regulate the overall computing power distribution of the system and maintain the overall health and balance of the system.

[0030] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes: determining the anomaly level of the abnormal situation of the computing power device; and adopting a processing strategy corresponding to the anomaly level.

[0031] In this embodiment of the application, by introducing anomaly level classification and differentiated processing strategies, task interruption caused by system overreaction can be avoided, thereby ensuring that anomalies of different natures can be resolved in the fastest and most reasonable way, thus guaranteeing the long-term stability and business continuity of mobile edge computing task offloading.

[0032] Secondly, an apparatus for offloading mobile edge computing tasks is provided. The apparatus includes: an acquisition module for acquiring task information of a mobile edge computing task in a mobile edge computing system, the task information including computing power requirement information for completing the task; a determination module for determining a target computing power device within a computing power device trust circle including at least one computing power device, based on the computing power requirement information, wherein the comprehensive trust value of each computing power device included in the computing power device trust circle is greater than or equal to a first trust threshold, and the comprehensive trust value of each computing power device is determined based on the hardware parameters, operating status parameters, historical behavior parameters, and multimodal parsing parameters of the computing power device; and a transceiver module for sending the mobile edge computing task to the target computing power device.

[0033] In conjunction with the second aspect, in some implementations of the second aspect, the device further includes: a data acquisition module, used to acquire the operating status data and multimodal analysis data of the computing power device through the computing power device in the mobile edge computing system; and a data analysis module, used to perform in-depth analysis on the acquired operating status data and multimodal analysis data of the computing power device through a multimodal large model to obtain the operating status parameters and multimodal analysis parameters of the computing power device.

[0034] In conjunction with the second aspect, in some implementations of the second aspect, the determining module is further configured to: determine the comprehensive trust value of computing devices in the mobile edge computing system; and define the set of computing devices whose comprehensive trust value is greater than or equal to a first trust threshold as the trust circle of the computing devices; wherein, the comprehensive trust value of the computing devices in the mobile edge computing system... Determined by the following equation: .

[0035] In conjunction with the second aspect, in some implementations of the second aspect, the determining module is also used to: determine if the computing power device in the mobile edge computing system is experiencing an anomaly; and determine the comprehensive trust value of the computing power device according to the following equation. , ,in, This is the overall trust value before the computing equipment malfunctions. This is the overall trust value after a computing device malfunctions. It is a historical trust value weighting system, and .

[0036] In conjunction with the second aspect, in some implementations of the second aspect, the determining module is further configured to: determine the task type to which the mobile edge computing task belongs based on the computing power requirement information and business requirement information of the mobile edge computing task, and the minimum comprehensive trust value of the computing power device required for each task type; determine the target computing power device within the target computing power device trust circle that matches the task type to which the mobile edge computing task belongs, wherein the comprehensive trust value of each computing power device included in the target computing power device trust circle is greater than or equal to a second trust threshold, and the second trust threshold is the minimum comprehensive trust value of the computing power device required for the task type to which the mobile edge computing task belongs.

[0037] In conjunction with the second aspect, in some implementations of the second aspect, the determining module is further configured to: divide the mobile edge computing task into at least two sub-tasks based on the computing power requirement information and business requirement information of the mobile edge computing task, wherein the at least two sub-tasks include a first sub-task and a second sub-task; determine the task types to which the first sub-task and the second sub-task belong; determine a first target computing power device for completing the first sub-task within a first target computing power device trust circle that matches the task type to which the first sub-task belongs; and determine a second target computing power device for completing the second sub-task within a second target computing power device trust circle that matches the task type to which the second sub-task belongs.

[0038] In conjunction with the second aspect, in some implementations of the second aspect, the determining module is also used to: construct an objective function based on the task response latency of the mobile edge computing task, the network bandwidth usage, and the device computing power load balancing of the mobile edge computing system; and optimize and solve the objective function based on the constraints of the hardware parameters, operating status parameters, and comprehensive trust values ​​of the computing power devices of the mobile edge computing system to determine the target computing power devices.

[0039] Regarding the relevant technical solutions and beneficial effects of the second aspect, please refer to the first aspect and any of its possible implementations mentioned above. For the sake of brevity, they will not be elaborated here.

[0040] Thirdly, an apparatus for offloading mobile edge computing tasks is provided, the apparatus comprising: a memory for storing a computer program; and a processor for executing the computer program stored in the memory, such that the apparatus performs the method in any possible implementation of the first aspect described above.

[0041] Fourthly, an electronic device is provided, comprising units or modules for implementing the method in any of the possible implementations of the first aspect described above.

[0042] In the embodiments of this application, the electronic device includes, for example, a smart interactive display screen, a smart screen terminal, a display screen, etc.

[0043] Fifthly, a computer-readable storage medium is provided that stores instructions which, when executed by a processor, cause the processor to implement the method in any of the possible implementations of the first aspect described above.

[0044] In a sixth aspect, a computer program product is provided, comprising computer program code that, when executed on a computer, causes the computer to implement the method in any of the possible implementations of the first aspect described above.

[0045] In a seventh aspect, a computer program is provided that, when the computer program is run, causes the method in any of the possible implementations of the first aspect to be implemented.

[0046] It should be understood that the relevant technical solutions and beneficial effects of the third to seventh aspects mentioned above can be referred to the first aspect and any of its possible implementation methods. For the sake of brevity, they will not be elaborated here. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application; Figure 2 This is a schematic flowchart illustrating a task unloading method provided in an embodiment of this application; Figure 3 This is a schematic flowchart illustrating another task unloading method provided in an embodiment of this application; Figure 4 This is a schematic flowchart illustrating another method for task unloading provided in an embodiment of this application; Figure 5 This is a schematic diagram of another network architecture provided in an embodiment of this application; Figure 6 This is a schematic flowchart illustrating another task unloading method provided in the embodiments of this application; Figure 7 This is a schematic flowchart illustrating a method for determining a target computing power device provided in an embodiment of this application; Figure 8 This is a schematic flowchart illustrating another method for determining a target computing power device provided in an embodiment of this application; Figure 9This is a schematic flowchart illustrating another method for determining a target computing power device provided in an embodiment of this application; Figure 10 This is a schematic diagram of a task unloading execution process provided in an embodiment of this application; Figure 11 This is a schematic block diagram of a task unloading device provided in an embodiment of this application; Figure 12 This is a schematic block diagram of another task unloading device provided in the embodiments of this application; Figure 13 This is a schematic block diagram of another task unloading device provided in the embodiments of this application. Detailed Implementation

[0048] The technical solutions in this application will now be described with reference to the accompanying drawings.

[0049] This application will present various aspects, embodiments, or features relating to a system comprising multiple devices, components, modules, etc. It should be understood and appreciated that individual systems may include additional devices, components, modules, etc., and / or may not include all the devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these approaches are also possible. Furthermore, in the embodiments of this application, the words "exemplary," "for example," etc., are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" in the embodiments of this application should not be construed as being better or more advantageous than other embodiments or design schemes. Specifically, the use of the term "exemplary" is intended to present the concept in a concrete manner. The business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0050] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0051] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0052] In the embodiments of this application, the same reference numerals are used to denote the same component or part. For the same part in the embodiments of this application, only one part or component may be labeled with reference numerals in the figures. It should be understood that the reference numerals also apply to other identical parts or components. In addition, the various parts in the figures are not drawn to scale, and the dimensions and sizes of the parts shown in the figures are only exemplary and should not be construed as limiting this application.

[0053] In the embodiments of this application, unless otherwise specified, the steps of this application can be performed sequentially or randomly, preferably sequentially. For example, the method includes steps (a) and (b), indicating that the method may include steps (a) and (b) performed sequentially, or it may include steps (b) and (a) performed sequentially. For example, the mention that the method may also include step (c) indicates that step (c) can be added to the method in any order. For example, the method may include steps (a), (b), and (c), or it may include steps (a), (c), and (b), or it may include steps (c), (a), and (b), etc.

[0054] With the deepening evolution of industrial internet technology and the accelerated construction of smart factories, edge application scenarios such as drone inspection, unmanned vehicle logistics, and real-time edge monitoring are becoming increasingly diverse, leading to a significant increase in the concurrency and computing power requirements for edge artificial intelligence (AI) tasks such as image recognition, audio analysis, and multimodal sensor data processing. However, traditional cloud-centric centralized computing solutions have significant limitations when dealing with edge scenarios in factories. First, task response latency is high. Because the edge data source is far from the cloud center, data needs to be transmitted over long distances through the core network, resulting in high processing latency for tasks such as real-time inspection and anomaly identification, making it difficult to meet business requirements. Second, network bandwidth pressure is high. The factory edge continuously generates a large amount of multimodal raw data such as images and audio. If a full cloud processing mode is adopted, it will consume a large amount of core network bandwidth, which can easily cause network congestion. Third, data privacy and security are weak. The operating parameters of core equipment in the factory and some sensitive inspection images are at high risk of data leakage during the process of uploading to the cloud for centralized storage.

[0055] To address the aforementioned cloud processing challenges, edge computing solutions that offload computing power are increasingly being adopted. However, current edge task offloading methods still suffer from the following shortcomings in complex factory environments: First, the coordination efficiency of heterogeneous devices is low. End-side devices such as drones, unmanned vehicles, and edge servers within a factory exhibit significant heterogeneity in physical form and computing power. Existing solutions lack effective coordination mechanisms, easily leading to some devices having idle computing power while others are overloaded, resulting in low overall computing power utilization. Second, task offloading stability is insufficient. The remaining battery power, network connectivity, and idle status of end-side devices are dynamically changing. Traditional static task offloading strategies cannot adapt to these fluctuations in real time, easily leading to task offloading failures or decreased processing efficiency when devices experience sudden anomalies.

[0056] More generally, current edge task offloading strategies often rely solely on the basic availability of computing power for task allocation, lacking a comprehensive assessment and screening mechanism for the overall reliability of heterogeneous computing devices. This results in a blind allocation method that fails to accurately match task requirements with truly reliable computing devices when offloading edge computing tasks: on the one hand, it easily assigns computing tasks to nodes with poor actual operating conditions, leading to task processing failures or response timeouts, resulting in low overall task execution reliability; on the other hand, it is difficult to effectively prevent critical or sensitive tasks from flowing to computing devices with potential vulnerabilities, making it difficult to effectively guarantee data security during the mobile edge computing task offloading process.

[0057] Therefore, this application provides a method and apparatus for offloading mobile edge computing tasks, which can improve the data security and reliability of task offloading, and also improve the overall computing power collaboration efficiency.

[0058] Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application. For example... Figure 1 As shown, this application embodiment provides a collaborative network for a mobile edge computing system. This collaborative network can be divided into three layers, namely a bottom perception layer, a middle parsing layer, and a top scheduling layer.

[0059] Devices in the perception layer can sense computing power devices, collect data from them, and execute specific computing tasks. For example, these devices are responsible for sensing the usage habits of computing power devices, collecting multimodal data in real time or periodically, and their own operational status data (such as remaining power, idle percentage, etc.). Taking a smart factory as an example, the devices in the perception layer can cover various heterogeneous edge computing power devices within the smart factory. That is, the devices in the perception layer can include edge devices, which can refer to various physical devices deployed at the edge of the smart factory network that have a certain computing power and are highly heterogeneous. For example, edge devices can include one or more of the following: edge server 11, smart glasses 12, smart terminal 13, edge PC 14, edge box 15 for providing off-grid computing power, drone 16, and unmanned vehicle 17. These edge devices can also include MEC nodes that provide off-grid computing power. Edge devices can be responsible for sensing device usage habits and collecting multimodal data at the perception layer, and at the same time, they can be the main components of the computing power device trust circle in the scheduling layer, responsible for actually undertaking and executing the split and offloaded computing power tasks. In the embodiments of this application, the edge devices can be collectively referred to as computing power devices.

[0060] The intermediate parsing layer, for example, is a multimodal parsing layer. It can employ a large multimodal model to perform deep analysis of one or more multimodal data sources collected by the perception layer, such as images, audio, sensor data, and usage habits of inference services. Through deep parsing and feature extraction, this parsing layer can output accurate analytical parameters such as device operating status and task type priority to the top-level scheduling layer, providing crucial data support for subsequent assessments of computing power device trust, task classification, and task splitting. Figure 1 As shown, the devices for the parsing layer include, for example, a large model server 21.

[0061] The top layer is the scheduling layer, which is the core of task offloading and computing power coordination. This scheduling layer can calculate the comprehensive trust value of computing power devices by combining core parameters such as the idle percentage, remaining power, historical behavior, and multimodal parsing accuracy. This allows it to call upon edge devices with suitable comprehensive trust values ​​to coordinate task offloading. This scheduling layer includes, for example, an AI server 31.

[0062] In this embodiment, the three-layer architecture network can also be referred to as a three-layer architecture system or a mobile edge computing system. Through real-time monitoring and dynamic updates of trust levels during execution (e.g., increasing trust value when a device successfully completes a task, and decreasing it when power consumption drops sharply), the three-layer architecture network or system can dynamically adapt to real-time state changes of computing devices, ensuring the efficiency and stability of cross-device collaboration.

[0063] The embodiments in this application are only for reference. Figure 1 The network architecture shown is illustrated as an example, but this application is not limited to this. For example, the scheduling layer can also integrate a multimodal large model, and the device of the scheduling layer is, for example, a mobile edge device, to parse multimodal data and perform task offloading and computing power coordination. Thus, the network architecture can also include a perception layer and a scheduling layer, and the mobile edge computing system can include a mobile edge computing device and a computing power device.

[0064] Figure 2 This is a schematic flowchart of a task offloading method 100 provided in an embodiment of this application. The method 100 can be executed, for example, by a mobile edge computing device, such as a server or an AI server integrating multimodal large models. Figure 2 As shown, the method 100 includes: S110, Obtain task information for the mobile edge computing task of the mobile edge computing system. This task information includes the computing power requirement information for completing the task.

[0065] For example, a mobile edge computing device receives MEC tasks from a mobile edge computing system. These MEC tasks may include image recognition, factory inspection, audio analysis, sensor data processing, etc. The mobile edge computing device can obtain task information for the MEC task based on the MEC task. This task information may include computing power requirements for completing the MEC task.

[0066] In this embodiment, the computing power requirement information may include basic computing power information required to complete the task. This basic computing power information may include, for example, the total number of instruction cycles of the Central Processing Unit (CPU), that is, the total number of CPU cycles or rotations required to complete the task. Based on the number of CPU cycles and the frequency of the computing device, the time required to execute the task can be directly estimated.

[0067] This basic computing power information may, for example, include computing density, which can be the number of CPU cycles consumed per unit of data (such as per byte). This basic computing power information may also, for example, include the number of floating-point operations.

[0068] Different types of tasks can use different types of parameters for computing power requirements. For example, for computing power-intensive tasks such as AI model inference, scientific computing, or image processing, computing power requirements can include the number of floating-point operations required to complete the task; for tasks such as image recognition or video encryption, computing power requirements can include the computational density required to complete the task.

[0069] In the embodiments of this application, the computing power requirement information may include one or more parameters for evaluating the basic computational load or required time of the task. For example, the computing power requirement information may include one or more parameters such as the number of CPU cycles, computational density, and number of floating-point operations.

[0070] In this embodiment, the computing power requirement information may also include computing power resource quotas, such as the minimum and peak memory required for task execution; or, for example, information such as cache or video memory requirements required for task execution. In this embodiment, the computing power requirement information may also include related constraint information such as latency and power consumption, such as the maximum tolerable latency for completing the task and the maximum energy consumption budget.

[0071] S120, within the computing power device trust circle including at least one computing power device, the target computing power device is determined based on the computing power demand information. The comprehensive trust value of each computing power device included in the computing power device trust circle is greater than or equal to a first trust threshold. The comprehensive trust value of each computing power device is determined based on the hardware parameters, operating status parameters, historical behavior parameters, and multimodal parsing parameters of the computing power device.

[0072] For example, a mobile edge computing device can determine the comprehensive trust value of one or more computing devices within a mobile edge computing system based on trust value parameters. These trust value parameters may include, for example, hardware parameters, operational status parameters, historical behavior parameters, and multimodal parsing parameters of the computing devices. Based on the comprehensive trust value of one or more computing devices, the mobile edge computing device can classify the computing devices. For example, it can divide one or more computing devices with a comprehensive trust value greater than or equal to a first trust threshold into a computing device trust circle, and identify target computing devices within this computing device trust circle.

[0073] In this embodiment, the hardware parameters of the computing power device may include parameters reflecting the hardware performance of the computing power device, such as those used to evaluate the basic physical capabilities of the computing power device. For example, the hardware parameters of the computing power device may include one or more of the following: computing power information, storage capacity, network bandwidth, etc. These parameters can determine whether the computing power device has the basic suitability to undertake a specific task at the physical level.

[0074] In this embodiment, the operating status parameters of the computing power device may include parameters reflecting the operating status of the computing power device, such as those used to assess the current real-time health and load status of the computing power device. For example, the operating status parameters of the computing power device may include one or more of the following: remaining power, idle percentage, and continuous running time. These dynamic parameters help the system avoid devices that are about to lose power, are overloaded, or are unstable, ensuring the reliability of task offloading.

[0075] In this embodiment, the historical behavior parameters of the computing power device may include parameters reflecting the historical behavior of the computing power device, such as those used to assess the past credit and performance of the computing power device. For example, the historical behavior parameters of the computing power device may include one or more of the following: task completion rate, data transmission security, and task processing latency. By tracing whether the computing power device has a history of data leakage or high latency, it is possible to ensure that critical and sensitive tasks are not routed to unreliable nodes.

[0076] In this embodiment, the multimodal parsing parameters of the computing power device may include parameters reflecting the device's ability to parse multimodal data. For example, these parameters can be used to evaluate the device's actual business capabilities in processing specific types of heterogeneous data. For instance, the multimodal parsing parameters may include one or more of the following: image or audio parsing accuracy, sensor data acquisition precision, and inference service response speed. These parameters can reflect the parsing accuracy and response performance of the computing power device in real-world AI edge tasks.

[0077] S130 sends a mobile edge computing task to the target computing device.

[0078] This application embodiment comprehensively determines the trust value of computing power devices based on four dimensions: hardware parameters, operating status parameters, historical behavior parameters, and multimodal parsing parameters. It examines not only the physical capabilities of the computing power devices but also their past security reputation and data parsing accuracy. By setting a first trust threshold to filter target computing power devices, it effectively avoids assigning tasks involving privacy or sensitive data to unreliable computing power devices with a history of security vulnerabilities or poor parsing capabilities. This fundamentally protects data privacy and processing security, thereby significantly improving the data security and reliability of task offloading. Furthermore, because the comprehensive trust value of computing power devices in this application embodiment considers operating status parameters, the system can dynamically avoid computing power devices that are insufficient in computing power, about to lose power, or overloaded. This not only achieves precise matching between task computing power requirements and the actual available state of computing power devices, reducing task failure rates caused by sudden changes in the state of computing power devices, but also fully utilizes idle computing power resources in the edge network, improving the overall computing power coordination efficiency of the system.

[0079] The following will combine Figures 3 to 6 This application describes in detail the methods for determining the comprehensive trust value of computing devices and the methods for determining the trust circle of computing devices in its embodiments.

[0080] Figure 3 This is a schematic flowchart illustrating another task unloading method provided in an embodiment of this application. Figure 3 As shown, the method 100 includes: S140 collects operational status data and multimodal analysis data from computing devices in the mobile edge computing system.

[0081] In this embodiment, the mobile edge computing device can acquire the operating status data and multimodal parsing data of the computing power device through the computing power device. For example, the computing power device in the mobile edge computing system can periodically collect the operating status data or information of the computing power device, as well as the multimodal parsing data or information, or it can periodically collect relevant data or information. Taking a smart factory as an example, computing power devices such as drones, unmanned vehicles, edge servers, and edge boxes in a smart factory can collect their own operating status data, business multimodal data, inference service usage habit data, etc., in real time or periodically, and can transmit them to the multimodal parsing layer through the network, providing comprehensive and necessary data for system evaluation and processing.

[0082] S150 uses a multimodal large model to perform in-depth analysis on the collected operating status data and multimodal analytical data of computing power devices to obtain the operating status parameters and multimodal analytical parameters of the computing power devices.

[0083] In this embodiment, the mobile edge computing device can perform in-depth analysis of the operating status data and multimodal parsing data collected by the computing power device through a multimodal large model, extracting data features, such as outputting parameters like image or audio parsing accuracy, sensor data acquisition precision, task type identification results, and device operating status evaluation. This enables the mobile edge computing to obtain the operating status parameters and multimodal parsing parameters of the computing power device, providing support for subsequent comprehensive trust calculations, task classification, and other tasks.

[0084] In the embodiments of this application, the mobile edge computing device can determine the hardware parameters of the computing device based on the hardware-related data fed back by the computing device; similarly, the mobile edge computing device can determine the operating status parameters of the computing device based on the task execution status fed back by the computing device, such as whether the task is completed, the task processing delay, and whether the data transmission is secure.

[0085] In this embodiment, by introducing a multimodal large model and leveraging its powerful feature extraction capabilities to deeply mine the raw data, the true health status of computing devices and their processing capabilities for various types of data can be accurately quantified. This solves the problem of insufficient parsing accuracy in current MEC task offloading schemes when facing complex heterogeneous multimodal data in the factory area. Furthermore, this embodiment transforms the originally ambiguous underlying physical information into reliable, deep-level feature inputs through the high-precision operating status parameters and multimodal parsing parameters output by the large model. This avoids trust calculation biases caused by inaccurate assessment of computing device status, thus providing high-precision data support for comprehensive trust assessment. Moreover, based on the precise parameters parsed by the large model, this embodiment can gain insight into the real-time status of each heterogeneous computing device in the network, ensuring that specific edge computing tasks are accurately scheduled to the most stable and best-suited computing device for the task's computing power requirements. This significantly improves task processing accuracy, completion efficiency, and the utilization rate of idle computing power in the factory area.

[0086] Once the mobile edge computing device obtains the hardware parameters, operational status parameters, historical behavior parameters, and multimodal parsing parameters of the computing power devices, it can determine the comprehensive trust value of these computing power devices and further determine the trust circle of the computing power devices, for example... Figure 4 As shown. Figure 4 This illustration shows a schematic flowchart of another task unloading method 100 provided in an embodiment of this application. The method 100 may further include: S160, determine the overall trust value of computing devices in the mobile edge computing system.

[0087] In this embodiment, the mobile edge computing device can determine the comprehensive trust value of the computing power device based on one or more parameters selected from hardware parameters, operating status parameters, historical behavior parameters, and multimodal analysis parameters. For example, the mobile edge computing device can determine the comprehensive trust value of the computing power device based on its hardware parameters and operating status parameters, or based on its hardware parameters and historical behavior parameters, or based on its hardware parameters, operating status parameters, and historical behavior parameters.

[0088] In this embodiment, the mobile edge computing device can determine the overall trust value of the computing power devices in the mobile edge computing system according to the following equation. : , in, The hardware computing power trust value is determined based on the hardware parameters of the computing device. This refers to the operational status trust value of computing power equipment, determined based on operational status parameters. The historical behavior trust value of computing devices, determined based on historical behavior parameters. The multimodal analytical trust value for computing power devices, determined based on multimodal analytical parameters. The weighting coefficients are non-negative, and .

[0089] Specifically, This represents the hardware computing power trust value, which can be determined based on one or more of the following: computing power information, storage capacity information, and network bandwidth information. For example, the hardware computing power trust value can be determined based on the corresponding functional relationship or correspondence table between the hardware computing power trust value and one or more of the following: computing power information, storage capacity information, and network bandwidth information.

[0090] This refers to the operational status trust value of a computing power device, determined based on operational status parameters. This operational status trust value can be determined based on one or more of the following: remaining battery power, idle percentage, and continuous operating time. For example, the operational status trust value can be determined based on a corresponding functional relationship or correspondence table between the operational status trust value and one or more of these pieces of information.

[0091] The historical behavior trust value represents the computing power device and is determined based on historical behavior parameters. This historical behavior trust value can be determined based on one or more of the following information: task completion rate information, data transmission security information, and task processing latency information. For example, the historical behavior trust value can be determined based on the correspondence function relationship or correspondence table between the historical behavior trust value and one or more of these information.

[0092] The multimodal resolution trust value, representing the computing power device, is determined based on multimodal resolution parameters. This multimodal resolution trust value can be determined based on one or more of the following: image / audio parsing accuracy information, sensor data acquisition accuracy information, and inference service response speed information. For example, the multimodal resolution trust value can be determined based on the correspondence function relationship or correspondence table between the multimodal resolution trust value and one or more of these pieces of information.

[0093] In this embodiment of the application, the comprehensive trust value of the computing power device Hardware computing power trust value Running status trust value Trust value based on historical behavior Multimodal analysis of trust values The value range can be set to 0 to 1, that is, greater than or equal to zero and less than or equal to 1. These are non-negative weighting coefficients; for example, all four weighting coefficients could be 0.25.

[0094] S170, the set of computing power devices whose comprehensive trust value is greater than or equal to the first trust threshold is determined as the computing power device trust circle.

[0095] In this embodiment, the mobile edge computing device can classify computing devices within the system based on their comprehensive trust value. For example, the set of computing devices with a comprehensive trust value greater than or equal to a first trust threshold is defined as a computing device trust circle, where the first trust threshold is, for example, 0.3. Computing devices with a comprehensive trust value less than the first trust threshold are not included in the computing device trust circle and are prohibited from participating in task offloading to avoid affecting task processing efficiency and data security.

[0096] In this embodiment, a technical approach is adopted that combines the trust values ​​of four dimensions—hardware computing power, operating status, historical behavior, and multimodal analysis—with non-negative weight coefficients to calculate a weighted sum, thereby deriving a final comprehensive trust value and delineating a trust circle set accordingly. This approach can overcome the problems of difficulty in uniformly and quantitatively evaluating multidimensional parameters and the lack of a reliable hierarchical organization mechanism when facing complex edge heterogeneous devices. It can transform abstract multidimensional features into precise mathematical quantitative indicators, thereby enabling dynamic hierarchical and structured management of edge computing resources and ensuring the data security and reliability of task offloading.

[0097] To adapt to the complex and ever-changing business scenarios in smart factories, weighting coefficients can be set for these trust values, and these weighting coefficients can be dynamically adjusted according to the specific needs of the factory.

[0098] For example, when a smart factory's business scenario prioritizes security, such as inspections of sensitive areas or analysis of core equipment parameters, these high-priority sensitive tasks have extremely high requirements for data security and leak prevention. Therefore, the weight of historical behavior (γ) can be increased. By assigning higher weight to historical security records, the system can ensure that it values ​​the equipment's past creditworthiness, thereby accurately offloading tasks to equipment that has never had security vulnerabilities and has a high task completion rate.

[0099] For example, when a smart factory's business scenario prioritizes real-time performance, such as real-time identification of equipment faults within the factory area, response latency is extremely critical (e.g., a response latency of ≤500ms is required). To ensure processing speed, α (hardware computing power weight) and β (running state weight) can be increased. This adjustment makes the system pay more attention to whether the computing power of the device is currently strong, whether the battery is sufficient, and whether it is in a high idle state when selecting computing power devices, thereby ensuring that computing tasks can be executed as quickly as possible.

[0100] For example, when business scenarios heavily rely on the accurate analysis of complex data, such as image object detection and audio anomaly recognition, or when analyzing massive amounts of multimodal data generated in a factory, the δ (multimodal analysis weight) can be increased. This prioritizes the computing power devices that perform best in image and audio parsing accuracy or sensor data acquisition precision, ensuring the final quality of business processing.

[0101] For example, in an edge computing task for real-time equipment fault identification in a factory area, a mobile edge computing device evaluates edge box No. 1 deployed in a certain area of ​​the factory. Based on deep analysis of the large model and data collected by the perception layer, four basic trust values ​​of the computing device can be determined: based on the box's strong CPU computing power and bandwidth, the hardware computing power trust value of the box can be determined. The value is 0.9; based on the fact that the box is currently in a high idle percentage (e.g., 80% idle) and is directly powered without considering remaining power or other operating status parameters, the operating status trust value of the box can be determined. The trust value is 0.9. Based on historical behavior parameters such as the box's 100% task completion rate and lack of data leakage records, the historical behavior trust value of the box can be determined. The value is 1.0; based on the box's high resolution accuracy (e.g., 95% accuracy) for fault images and other multimodal resolution parameters, the box's multimodal resolution confidence value can be determined. It is 0.8.

[0102] Considering that the current business scenario involves real-time identification of equipment faults, which places high demands on the computing power response speed and real-time status of computing devices, the mobile edge computing device can dynamically allocate weight coefficients that favor real-time performance, such as α=0.4, β=0.4, γ=0.1, δ=0.1. Therefore, the current comprehensive trust value T of edge box No. 1 can be determined to be 0.90.

[0103] In this embodiment of the application, through this mechanism of flexibly configuring weight coefficients, the smart factory can dynamically and adaptively build a trust circle of computing power devices that best matches the characteristics of the current task when facing diverse edge AI tasks, thereby achieving the optimal solution for resource utilization and task processing.

[0104] In this embodiment of the application, the mobile edge computing device can classify all or part of the computing devices in the system into trust levels based on the comprehensive trust value of the computing devices. For example, the trust circle of the computing devices can be divided into three trust levels, and each trust level can have different trust circle permissions.

[0105] For example, a mobile edge computing device can define a first-level trust circle as the set of computing power devices whose overall trust value is greater than or equal to the first-level trust threshold. This first-level trust circle can be a core trust circle, and the first-level trust threshold is, for example, 0.8, meaning that the overall trust value of each computing power device included in the first-level trust circle is greater than or equal to 0.8. For example, the computing power devices within this trust circle mainly include edge servers and core edge devices (such as main inspection drones in the factory area and edge boxes in the core area) that have sufficient computing power, stable operation, good historical behavior, and high multimodal resolution accuracy. Figure 5 As shown in the schematic diagram of another network architecture provided by the representative embodiment of this application, the first-level trust circle 51 may include a first edge server 11a, a first smart glasses 12a, a first smart terminal 13a, and an edge box 15. The computing devices in this first-level trust circle have the highest priority task collaboration permissions, can undertake the offloading of sensitive tasks and high-priority tasks, and can also act as the core node of the trust circle, responsible for coordinating the task allocation of other trust level devices, etc.

[0106] For example, a mobile edge computing device can define a second-level trust circle as a set of computing devices with a comprehensive trust value greater than or equal to the second-level trust threshold. This second-level trust circle can be a general trust circle, and the second-level trust threshold is, for example, 0.5, meaning that the comprehensive trust value of each computing device included in the second-level trust circle is greater than or equal to 0.5 and less than 0.8. For example, the computing devices within this trust circle mainly include edge devices with moderate computing power, relatively stable operating status, and no abnormal historical behavior (such as ordinary inspection drones, unmanned vehicles, and edge terminals). Figure 5 As shown, the second-level trust circle 52 may include a second smart glasses 12b, an edge personal computer 14, a first drone 16a, and a first unmanned vehicle 17a. The computing power devices in this second-level trust circle can offload regular tasks, accept scheduling from the computing power devices in the first-level trust circle, and assist in completing sub-tasks after task decomposition.

[0107] For example, a mobile edge computing device can define a third-level trust circle as a set of computing devices with a comprehensive trust value greater than or equal to the third-level trust threshold. This third-level trust circle can be a temporary trust circle, and the third-level trust threshold is, for example, 0.3, meaning that the comprehensive trust value of each computing device included in the third-level trust circle is greater than or equal to 0.3 and less than 0.5. For example, the computing devices within this trust circle mainly include edge devices with limited computing power, large fluctuations in operating status, but no history of adverse behavior (such as temporarily deployed edge terminals and mobile inspection devices). Figure 5 As shown, the third-level trust circle 53 may include a second edge server 11b, a second smart terminal 13b, a second drone 16b, and a second unmanned vehicle 17b. The computing power devices in this third-level trust circle can undertake simple, non-sensitive sub-tasks only when the computing power devices in the first or second-level trust circles are insufficient, and their operating status and task completion status need to be monitored in real time.

[0108] For computing power devices with a comprehensive trust value of less than 0.3, they will not be included in any level of computing power device trust circle and will be prohibited from participating in task offloading to avoid affecting task processing efficiency and data security. In this embodiment, the scope of the computing power device trust circle is not fixed and can be updated in real time according to the dynamic changes in the device trust value to ensure the effectiveness of the trust circle.

[0109] In this embodiment, the comprehensive trust value of the computing power device can be dynamically updated by combining its real-time operating status, task completion status, and multimodal analysis results, thereby further dynamically updating the trust circle of the computing power device. The comprehensive trust value update can be periodic, real-time, or a combination of both. For example, the update cycle can be set to 5 minutes to ensure that the comprehensive trust value accurately reflects the current state of the computing power device, thereby optimizing task offloading strategies and improving task processing efficiency. The following will combine... Figure 6 The schematic flowchart shown in this application embodiment illustrates another task unloading method, describing a handling scheme in the event of an anomaly in the computing power device.

[0110] like Figure 6 As shown, the method 100 may include: S180, It is determined that there is an abnormality in the computing equipment in the mobile edge computing system.

[0111] In this embodiment, the mobile edge computing device can establish a real-time anomaly detection mechanism. For example, it can promptly detect various anomalies within the system through collaborative monitoring between the perception layer and the scheduling layer. In this embodiment, the anomalies of the computing power device can include one or more of the following: abnormal operation of the computing power device, network anomaly of the computing power device, task anomaly of the computing power device, and trust level anomaly of the computing power device.

[0112] For example, operational anomalies in computing power equipment mainly involve sudden problems at the physical and state levels, such as a sudden drop in remaining battery power, computing power overload, or equipment failure. Network anomalies in computing power equipment mainly involve quality issues in communication links, such as data transmission interruptions, sudden bandwidth drops, or excessively high latency. Task anomalies in computing power equipment mainly involve obstacles at the execution level of specific business processes, such as task processing failures, low parsing accuracy, or subtask dependency conflicts. Trust anomalies in computing power equipment mainly involve issues at the equipment security and trust level, such as a sudden drop in trust value or malicious behavior by the equipment.

[0113] S161, In cases where it is determined that the computing power equipment is experiencing an abnormal condition, the comprehensive trust value of the computing power equipment shall be determined according to the following equation. , , in, This is the overall trust value before the computing equipment malfunctions. This is the overall trust value after a computing device malfunctions. It is a historical trust value weighting system, and .

[0114] In the embodiments of this application, This is the overall trust value after a computing device malfunctions. It can be determined by the following equation:

[0115] in, These are the hardware computing power trust value, operating status trust value, historical behavior trust value, and multimodal analysis trust value after the computing power equipment malfunctions. , , , The weighting coefficients are consistent with those used when constructing the trust circle.

[0116] In this application embodiment, the following adjustment rules can be adopted for trust level updates in different scenarios. Regarding task completion, if the computing device successfully completes the task, and the processing latency and parsing accuracy meet the requirements, It can improve performance by 5%-10%; however, if the task fails or the results are unsatisfactory, This can reduce costs by 10%-20%. For example, regarding operational status, if the remaining battery power of the computing device is below 20% and the idle percentage is below 30%, then... It can reduce costs by 15%-25%; if the computing power equipment returns to normal operation, then... It can gradually recover. For example, regarding data security, if a computing device experiences data leakage or transmission anomalies, the trust value can be directly lowered to below 0.3, removing it from the trust circle and issuing an anomaly warning. For another example, regarding multimodal parsing, if the accuracy of multimodal data parsing by the computing device improves, then… It can improve accuracy by 5%-10%; if accuracy decreases, then... It can reduce it by 10%-15%.

[0117] In this embodiment, when a computing device malfunctions, a weighted coefficient of historical trust value is introduced to dynamically update the comprehensive trust value by weighting and iterating the trust values ​​before and after the malfunction. This overcomes the technical problem that static trust assessment cannot reflect sudden situations (such as sudden power drops or network fluctuations) that occur in dynamic environments. It can then combine the historical performance and latest status of the computing device to achieve keen detection of abnormal situations and closed-loop management of the trust circle of computing devices, thereby enhancing the stability of the credit assessment of computing devices.

[0118] In this embodiment of the application, the method further includes: determining the anomaly level of the abnormal situation of the computing power device; and adopting a processing strategy corresponding to the anomaly level.

[0119] Mobile edge computing devices can classify anomalies into three levels based on their severity and impact, with different levels corresponding to different processing priorities. Specifically, anomalies affecting core task execution, causing task unloading interruptions, or data leaks can be classified as Level 1 or severe anomalies. Examples include core trust circle device failures, network transmission interruptions, and malicious data leaks by devices. These level anomalies have the highest processing priority and require immediate response. Anomalies affecting regular task execution and causing task delays can be classified as Level 2 or general anomalies. Examples include ordinary trust circle device overload, network bandwidth degradation, and subtask processing timeouts. These level anomalies have a medium processing priority and require a response within one minute. Anomalies that do not affect task execution and only require subsequent optimization can be classified as Level 3 or minor anomalies. Examples include slightly low battery levels in temporary trust circle devices and a slight decrease in parsing accuracy. These level anomalies have the lowest processing priority and can be processed within the trust level update cycle.

[0120] Correspondingly, mobile edge computing devices can adopt processing strategies corresponding to the anomaly level. For example, differentiated processing strategies can be formulated for different levels of anomalies to ensure that the anomalies are resolved quickly and effectively.

[0121] For example, in the case of a Level 1 anomaly, the execution of tasks by the abnormal device can be immediately stopped, it removed from the trust circle, an emergency dispatch mechanism can be triggered, and the sub-tasks undertaken by the abnormal device can be reassigned to other devices in the core trust circle. If the anomaly is a network anomaly, the backup edge network can be switched immediately to restore data transmission. If the anomaly is a data leakage anomaly, data transmission can be immediately terminated, the leaked data can be encrypted, device security vulnerabilities can be investigated, abnormal device information can be recorded, and it can be prohibited from rejoining the trust circle.

[0122] For example, for a level 2 anomaly, new task allocation to the anomalous device can be paused, and its load can be adjusted (e.g., reducing the number of subtasks). If the anomaly is due to device overload, some subtasks can be migrated to idle devices of the same trust level. If the anomaly is due to decreased network bandwidth, data transmission methods can be optimized (e.g., compressing multimodal data) to reduce bandwidth consumption. If the anomaly is due to subtask processing timeout, processing time can be extended, and the device's trust value can be adjusted, lowering its priority for subsequent task allocation.

[0123] For example, for a Level 3 anomaly, the anomaly information can be recorded, and the device trust value can be adjusted when the trust level is updated, reminding plant management personnel to perform equipment maintenance (such as charging or debugging). For instance, if the anomaly persists, the device trust level can be downgraded until the anomaly is resolved.

[0124] In this embodiment of the application, by introducing anomaly level classification and differentiated processing strategies, task interruption caused by system overreaction can be avoided, thereby ensuring that anomalies of different natures can be resolved in the fastest and most reasonable way, thus guaranteeing the long-term stability and business continuity of mobile edge computing task offloading.

[0125] In this embodiment, after resolving anomalies, a recovery mechanism for devices and tasks can be established. For example, for faulty devices, after repair and debugging, their trust value can be recalculated, and they can be re-included in the corresponding trust circle based on the trust value, gradually allocating tasks. For network anomalies, task execution progress can be synchronized after normal transmission is restored to ensure continuous task execution. For task anomalies, sub-tasks can be reallocated to complete incomplete task steps, ensuring the final completion of the task. Furthermore, the anomaly handling process can be recorded, the causes of anomalies analyzed, and trust circle management and task scheduling strategies optimized to reduce the occurrence of subsequent anomalies.

[0126] The following will combine Figures 7 to 9 This application describes in detail the method for determining a target computing power device provided in its embodiments. For example... Figure 7 As shown, the method 120 includes: S121. Based on the computing power requirement information and business requirement information of the mobile edge computing task, determine the task type to which the mobile edge computing task belongs, and the minimum comprehensive trust value of the computing power equipment required for each task type.

[0127] In this embodiment, the task information for a mobile edge computing task may further include business requirement information, which includes one or more of the following: business priority information, business sensitivity information, business processing accuracy information, business data security requirements information, and business response latency requirements information. Based on the business requirement information, the mobile edge computing device can determine the task type to which the mobile edge computing task belongs, and each task type corresponds to the minimum comprehensive trust value of the required computing power equipment.

[0128] In this embodiment of the application, the task type may include a first task type, a second task type, and a third task type, wherein the minimum comprehensive trust value of the computing power device required for the task of the first task type is greater than the minimum comprehensive trust value of the computing power device required for the task of the second task type, and the minimum comprehensive trust value of the computing power device required for the task of the second task type is greater than the minimum comprehensive trust value of the computing power device required for the task of the third task type.

[0129] For example, in conjunction with factory business scenarios, mobile edge computing devices can classify mobile edge computing tasks into three categories, such as the first task type, the second task type, and the third task type, and can formulate different offloading strategies based on task priority, sensitivity, computing power requirements, etc.

[0130] The first task type can be a high-priority sensitive task, such as real-time identification of equipment faults in the factory area, processing of inspection data in sensitive areas, and analysis of operating parameters of core equipment. These tasks have extremely high requirements for response latency (≤500ms), data security, and processing accuracy. They are only allowed to be undertaken by devices in the core trust circle and are preferentially offloaded to edge servers in the core trust circle to ensure the security and real-time performance of task processing.

[0131] The second type of task can be a medium-priority routine task, such as routine area inspection, material transportation route planning, and routine sensor data collection and analysis. These tasks have moderate requirements for response latency (≤1000ms) and computing power. They can be undertaken collaboratively by core trust circle and ordinary trust circle devices, and tasks can be allocated according to the device's idle status and computing power matching degree.

[0132] The third task type can be a low-priority, simple task, such as data backup, historical data statistics, and environmental monitoring of non-critical areas. These tasks have low response latency requirements (≤3000ms) and low computing power requirements. They can be undertaken by temporary trust circle devices or processed when core or ordinary trust circle devices are idle, making full use of idle computing power.

[0133] S122, within the trust circle of computing power devices, which includes at least one computing power device, the target computing power device is determined based on computing power demand information, including: The target computing devices are determined within the trust circle of the target computing devices that matches the task type to which the mobile edge computing task belongs. The overall trust value of each computing device included in the trust circle is greater than or equal to a second trust threshold, which is the minimum overall trust value of the computing devices required for the task type to which the mobile edge computing task belongs.

[0134] In this embodiment, for example, if the minimum overall trust value of the computing power device required for a task of the first task type is 0.8, then this type of task can be executed by computing power devices within the first level trust circle. That is, for a task of the first task type, the target computing power device can be determined within the first level trust circle. As another example, if the minimum overall trust value of the computing power device required for a task of the second task type is 0.5, then this type of task can be executed by computing power devices within the second level trust circle or the first level trust circle. That is, for a task of the second task type, the target computing power device can be determined within the second level trust circle and / or the first level trust circle. As yet another example, if the minimum overall trust value of the computing power device required for a task of the third task type is 0.3, then this type of task can be executed by computing power devices within the first, second, or third level trust circles. That is, for a task of the third task type, the target computing power device can be determined within the first, second, or third level trust circles.

[0135] In the embodiments of this application, the mobile edge computing device can determine the target computing power device within a trust circle level that matches the task type to which the mobile edge computing task belongs. For example, for a task of the first task type, the target computing power device can be determined within the first level trust circle; for a task of the second task type, the target computing power device can be determined within the second level trust circle; and for a task of the third task type, the target computing power device can be determined within the third level trust circle.

[0136] In this embodiment, task types are determined based on business requirements including priority, sensitivity, and latency, combined with computing power requirements. A minimum comprehensive trust value threshold is set for different task types, and the trust circle of the target computing power device is matched accordingly. This can overcome the resource mismatch problem that is easily caused by different business requirements for security, latency, sensitivity, or accuracy. In turn, it can ensure extremely low response latency and extremely high security for high-priority sensitive tasks, while preventing low-priority tasks from crowding out high-quality resources, thereby enabling fine-grained hierarchical allocation of computing power.

[0137] Figure 8 A schematic flowchart illustrating another method for determining a target computing power device provided in an embodiment of this application is shown. Figure 8 As shown, the method 120 includes: S123, based on the computing power requirement information and business requirement information of the mobile edge computing task, the mobile edge computing task is divided into at least two sub-tasks, which include a first sub-task and a second sub-task.

[0138] In this embodiment, the mobile edge computing device can divide a mobile edge computing task into two or more sub-tasks based on the computing power requirement information and business requirement information of the mobile edge computing task. For example, the mobile edge computing device can divide the mobile edge computing task into multiple sub-tasks, each with different computing power requirements; or each sub-task with different task priorities; or each sub-task with different sensitivity requirements; or each sub-task with different business processing accuracy; or each sub-task with different business data security requirements.

[0139] S124, determine the task type to which the first and second subtasks belong.

[0140] In this embodiment, the mobile edge computing device can determine the task type of the first subtask and the task type of the second subtask based on the computing power requirement information and service requirement information of the subtasks. Each task type corresponds to a minimum comprehensive trust value of the required computing power equipment. The task types of the first and second subtasks can be the same or different.

[0141] S125, determining the target computing device within the target computing device trust circle that matches the task type to which the mobile edge computing task belongs, including: determining the first target computing device for completing the first sub-task within the first target computing device trust circle that matches the task type to which the first sub-task belongs; S126, determining the target computing device within the target computing device trust circle that matches the task type to which the mobile edge computing task belongs, including: determining the second target computing device for completing the second sub-task within the second target computing device trust circle that matches the task type to which the second sub-task belongs.

[0142] In this embodiment of the application, by decomposing the mobile edge computing task into at least two sub-tasks based on computing power and business needs, and determining the target computing power device within the trust circle of one or more target computing power devices that match the type of each sub-task, the bottleneck problem of the upper limit of computing power of a single trusted computing power device can be overcome. This can improve the efficiency of parallel collaborative processing of multiple computing power devices in the system, thereby greatly improving the overall throughput of the task and significantly shortening the task response time.

[0143] In this embodiment, the mobile edge computing task is divided into at least two sub-tasks, including: when the task type of the mobile edge computing task is a first task type or a second task type, the mobile edge computing task is divided into at least two sub-tasks, wherein the task types of the first sub-task and the second sub-task are different. The task type may include a first task type, a second task type, and a third task type, wherein the minimum comprehensive trust value of the computing power device required for the task of the first task type is greater than the minimum comprehensive trust value of the computing power device required for the task of the second task type, and the minimum comprehensive trust value of the computing power device required for the task of the second task type is greater than the minimum comprehensive trust value of the computing power device required for the task of the third task type.

[0144] In this embodiment, by setting a tiered minimum comprehensive trust value threshold and forcibly breaking down higher-level computing tasks into sub-tasks with different types of requirements, the security and reliability of higher-level task processing can be better guaranteed. On the other hand, it can overcome the problem of excessive waste of high-level computing resources caused by the inconsistency of security and computing power requirements of each stage of the task. This enables accurate matching based on the actual characteristics of the sub-tasks, thereby maximizing the utilization of idle computing power in the trust circle of computing power devices at all levels and maximizing the utilization of global computing power resources of the system.

[0145] For example, in this embodiment, high-priority tasks can be split. For instance, when the computing power of a single computing device is insufficient, a task splitting algorithm based on multimodal data features can be used to divide the task into multiple sub-tasks. The computing power requirement of each sub-task matches the computing power of the computing devices within the trust circle, ensuring that the sub-tasks can be completed efficiently. Task splitting can follow the principle of "minimizing transmission overhead and maximizing parallel processing," and the splitting process is as follows: The multimodal parsing layer extracts features from the multimodal data corresponding to the task to determine key parameters such as the task's computing power requirements, data transmission volume, and processing priority. The scheduling layer can divide the task into n subtasks (n≥2) according to the task type and computing power requirements. The computing power requirement of each subtask does not exceed the maximum computing power of a single trust circle device, and the dependencies between subtasks are as low as possible to facilitate parallel processing. The scheduling layer can prioritize each subtask, and high-priority subtasks are preferentially allocated to devices with sufficient computing power and high trust level to ensure the overall efficiency of task completion.

[0146] For example, the image recognition task for factory area drone inspection can be broken down into three sub-tasks: image preprocessing (low computing power requirement), target detection (medium computing power requirement), and anomaly recognition (high computing power requirement). These sub-tasks can be assigned to computing devices in the temporary trust circle, the ordinary trust circle, and the core trust circle, respectively, thereby achieving parallel processing and shortening the task response time.

[0147] Figure 9 A schematic flowchart illustrating another method for determining a target computing power device provided in an embodiment of this application is shown. Figure 9 As shown, the method 120 includes: S127. Construct an objective function based on the task response latency of mobile edge computing tasks, network bandwidth usage, and the device computing power load balancing of the mobile edge computing system.

[0148] In this embodiment of the application, the objective function may include: ; in, To reduce the task response latency of mobile edge computing tasks, Network bandwidth usage for mobile edge computing tasks. To improve the load balancing of computing power in mobile edge computing systems, The weighting coefficients are non-negative, and ; S128: Based on the constraints of hardware parameters, operating status parameters, and comprehensive trust values ​​of the computing power equipment of the mobile edge computing system, the objective function is optimized and the target computing power equipment is determined.

[0149] In this embodiment of the application, the constraints may include: , , ; in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

[0150] Specifically, mobile edge computing devices can employ the Blackwing Kite optimization algorithm to offload and schedule tasks within a trust circle. The core objectives are to minimize task response latency, reduce network bandwidth usage, and balance device computing load. During scheduling, a scheduling objective function is constructed, using factors such as the trust value of devices within the trust circle, idle percentage, remaining battery power, computing power matching degree, and data transmission distance as core optimization objectives.

[0151] In this embodiment, the mobile edge computing device can employ the Black-winged Kite optimization algorithm to optimize and solve the objective function based on constraints, thereby determining the target computing power device. Alternatively, the mobile edge computing device can employ other optimization algorithms to determine the target computing power device. The solution process will be explained in detail below using the Black-winged Kite optimization algorithm as an example.

[0152] The Black-winged Kite Optimization Algorithm simulates the foraging behavior of black-winged kites to iteratively optimize task unloading schemes and quickly find the optimal subtask allocation scheme. The specific steps are as follows: Initialize the population: For example, treat each subtask allocation scheme as an individual and set the population size according to the number of subtasks (e.g., default population size = number of subtasks × 5). Fitness calculation: For example, using the objective function F as the fitness function, calculate the fitness value of each individual. The smaller the fitness value, the better the allocation scheme. Iterative optimization: For example, through selection, crossover, and mutation operations, individuals with good fitness values ​​are retained, while inferior individuals are eliminated, and the population is continuously updated iteratively until a preset number of iterations is reached (e.g., default number of iterations = 100) or the fitness value tends to stabilize. Output the optimal solution: After the iteration is completed, the individual with the smallest fitness value can be output, which is the optimal subtask unloading and allocation scheme to ensure that the task is completed efficiently and stably.

[0153] In this embodiment, by constructing an objective function using three core indicators—comprehensive task response latency, network bandwidth usage, and device computing power load balancing—and optimizing the solution under constraints of hardware, state, and trust values, the technical problem of finding the optimal resource allocation solution quickly by relying solely on simple rules in scenarios with massive heterogeneous devices and multiple concurrent tasks can be overcome. This achieves the technical effect of improving the solution accuracy of multi-objective combination optimization and quickly locking the optimal cross-device collaborative unloading strategy for tasks globally.

[0154] In this embodiment of the application, the objective function may include: ; in, To reduce the task response latency of mobile edge computing tasks, Network bandwidth usage for mobile edge computing tasks. To improve the load balancing of computing power in mobile edge computing systems, For computing power equipment Overall trust value The weighting coefficients are non-negative, and ; Constraints may include: , , ; in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

[0155] in, Weight the trust value (e.g., default) =0.2), This is a trust value penalty item; the lower the trust value, the larger the penalty. This helps avoid assigning tasks to devices with low trust levels. By adding a trust value penalty item, task allocation schemes can be further optimized, improving the security and reliability of task unloading.

[0156] In this embodiment, by defining a specific linear objective function formula and clearly defining constraints on remaining computing power, electricity consumption, and minimum trust value, a robust mathematical and logical boundary can be provided for the offloading method. This ensures that the algorithm will not over-allocate tasks during optimization, fully guaranteeing the safety baseline of the selected computing devices in terms of physical electricity consumption, computing power, and trust value, thereby improving the data security and reliability of task offloading. Furthermore, in this embodiment, by introducing "..." into the objective function... "This trust value penalty item for computing power devices can directly impose penalty costs on low-trust computing power devices at the mathematical calculation level, thereby effectively overcoming the defect of allocating tasks to idle but unreliable computing power devices in order to unilaterally pursue low latency, and further improving the data security and reliability of task offloading."

[0157] In this embodiment of the application, the device computing power load balancing degree of the mobile edge computing system It can be determined by the following equation: , in, The number of computing devices in a mobile edge computing system. For computing power equipment The computing power already used For computing power equipment Total computing power L represents the average computing power utilization rate of computing devices in a mobile edge computing system. The smaller the L value, the more balanced the computing power load of the computing devices, which can effectively avoid the situation where some devices are overloaded and others are idle.

[0158] In this embodiment of the application, by statistically analyzing the deviation between the computing power utilization ratio of each computing power device and the overall average computing power utilization rate of the system, the computing power load balance of the computing power devices can be specifically calculated and quantified. This can overcome the problem that some computing power devices are easily overloaded while others are idle during dynamic operation of the system. In this way, it can accurately measure and regulate the overall computing power distribution of the system and maintain the overall health and balance of the system.

[0159] Figure 10 This is a schematic diagram illustrating a task unloading execution process provided in an embodiment of this application. Figure 10 As shown, the method 300 includes: S310, data acquisition and sensing.

[0160] For example, the edge sensing devices in the perception layer, such as drones, unmanned vehicles, and edge servers, are responsible for collecting their own operational status data (such as remaining power and idle percentage), multimodal business data (images, audio, etc.), and usage habit data of inference services in real time, and transmitting this data to the parsing layer through the edge network.

[0161] S320, Multimodal Data Analysis.

[0162] For example, the multimodal parsing layer uses a large multimodal model to perform deep analysis and feature extraction on the multimodal data transmitted from the perception layer, and outputs key parameters such as task type recognition results, image / audio parsing accuracy, and device operating status, thereby providing data support for subsequent trust assessment and task scheduling.

[0163] S330, Trust Circle Construction and Update.

[0164] For example, the scheduling layer can calculate the comprehensive trust value for some or all computing devices in the mobile edge computing system based on the parameters passed by the multimodal parsing layer and the historical behavior data of the devices, through a multi-dimensional trust evaluation model. It can also classify each computing device into trust levels (core, ordinary, temporary) to complete the construction and dynamic elimination of trust circles.

[0165] S340, Task Reception and Classification.

[0166] For example, after receiving AI tasks (such as inspections and monitoring) at the edge of the factory, the scheduling layer can classify the tasks into three categories—high, medium, and low—based on their sensitivity, priority, and computing power requirements, and specify the corresponding minimum trust threshold.

[0167] S350, task splitting and scheduling.

[0168] For example, the scheduling layer can decompose complex tasks with medium to high priority into multiple sub-tasks using a task splitting algorithm, and can run the Blackwing Kite optimization algorithm to formulate the optimal task cross-device allocation scheme by combining the device's trust value, computing power, remaining power, etc.

[0169] S360, task execution and monitoring.

[0170] For example, the end devices in each trust circle of the perception layer are responsible for actually receiving and executing the assigned tasks or sub-tasks in parallel, while providing real-time feedback on progress and status. The scheduling layer, on the other hand, is responsible for monitoring the execution status in real time. If abnormal conditions such as insufficient power or task failure are detected, it will immediately trigger and implement a reallocation of sub-tasks.

[0171] S370, Trust Update and Result Feedback.

[0172] For example, after a task is completed, the scheduling layer can recalculate and update the overall trust value of each computing device using a dynamic trust update formula based on the task completion quality and operational status changes of each device. This allows for adjustments to the trust circle, and the processing results are then uniformly summarized and fed back to the plant management platform to complete the closed loop.

[0173] Therefore, this application embodiment determines the comprehensive trust value of computing power devices based on four dimensions: hardware parameters, operating status parameters, historical behavior parameters, and multimodal analysis parameters. It not only examines the physical capabilities of the computing power devices but also their historical security reputation and data parsing accuracy. Furthermore, by setting a "first trust threshold" to filter target computing power devices, computing tasks are ensured to circulate only among qualified and trusted computing power devices. This fundamentally avoids the risk of data leakage caused by blindly uninstalling, not only protecting data privacy and processing security but also effectively reducing task processing failure rates by matching actual operating status and improving the overall computing power collaboration efficiency of the system.

[0174] It should be understood that, in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0175] The above text combines Figures 1 to 10 The method for task unloading according to embodiments of this application is described in detail below. Figures 11 to 13 This application provides a detailed description of the relevant apparatus for task unloading according to embodiments of the present application.

[0176] Figure 11 This is a schematic block diagram of a task unloading device provided in an embodiment of this application. Figure 11 As shown, the device 600 includes: The acquisition module 610 is used to acquire task information of the mobile edge computing task of the mobile edge computing system. The task information includes computing power requirement information for completing the task. The determination module 620 is used to determine the target computing power device within the computing power device trust circle that includes at least one computing power device, based on the computing power demand information. The comprehensive trust value of each computing power device included in the computing power device trust circle is greater than or equal to a first trust threshold. The comprehensive trust value of each computing power device is determined based on the hardware parameters, operating status parameters, historical behavior parameters and multimodal parsing parameters of the computing power device. The transceiver module 630 is used to send mobile edge computing tasks to the target computing device.

[0177] In the embodiments of this application, such as Figure 12 As shown, the device 600 also includes: The acquisition module 640 is used to acquire the operating status data and multimodal parsing data of the computing power devices in the mobile edge computing system. The parsing module 650 is used to perform in-depth analysis on the collected operating status data and multimodal parsing data of computing power devices through a multimodal large model, and to obtain the operating status parameters and multimodal parsing parameters of the computing power devices.

[0178] In this embodiment of the application, the determining module 620 is further configured to: determine the comprehensive trust value of the computing power devices in the mobile edge computing system; and determine the set of computing power devices whose comprehensive trust value is greater than or equal to the first trust threshold as the computing power device trust circle; Among them, the comprehensive trust value of computing devices in mobile edge computing systems Determined by the following equation: , in, The hardware computing power trust value is determined based on the hardware parameters of the computing device. This refers to the operational status trust value of computing power equipment, determined based on operational status parameters. The historical behavior trust value of computing devices, determined based on historical behavior parameters. The multimodal analytical trust value for computing power devices, determined based on multimodal analytical parameters. The weighting coefficients are non-negative, and .

[0179] In this embodiment, the determining module 620 is specifically used to: determine that the computing power device in the mobile edge computing system is experiencing an abnormal situation; and determine the comprehensive trust value of the computing power device according to the following equation. , , in, This is the overall trust value before the computing equipment malfunctions. This is the overall trust value after a computing device malfunctions. It is a historical trust value weighting system, and .

[0180] In this application embodiment, the abnormal situation of the computing power device includes at least one of the following situations: abnormal operation of the computing power device, abnormal network of the computing power device, abnormal task of the computing power device, and abnormal trust level of the computing power device.

[0181] In this embodiment, the determining module 620 is further configured to: determine the task type to which the mobile edge computing task belongs based on the computing power requirement information and business requirement information of the mobile edge computing task, and the minimum comprehensive trust value of the computing power device required for each task type; determine the target computing power device within the target computing power device trust circle that matches the task type to which the mobile edge computing task belongs, wherein the comprehensive trust value of each computing power device included in the target computing power device trust circle is greater than or equal to a second trust threshold, and the second trust threshold is the minimum comprehensive trust value of the computing power device required for the task type to which the mobile edge computing task belongs, wherein the task information also includes business requirement information, and the business requirement information includes at least one of the following: business priority information, business sensitivity information, business processing accuracy information, business data security requirement information, and business response latency requirement information.

[0182] In this embodiment of the application, the determining module 620 is further configured to: divide the mobile edge computing task into at least two sub-tasks based on the computing power requirement information and business requirement information of the mobile edge computing task, wherein the at least two sub-tasks include a first sub-task and a second sub-task; determine the task type to which the first sub-task and the second sub-task belong; determine a first target computing power device for completing the first sub-task within a first target computing power device trust circle that matches the task type to which the first sub-task belongs; and determine a second target computing power device for completing the second sub-task within a second target computing power device trust circle that matches the task type to which the second sub-task belongs.

[0183] In this embodiment, the determining module 620 is specifically used to: when the mobile edge computing task belongs to a first task type or a second task type, split the mobile edge computing task into at least two sub-tasks, wherein the first sub-task and the second sub-task have different task types. The task types include a first task type, a second task type, and a third task type, wherein the minimum comprehensive trust value of the computing power equipment required for a task of the first task type is greater than the minimum comprehensive trust value of the computing power equipment required for a task of the second task type, and the minimum comprehensive trust value of the computing power equipment required for a task of the second task type is greater than the minimum comprehensive trust value of the computing power equipment required for a task of the third task type.

[0184] In this embodiment, the determining module 620 is specifically used to: construct an objective function based on the task response latency of the mobile edge computing task, the network bandwidth usage, and the device computing power load balancing of the mobile edge computing system; and optimize the objective function based on the constraints of the hardware parameters, operating status parameters, and comprehensive trust value of the computing power devices of the mobile edge computing system to determine the target computing power devices.

[0185] In this embodiment of the application, the objective function includes: ;in, To reduce the task response latency of mobile edge computing tasks, Network bandwidth usage for mobile edge computing tasks. To improve the load balancing of computing power in mobile edge computing systems, The weighting coefficients are non-negative, and The constraints include: , , ;in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

[0186] In this embodiment of the application, the objective function includes: ;in, To reduce the task response latency of mobile edge computing tasks, Network bandwidth usage for mobile edge computing tasks. To improve the load balancing of computing power in mobile edge computing systems, For computing power equipment Overall trust value The weighting coefficients are non-negative, and . In this embodiment of the application, the device computing power load balancing degree Determined by the following equation: , in, The number of computing devices in a mobile edge computing system. For computing power equipment The computing power already used For computing power equipment Total computing power This represents the average computing power utilization rate of computing devices in a mobile edge computing system.

[0187] like Figure 13 As shown in the figure, this application embodiment also provides a task offloading device 800, which may include a processor 810, a memory 820, and a bus system 830. The processor 810 and the memory 820 are connected via the bus system 830. The memory 820 is used to store instructions or computer programs, and the processor 810 is used to execute the instructions or computer programs stored in the memory 820, so that the processor 810 performs the following tasks: Figures 1 to 10 The method described in [the document / document].

[0188] This application also provides an electronic device, which includes units or modules for implementing the methods described in the above embodiments.

[0189] In the embodiments of this application, the electronic device includes, for example, a smart interactive display screen, a smart screen terminal, a display screen, an MEC node, an edge box, an edge server, smart glasses, a smart terminal, an edge personal computer, a drone, or an unmanned vehicle.

[0190] This application also provides a computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the methods described above.

[0191] This application also provides a computer program product, which includes computer program code that, when run on a computer, causes the computer to perform the methods described in the above embodiments.

[0192] This application also provides a computer program that, when run, enables the methods described in the above embodiments to be implemented.

[0193] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0194] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0195] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0196] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0197] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0198] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0199] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for offloading mobile edge computing tasks, characterized in that, include: Obtain task information for mobile edge computing tasks in a mobile edge computing system, including computing power requirement information for completing the task. Within a computing power device trust circle that includes at least one computing power device, a target computing power device is determined based on the computing power demand information. The comprehensive trust value of each computing power device included in the computing power device trust circle is greater than or equal to a first trust threshold. The comprehensive trust value of the computing power device is determined based on the hardware parameters, operating status parameters, historical behavior parameters, and multimodal parsing parameters of the computing power device. The mobile edge computing task is sent to the target computing device.

2. The method according to claim 1, characterized in that, The method further includes: The mobile edge computing system collects the operating status data and multimodal analysis data of the computing devices. The operating status data and multimodal analysis data of the collected computing power equipment are deeply analyzed by using a multimodal large model to obtain the operating status parameters and multimodal analysis parameters of the computing power equipment.

3. The method according to claim 1 or 2, characterized in that, The method further includes: Determine the overall trust value of the computing devices in the mobile edge computing system; The set of computing power devices whose comprehensive trust value is greater than or equal to the first trust threshold is defined as the computing power device trust circle. Among them, the comprehensive trust value of the computing devices in the mobile edge computing system Determined by the following equation: ， in, The hardware computing power trust value of the computing power device is determined based on the hardware parameters. The operating status trust value of the computing power device, determined based on the operating status parameters. The historical behavior trust value determined by the computing power device based on the historical behavior parameters. The multimodal analytical trust value determined by the computing power device based on the multimodal analytical parameters. The weighting coefficients are non-negative, and .

4. The method according to claim 3, characterized in that, Determining the overall trust value of the computing devices in the mobile edge computing system includes: It was determined that the computing power equipment in the mobile edge computing system was malfunctioning; The overall trust value of the computing power device is determined according to the following equation. , ， in, This is the overall trust value before the computing power device malfunctions. This is the overall trust value after the computing power device malfunctions. It is a historical trust value weighting system, and .

5. The method according to claim 4, characterized in that, The abnormal conditions of the computing power device include at least one of the following: abnormal operation of the computing power device, abnormal network of the computing power device, abnormal task of the computing power device, and abnormal trust level of the computing power device.

6. The method according to any one of claims 1 to 5, characterized in that, The task information also includes business requirement information, which includes at least one of the following: business priority information, business sensitivity information, business processing accuracy information, business data security requirements information, and business response delay requirements information. The method further includes: Based on the computing power requirement information and the business requirement information of the mobile edge computing task, the task type to which the mobile edge computing task belongs is determined, and the minimum comprehensive trust value of the computing power equipment required for each task type is determined. The step of determining the target computing power device based on the computing power demand information within a computing power device trust circle that includes at least one computing power device includes: The target computing device is determined within the trust circle of the target computing device that matches the task type to which the mobile edge computing task belongs. The comprehensive trust value of each computing device included in the trust circle of the target computing device is greater than or equal to a second trust threshold. The second trust threshold is the minimum comprehensive trust value of the computing device required for the task type to which the mobile edge computing task belongs.

7. The method according to claim 6, characterized in that, The method further includes: Based on the computing power requirement information and the business requirement information of the mobile edge computing task, the mobile edge computing task is divided into at least two sub-tasks, the at least two sub-tasks including a first sub-task and a second sub-task; Determine the task types to which the first subtask and the second subtask belong; The step of determining the target computing power device within the trust circle of the target computing power device that matches the task type to which the mobile edge computing task belongs includes: Within the trust circle of the first target computing power device that matches the task type to which the first sub-task belongs, determine the first target computing power device used to complete the first sub-task; Within the trust circle of the second target computing power device that matches the task type to which the second subtask belongs, determine the second target computing power device used to complete the second subtask.

8. The method according to claim 7, characterized in that, The task types include a first task type, a second task type, and a third task type, wherein the minimum comprehensive trust value of the computing power equipment required for the task of the first task type is greater than the minimum comprehensive trust value of the computing power equipment required for the task of the second task type, and the minimum comprehensive trust value of the computing power equipment required for the task of the second task type is greater than the minimum comprehensive trust value of the computing power equipment required for the task of the third task type. The step of splitting the mobile edge computing task into at least two sub-tasks includes: If the task type of the mobile edge computing task is either the first task type or the second task type, the mobile edge computing task is split into at least two sub-tasks, wherein the task types of the first sub-task and the second sub-task are different.

9. The method according to any one of claims 1 to 8, characterized in that, The target computing power device includes: Based on the task response latency of the mobile edge computing task, the network bandwidth usage, and the device computing power load balancing of the mobile edge computing system, an objective function is constructed. Based on the constraints of the hardware parameters, operating status parameters, and comprehensive trust value of the computing power device of the mobile edge computing system, the objective function is optimized and solved to determine the target computing power device.

10. The method according to claim 9, characterized in that, The objective function includes: ; in, The task response latency of the mobile edge computing task, This refers to the network bandwidth usage of the mobile edge computing task. The device computing power load balancing of the mobile edge computing system. The weighting coefficients are non-negative, and ; The constraints include: , , ; in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

11. The method according to claim 9, characterized in that, The objective function includes: ; in, The task response latency of the mobile edge computing task, This refers to the network bandwidth usage of the mobile edge computing task. The device computing power load balancing of the mobile edge computing system. For computing power equipment Overall trust value The weighting coefficients are non-negative, and ; The constraints include: , , ; in, For computing power equipment The remaining computing power For computing power equipment Undertake the task Required computing power For computing power equipment The remaining battery power, The minimum battery level threshold, For computing power equipment Overall trust value In order to complete the mission The minimum overall trust value of the computing power equipment required for each task type.

12. The method according to claim 9 or 10, characterized in that, The device computing power load balancing of the mobile edge computing system Determined by the following equation: ， in, This refers to the number of computing devices in the mobile edge computing system. For computing power equipment The computing power already used For computing power equipment Total computing power This refers to the average computing power utilization rate of the computing devices in the mobile edge computing system.

13. An apparatus for offloading mobile edge computing tasks, characterized in that, Includes units or modules for implementing the method as described in any one of claims 1 to 12.

14. An apparatus for offloading mobile edge computing tasks, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program stored in the memory to cause the apparatus to perform the method as described in any one of claims 1 to 12.

Images