Long-term resource allocation method and device for heterogeneous edge network, equipment and storage medium

The long-term average energy consumption problem of edge networks is decomposed through the Liyapunov optimization method, and the resource optimization of heterogeneous edge networks is achieved, solving the problem of inefficient resource management in the medium and long-term technology, ensuring the stability and energy efficiency of the system.

CN118945126APending Publication Date: 2024-11-12PENG CHENG LAB
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410937504.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing edge network resource optimization solutions mainly focus on short-term cost minimization, throughput maximization or latency minimization, ignoring the tolerance of some applications for latency and the long-term performance of the system, resulting in inefficiency in long-term operation. High energy consumption and poor network stability.

Method used

The Liyapunov optimization method is used to optimize the long-term average energy consumption problem of heterogeneous edge networks, and decompose it into a single time slot optimization problem. The computing resource allocation of each edge server, computing resource allocation of equipment and transmit power allocation are obtained through decoupling. Problems are solved separately to achieve long-term resource optimization.

Benefits of technology

It effectively solves the long-term resource management problems of heterogeneous edge networks, ensures the long-term stability and energy efficiency optimization of the system, and enables it to operate efficiently in a dynamic environment for a long time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118945126A_ABST
    Figure CN118945126A_ABST
Patent Text Reader

Abstract

The invention discloses a long-term resource allocation method and device for a heterogeneous edge network, equipment and a storage medium, and relates to the technical field of the Internet of Things, and the method comprises the steps: carrying out the problem optimization of an original long-term average energy consumption problem corresponding to the heterogeneous edge network based on a Lyapunov optimization mode, and determining a single time slot optimization problem; performing problem decoupling on the single time slot optimization problem to obtain a computing resource allocation problem of each edge server, a computing resource allocation problem of each device and a transmitting power allocation problem of each device; and solving the problem to obtain the target computing resource of each edge server, the target computing resource of each device and the target transmitting power of each device in each time slot so as to perform long-term resource allocation of the heterogeneous edge network. Through the mode, the problem of long-term resource management of the heterogeneous edge network can be effectively solved, the long-term stability and energy efficiency optimization of the system are ensured, and the heterogeneous edge network can efficiently run in a dynamic environment for a long time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of Internet of Things technology, and in particular to a method, device, equipment and storage medium for long-term resource allocation in a heterogeneous edge network. Background Art

[0002] With the rapid expansion of IoT applications, traditional IoT devices are inefficient in processing large amounts of computing tasks. Mobile edge computing (MEC) technology has emerged to reduce latency and improve service quality by providing computing power at the edge of the network. However, in the face of the rapid expansion of IoT applications, the increase in base station deployment density and the number of users, improper resource allocation will lead to higher energy consumption and lower spectrum efficiency. Especially in an environment where heterogeneous base stations coexist, how to optimize resources to reduce interference and improve network performance becomes the key. Compared with a single MEC server, servers in a heterogeneous edge network have different computing and transmission capabilities, and there are significant differences in user task types. This makes the resource allocation design of heterogeneous edge networks supporting MEC more complicated.

[0003] However, existing edge network resource optimization solutions mainly focus on minimizing costs, maximizing throughput, or minimizing latency in the short term, ignoring the tolerance of certain applications to latency and the long-term performance of the system. Short-term optimization methods are often inefficient in long-term operation and cannot adapt to dynamically changing environments, resulting in high energy consumption and poor network stability. In addition, due to the random arrival and time correlation of computing tasks, long-term optimal decisions are challenging. Summary of the invention

[0004] The main purpose of this application is to provide a long-term resource allocation method, device, equipment and storage medium for a heterogeneous edge network, aiming to solve the technical problem of how to achieve long-term resource optimization of a heterogeneous edge network.

[0005] To achieve the above objectives, the present application proposes a long-term resource allocation method for a heterogeneous edge network, the long-term resource allocation method for a heterogeneous edge network comprising:

[0006] Based on the Lyapunov optimization method, the original long-term average energy consumption problem corresponding to the heterogeneous edge network is optimized to determine the single time slot optimization problem;

[0007] Decoupling the single time slot optimization problem, obtaining the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device;

[0008] Solve the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device respectively, and obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device;

[0009] Long-term resource allocation of the heterogeneous edge network is performed based on target computing resources of each edge server, target computing resources of each device, and target transmission power of each device in each time slot.

[0010] In one embodiment, the problem optimization of the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method, before determining the single time slot optimization problem, also includes:

[0011] The task volume is defined through the digital twin network corresponding to the heterogeneous edge network to determine the actual task volume of each device and the time slot unloading task volume of each device;

[0012] Derivation of problematic conditions is performed based on the actual task load of each device and the time slot unloading task load of each device to determine the task constraint conditions;

[0013] A problem is constructed according to the task constraints, the server energy consumption of each edge server, and the device energy consumption of each device to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

[0014] In one embodiment, the task volume is defined by the digital twin network corresponding to the heterogeneous edge network to determine the actual task volume of each device and the time slot unloading task volume of each device, including:

[0015] Determine the expected processing rate of each device and the processing rate deviation value of each device through the digital twin network corresponding to the heterogeneous edge network;

[0016] The task volume is defined according to the task processing cycle of each device, the time slot length, the expected processing rate of each device and the processing rate deviation of each device, and the actual task volume of each device is determined;

[0017] The task volume is defined according to the channel power gain information of each device, the target total bandwidth and the device association information of each device, and the time slot offloading task of each device is determined.

[0018] In one embodiment, the problem condition derivation is performed based on the actual task amount of each device and the time slot unloading task amount of each device to determine the task constraint condition, including:

[0019] Determine the target task volume of each device based on the actual task volume of each device and the time slot unloading task volume of each device;

[0020] Determine the first queue length in the task buffer of each device according to the target task amount and the arrived task amount of each device;

[0021] Defining the task volume according to the device association information of each edge device, the estimated processing rate of each edge server, the processing rate deviation of each edge server, and the task processing cycle of each edge server, and determining the processing task volume of each edge server;

[0022] Determine the second queue length in the task buffer of each edge server according to the processing task amount of each edge server and the time slot unloading task amount of each device;

[0023] A task constraint set is determined according to a network stability definition, the first queue length, and the second queue length.

[0024] In one embodiment, the problem construction is performed according to the task constraint conditions, the server energy consumption of each edge server, and the device energy consumption of each device to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network, including:

[0025] Energy consumption is defined based on the energy consumption weight of each edge server, the energy consumption weight of each device, the server energy consumption of each edge server, and the device energy consumption of each device, and the long-term average energy consumption of the heterogeneous edge network is determined;

[0026] Construct resource constraints based on server resource information of each edge server and device resource information of each device;

[0027] A problem is constructed according to the resource optimization goal, the long-term average energy consumption, the resource constraints and the task constraints to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

[0028] In one embodiment, the problem optimization of the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method to determine the single time slot optimization problem includes:

[0029] A function is defined according to a first queue length in a task buffer of each device and a second queue length in a task buffer of each edge server to determine a target Lyapunov function;

[0030] determining a conditional Lyapunov drift based on the target Lyapunov function;

[0031] determining a target Lyapunov optimization function based on the conditional Lyapunov drift, the stability control parameter, and the long-term network energy consumption of the heterogeneous edge network;

[0032] The original long-term average energy consumption problem corresponding to the heterogeneous edge network is optimized according to the target Lyapunov optimization function to determine a single time slot optimization problem.

[0033] In one embodiment, respectively solving the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device to obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device include:

[0034] Solve the computing resource allocation problem of each device in each time slot in a closed form to obtain the target computing resource of each device in each time slot;

[0035] The transmission power allocation problem of each device in each time slot is solved by target fractional programming to obtain the target transmission power of each device in each time slot;

[0036] The computing resource allocation problem of each edge server in each time slot is solved by the Lagrangian method of target binary search, and the target computing resources of each edge server in each time slot are obtained.

[0037] In addition, to achieve the above purpose, the present application also proposes a long-term resource allocation device for a heterogeneous edge network, and the long-term resource allocation device for a heterogeneous edge network includes:

[0038] The optimization module is used to optimize the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method and determine the single time slot optimization problem;

[0039] A decoupling module is used to decouple the single time slot optimization problem to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device;

[0040] A solution module is used to solve the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, respectively, to obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device;

[0041] The allocation module is used to perform long-term resource allocation of the heterogeneous edge network based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

[0042] In addition, to achieve the above-mentioned purpose, the present application also proposes a long-term resource allocation device for a heterogeneous edge network, the device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the long-term resource allocation method for a heterogeneous edge network as described above.

[0043] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the long-term resource allocation method for heterogeneous edge networks as described above are implemented.

[0044] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the long-term resource allocation method for heterogeneous edge networks as described above.

[0045] The present application optimizes the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method, and determines the single time slot optimization problem; decouples the single time slot optimization problem to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device; solves the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, respectively, to obtain the target computing resource of each edge server, the target computing resource of each device, and the target transmission power of each device; based on the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device, the long-term resource allocation of the heterogeneous edge network is performed. In the above way, using the Lyapunov optimization method, the complex non-convex optimization problem of long-term energy minimization is decomposed into a low computational complexity problem with a deterministic upper bound for each time slot. Then, the low computational complexity problem is decoupled into three independent sub-problems, and each sub-problem is solved separately, which can effectively solve the problem of long-term resource management of heterogeneous edge networks, ensure the long-term stability and energy efficiency optimization of the system, and enable heterogeneous edge networks to operate efficiently and long-term in a dynamic environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0048] Figure 1 A flow chart of a first embodiment of a method for long-term resource allocation in a heterogeneous edge network of the present application;

[0049] Figure 2 A schematic diagram of a heterogeneous edge network assisted by a digital twin of the long-term resource allocation method for a heterogeneous edge network provided in Example 1 of the present application;

[0050] Figure 3 A flow chart of the second embodiment of the long-term resource allocation method for heterogeneous edge networks provided in the present application;

[0051] Figure 4 A flow chart of a third embodiment of a method for long-term resource allocation in a heterogeneous edge network of the present application;

[0052] Figure 5 A flow chart of a fourth embodiment of a method for long-term resource allocation in a heterogeneous edge network of the present application;

[0053] Figure 6 A brief flowchart of a method for long-term resource allocation in a heterogeneous edge network provided in Embodiment 4 of the present application;

[0054] Figure 7 A schematic diagram of the module structure of a long-term resource allocation device for a heterogeneous edge network according to an embodiment of the present application;

[0055] Figure 8 A schematic diagram of the device structure of the hardware operating environment involved in the long-term resource allocation method for a heterogeneous edge network in an embodiment of the present application.

[0056] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0058] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0059] The main solution of the embodiment of the present application is: based on the Lyapunov optimization method, the original long-term average energy consumption problem corresponding to the heterogeneous edge network is optimized to determine the single time slot optimization problem; the single time slot optimization problem is decoupled to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device; the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device are solved respectively to obtain the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device; the long-term resource allocation of the heterogeneous edge network is performed based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

[0060] Existing edge network resource optimization solutions mainly focus on minimizing costs, maximizing throughput, or minimizing latency in the short term, ignoring the tolerance of certain applications to latency and the long-term performance of the system. Short-term optimization methods are often inefficient in long-term operation and cannot adapt to dynamically changing environments, resulting in high energy consumption and poor network stability. In addition, due to the random arrival and time correlation of computing tasks, long-term optimal decisions are challenging.

[0061] The present application provides a solution, which uses the Lyapunov optimization method to decompose the complex non-convex optimization problem of long-term energy minimization into a low computational complexity problem with a deterministic upper bound for each time slot. Then, the low computational complexity problem is decoupled into three independent sub-problems, and each sub-problem is solved separately, which can effectively solve the problem of long-term resource management of heterogeneous edge networks, ensure the long-term stability and energy efficiency optimization of the system, and enable heterogeneous edge networks to operate efficiently and long-term in a dynamic environment.

[0062] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or a long-term resource allocation device of a heterogeneous edge network capable of realizing the above functions, etc. The following takes the long-term resource allocation device of a heterogeneous edge network as an example of the execution subject to illustrate this embodiment and the following embodiments.

[0063] Based on this, the embodiment of the present application provides a long-term resource allocation method for a heterogeneous edge network, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the long-term resource allocation method for heterogeneous edge networks of the present application.

[0064] In this embodiment, the long-term resource allocation method of the heterogeneous edge network includes steps S10 to S40:

[0065] Step S10, optimizing the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method, and determining the single time slot optimization problem.

[0066] It should be noted that the method of this embodiment is aimed at edge networks with heterogeneous cellular base stations. In order to solve the problems of long-term resource management and maintaining network stability, a long-term resource management solution based on Lyapunov optimization of joint communication and computing is designed, with the goal of maintaining network stability in the long term and minimizing system energy consumption.

[0067] It can be understood that in this embodiment, a system model of a digital twin-assisted heterogeneous edge network is proposed to derive the long-term energy consumption cost and network stability of devices and edge servers. Then, under the constraint of long-term network stability, the resource allocation problem of joint communication and computing is described with the goal of minimizing the energy consumption of the system in the long term. The system model of a digital twin-assisted heterogeneous edge network in this embodiment consists of a heterogeneous edge network and its digital twin network, such as Figure 2 shown.

[0068] In the specific implementation, in the heterogeneous edge network, there are 1 macro base station and M small base stations, the set is M = {0, 1, ..., M}. All base stations are equipped with corresponding edge servers. There are N devices in the system, the set is N = {1, ..., N}. The directed graph G = (N, M, ε) is used to represent the network, and ε represents the association between the device and the base station. If device n is connected to base station m, the link will be recorded in the edge set ε. The system time is a discrete time slot with a time slot length of τ. For convenience, the index set of the time slot is represented as T = {0, 1, ... T}.

[0069] It should be noted that in the digital twin network, the device collects the data that needs to be processed, while ensuring that the edge server has enough data for modeling the digital twin system. Small base stations can process the data unloaded by devices in the coverage area, while locally building virtual models of devices and small base stations, and synchronizing parameters to macro base stations. If the devices are not within the coverage of any small base station, they will unload the data to the macro base station with sufficient computing power and synchronize the parameters. The digital twin system of the heterogeneous edge network can be built and maintained on the macro base station, and the virtual model includes the wireless network topology, the communication model between the device and the base station, the status and channel information, etc. Based on these virtual models and data, efficient mathematical optimization or intelligent algorithms can be used to learn optimization strategies and provide policy feedback for the operation of the physical network in real time.

[0070] It can be understood that digital twins can establish a digital representation system similar to the real environment to obtain the estimated performance value of the system without knowing the implementation details of the devices and edge servers in the system.s It can be expressed as: DT s ={E s ,R s}. s is the state of the edge server in the entire MEC system, R s It is a resource allocation strategy based on feedback. Based on the real-time update information of physical objects, the digital services of the digital twin layer provide comprehensive functions to automatically manage and control the system. These services can be listed as data collection, visualization, analysis, decision-making, and real-time optimization. Digital twin services need to provide optimization solutions in a timely manner in terms of task offloading, estimated processing rate, and transmission power allocation to ensure the performance of the entire system. For edge server m, its corresponding digital twin It can be expressed as: in represents the estimated processing rate of the edge server in the digital twin, is the deviation value of the processing rate estimation between the real edge server and its digital twin. For device n, its corresponding digital twin It can be expressed as: in is the expected processing rate of device n in the digital twin, is the estimated deviation in processing rate between the real device and its digital twin.

[0071] In the specific implementation, assume that the location of IoT device n is l n , and the location of the macro base station / small base station m is l m Therefore, the distance from device n to the macro base station / small base station m is r nm =||l n -l m ||.d m Defined as the coverage radius of all small base stations m. If the distance between device n and small base station m is less than d m (Right now ,r nm <d m ), then device n can communicate with the base station. Here, we consider that the device communicates with the nearest base station to perform computation offloading. Using the associated variable a mn Indicates the association between device n and edge server m, a mn =1 indicates that device n is associated with edge server m, and the edge server can serve the offloading task of the device, where That is, device n can only be associated with one edge server. Therefore, the association variable is

[0072] It should be noted that the wireless channel between the IoT device and the edge server is an independent and identically distributed frequency flat block fading. The power gain of the small-scale fading channel from device n to small base station m in the tth time slot is expressed as γ nm (t), assuming it has a bounded mean, i.e. The power gain of a small-scale fading channel is exponentially distributed with a unit mean, i.e., γ nm (t)~Exp(λ). Therefore, the channel power gain from device n to edge server m is: Γ nm (t)≡γ nm (t)g 0 (r 0 / r nm ) θ . Where g 0 is the path loss constant, θ is the path loss exponent, r 0 is the reference distance. Given the total bandwidth W, according to the Shannon-Hartley theorem, the wireless communication data rate between device n and base station m can be expressed as in p tx,n (t) is the transmission power, and its maximum value is p tx,max .

[0073] It should be noted that the actual task volume of device n can be calculated by calculating the CPU cycles required for device n to calculate a bit of computing task, the expected processing rate of device n, and the estimated deviation of the processing rate between the real device and its digital twin. The data rate of wireless communication between device n and base station m The amount of unloaded tasks of device n during time slot t can be determined n∈N.

[0074] In the specific implementation, in this embodiment, the long-term network stability of the system is described by a dual-task queue model on the device and the server. Assume that the device is running an independent fine-grained task: At the beginning of the tth time slot, the A of the calculation task n (t) bits arrive at IoT device n, and can be processed starting from time slot t+1. Without loss of generality, assume that A in different time slots n (t) In [A n,min ,A n,max ] are independent and identically distributed, and obey the average distribution A n (t)~[A n,min ,A n,max ],E[A n (t)] = λ, n∈N. In each time slot, the number of The computation task of bits will be offloaded to the edge server. The tasks that have arrived but not yet offloaded will be queued in the task buffer of each mobile device with a large enough capacity, and the queue length of the task buffer at the beginning of the tth time slot is denoted as Q n (t). where Q n (t+1)=max{Q n (t)-D n (t),0}+A n (t). Among them, represents the amount of tasks computed locally and offloaded from device n to the associated edge server m.

[0075] It should be noted that each edge server maintains a corresponding task buffer for device n within its coverage area, which is used to store tasks that have been offloaded but not yet executed. The queue length of the task buffer for maintaining device n starting at time slot t on the base station is represented as T n (t). Dynamic update formula of queue length: Among them, n (t) is the size of the computational task leaving the edge task buffer. The CPU cycles provided by the edge server can be allocated to the tasks offloaded from different devices for computation. The task volume scheduling decision of the edge server in the tth time slot can be determined by the CPU computational resources allocated by the edge server m to the device n within the coverage range (i.e., ): Assume that in order to process one bit of computational task offloaded from mobile device n to base station m, L m CPU cycles. The computing resources of the edge server can be allocated to multiple user devices to maintain their digital twins in the server. represents the estimated computing resources allocated to the digital twin of IoT device n. This constraint ensures that the number of CPU cycles planned for task processing on the edge server should not be greater than the available CPU cycles on the edge server. According to the definition of network stability, the task queue is stable if all computing tasks satisfy the following constraints:

[0076] It can be understood that the energy consumption of the heterogeneous edge network consists of four parts. For local devices, the first is the energy consumption for local calculations, and the second is the energy consumption for uplink wireless transmission for offloading. For base station m, the third is the energy consumption of calculations with computing resources allocated to associated devices. The fourth is the energy consumption of downlink wireless transmission to offload the calculation results to the device. Since the size of the result is very small, the energy consumption of downlink transmission is ignored in this embodiment.

[0077] In the specific implementation, for the energy consumption of the device: In the digital twin system, the energy consumption of executing tasks locally Where κ is a constant representing the effective switching capacitance of the local device CPU, which is related to the chip structure. Assuming that the deviation between the IoT device and its digital twin can be obtained in advance, the computational energy consumption gap between the estimated value and the actual value of the digital twin is The actual computing energy consumption of device n in local computing tasks It can be obtained that the transmission power consumption of local device n is: The energy consumed by the local device n includes local computing energy consumption and transmission power energy consumption:

[0078] It should be noted that for edge servers, the energy consumption depends on how busy the edge server is, which can be calculated by the available CPU frequency. The energy consumption of executing tasks on edge server m is Where ζ is a constant representing the effective switching capacitance of the edge server m CPU, which is related to the chip structure. Assuming that the deviation between the edge server and its digital twin can be obtained in advance, the computational energy consumption gap between the estimated value and the actual value of the digital twin is It can be obtained by the following formula: The actual computing energy consumption of the edge server m allocated to the offload task of device n can be obtained by and The sum of is obtained and can be expressed as:

[0079] It is understandable that in this embodiment, the focus is on the energy consumption of the task execution process in the device and edge server CPU, as well as the transmission energy consumption for computing offload. The energy consumed for other purposes, such as powering the screen of the mobile device and supporting basic operations in the MEC system, is ignored for simplicity. Therefore, the long-term average energy consumption of the heterogeneous cellular network is used as the performance indicator, which is defined as follows: For convenience, use Among them, w 1 ,w 2 are the energy consumption weights of edge servers and local devices, respectively. These parameters can be adjusted to solve the power consumption costs of different nodes in the MEC system and balance the power consumption of mobile devices and MEC servers. In this paper, these parameters are assumed to be constant. In the digital twin-assisted heterogeneous cellular network resource optimization, the optimization goal is to minimize the long-term total energy consumption of the system:

[0080]

[0081] in

[0082] Constraint a represents the maximum computing resources owned by base station m. Constraint b represents the resource upper limit of IoT device n. Constraint c is the maximum value constraint of the transmission power of any IoT device n. Constraint d ensures the stability of the device task queue. The mandatory task buffer is average rate stable, which ensures that all arriving computing tasks can be completed within a bounded delay. Variable 1: Estimated CPU computing frequency assigned to IoT device n by edge server m on the digital twin Variable 2: Estimated CPU computation frequency of IoT device n on the digital twin Variable 3: p tx,n (t) is the transmission power of IoT device n. In this embodiment, the original long-term average energy consumption problem of the heterogeneous edge network refers to

[0083] In specific implementation, Lyapunov optimization algorithm is a powerful method for solving optimization problems with long-term goals and constraints, which requires less prior information about task arrival, channel state information and task buffer. The main idea behind Lyapunov optimization is to transform the optimization problem with long-term goals into a series of sub-problems with short-term goals and transform the long-term constraints into constraints with queue stability. In addition, Lyapunov optimization optimizes sub-problems through online algorithms with low computational complexity.

[0084] It should be noted that in order to solve the original long-term average energy consumption problem of heterogeneous edge networks, the original long-term average energy consumption problem is decoupled into a simpler single-slot optimization problem through the Lyapunov optimization method:

[0085] V is a control parameter, which is used to balance the optimal objective function and the stability of the system queue.

[0086] Step S20, decoupling the single time slot optimization problem to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device.

[0087] It should be noted that the single-time slot optimization problem is still a highly complex non-convex problem. In order to solve this problem, we decompose the problem into three sub-problems: the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device. By alternately solving the sub-problems, the algorithm can reach a convergence state.

[0088] Step S30, respectively solving the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, to obtain the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

[0089] It should be noted that in each time slot, the device CPU computing resource variables for the computing resource allocation problem of each device are determined in the form of a closed solution. In each time slot, for the device transmit power variables for the transmit power allocation problem of each device, a fractional programming algorithm based on quadratic transformation is proposed.

[0090] In each time slot, for the CPU computing resource variables of the edge servers in the computing resource allocation problem of each edge server, a Lagrangian algorithm based on binary search is proposed. After solving the problem in the above way, the target computing resources of each device, the target transmission power of each device, and the target computing resources of each edge server in each time slot can be obtained, thereby obtaining the optimal resource configuration for each time slot.

[0091] Step S40, performing long-term resource allocation of the heterogeneous edge network based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

[0092] It should be noted that based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device, the optimal resource configuration of the heterogeneous edge network in each time slot can be obtained, and the device and edge server task queues can be updated through the formula until the optimal resource allocation strategy is obtained to achieve long-term resource allocation for the heterogeneous edge network.

[0093] This embodiment optimizes the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method to determine the single time slot optimization problem; decouples the single time slot optimization problem to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device; solves the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, respectively, to obtain the target computing resource of each edge server, the target computing resource of each device, and the target transmission power of each device; based on the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device, the long-term resource allocation of the heterogeneous edge network is performed. In the above way, using the Lyapunov optimization method, the complex non-convex optimization problem of long-term energy minimization is decomposed into a low computational complexity problem with a deterministic upper bound for each time slot. Then, the low computational complexity problem is decoupled into three independent sub-problems, and each sub-problem is solved separately, which can effectively solve the problem of long-term resource management of heterogeneous edge networks, ensure the long-term stability and energy efficiency optimization of the system, and enable heterogeneous edge networks to operate efficiently and long-term in a dynamic environment.

[0094] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 3 , before step S10, the long-term resource allocation method for heterogeneous edge networks further includes steps S01 to S03:

[0095] Step S01, define the task volume through the digital twin network corresponding to the heterogeneous edge network, and determine the actual task volume of each device and the time slot unloading task volume of each device.

[0096] It should be noted that the expected processing rate of device n can be determined through the digital twin network corresponding to the heterogeneous edge network. The estimated deviation of the processing rate between device n and its digital twin By computing one bit of computing task through device n, the required L n CPU cycles, as well as The actual amount of tasks that can be performed locally by device n and the amount of offloaded tasks of device n during time slot t In this embodiment, the amount of unloaded tasks of device n during time slot t is That is the time slot unloading task amount of the device.

[0097] In a feasible implementation, step S01 may include steps A21 to A23:

[0098] Step A21, determining the expected processing rate of each device and the processing rate deviation value of each device through the digital twin network corresponding to the heterogeneous edge network.

[0099] It should be noted that the estimated deviation of the processing rate between device n and its digital twin is This is the processing rate deviation of the device.

[0100] Step A22, defining the task volume according to the task processing cycle of each device, the time slot length, the expected processing rate of each device and the processing rate deviation of each device, and determining the actual task volume of each device.

[0101] It should be noted that the task processing cycle of a device refers to the CPU cycle required for device n to calculate a computing task, and the time slot length is τ. The actual task volume of each device can be determined by the task processing cycle, time slot length, expected processing rate of each device, and processing rate deviation of each device. Specifically, assuming that IoT device n has a certain computing power, the task volume of the local computing is Where L n : To calculate a bit of computing task for a hypothetical device n, it will take L n CPU cycles. There is a deviation between the predicted value and the actual value of the digital twin. Therefore, the actual task load of device n is

[0102] Step A23, defining the task amount according to the channel power gain information of each device, the target total bandwidth and the device association information of each device, and determining the time slot offloading task of each device.

[0103] It should be noted that the channel power gain information includes the channel power gain from the device to the edge server. The target total bandwidth refers to the given total bandwidth W. The device association information refers to the association relationship between the device and the edge server, where Given the total bandwidth W, according to the Shannon-Hartley theorem, the wireless communication data rate between device n and base station m can be expressed as In time slot t, the device sends its task to the Unload to the base station. The unloading task of device n during time slot t can be expressed as

[0104] Step S02, deriving problematic conditions based on the actual task load of each device and the time slot unloading task load of each device, and determining task constraint conditions.

[0105] It should be noted that the long-term network stability of the system is described by a dual-task queue model on the device and the server. Assume that the device is running independent fine-grained tasks: At the beginning of the tth time slot, the A of the calculation task n (t) bits arrive at IoT device n, and can be processed starting from time slot t+1. Without loss of generality, assume that A in different time slots n (t) In [A n,min ,A n,max ] are independent and identically distributed, and obey the average distribution A n (t)~[A n,min ,A n,max ],E[A n (t)] = λ, n∈N. In each time slot, the number of The computation task of bits will be offloaded to the edge server. The tasks that have arrived but not yet offloaded will be queued in the task buffer of each mobile device with a large enough capacity, and the queue length of the task buffer at the beginning of the tth time slot is denoted as Q n (t). where Q n (t+1)=max{Q n (t)-D n (t),0}+A n (t). Among them, represents the amount of tasks computed locally and offloaded from device n to the associated edge server m.

[0106] It can be understood that each edge server maintains a corresponding task buffer for device n within the coverage area, which is used to store tasks that have been offloaded but not yet executed. The queue length of the task buffer for maintaining device n starting at time slot t on the base station is represented as T n (t). Dynamic update formula of queue length: Among them, n (t) is the size of the computational task leaving the edge task buffer. The CPU cycles provided by the edge server can be allocated to the tasks offloaded from different devices for computation. The task volume scheduling decision of the edge server in the tth time slot can be determined by the CPU computational resources allocated by the edge server m to the device n within the coverage range (i.e., ): Assume that in order to process one bit of computational task offloaded from mobile device n to base station m, L mCPU cycles. The computing resources of the edge server can be allocated to multiple user devices for maintaining their digital twins in the server. This constraint ensures that the number of CPU cycles scheduled for task processing on the edge server should not be greater than the available CPU cycles on the edge server. According to the definition of network stability, the task queue is stable if all computing tasks satisfy the following constraints: In this embodiment, These are task constraints.

[0107] In a feasible implementation, step S02 may include steps B21 to B25:

[0108] Step B21, determining the target task volume of each device based on the actual task volume of each device and the time slot unloading task volume of each device.

[0109] It should be noted that the target task volume of a device refers to the sum of the actual task volume of the device and the time slot unloading task volume of the device. The target task volume of device n is:

[0110] Step B22, determining the first queue length in the task buffer of each device according to the target task volume and the arrived task volume of each device.

[0111] It should be noted that the first queue length in the task buffer of the device refers to the queue length Q of the task buffer of the device at the beginning of the tth time slot. n (t), evolves according to the following equation: Q n (t+1)=max{Q n (t)-D n (t),0}+A n (t).

[0112] Step B23, defining the task volume according to the device association information of each edge device, the estimated processing rate of each edge server, the processing rate deviation of each edge server and the task processing cycle of each edge server, and determining the processing task volume of each edge server.

[0113] It should be noted that the estimated processing rate of edge server m is The processing rate deviation of edge server m refers to the deviation value of the processing rate estimation between the edge server and its digital twin The task processing cycle of edge server m refers to the CPU cycles required to process a one-bit computing task offloaded from mobile device n to edge server m.

[0114] It can be understood that the processing task volume of the edge server refers to the size of the computing tasks that leave the edge task buffer. Each edge server maintains a corresponding task buffer for device n within the coverage area to store tasks that have been offloaded but not yet executed. The processing task volume of edge server m is The CPU cycles provided by the edge server can be allocated to the tasks offloaded from different devices for computation. The task scheduling decision of the edge server in the tth time slot can be determined by the CPU computing resources allocated by the edge server m to the device n within the coverage area (i.e. ). Where L m For edge server m to process a bit of computing task offloaded from mobile device n to base station m, it will require L m CPU cycles.

[0115] In a specific implementation, the computing resources of the edge server can be allocated to multiple user devices to maintain their digital twins in the server. represents the estimated computational resources allocated to the digital twin of IoT device n. This constraint ensures that the number of CPU cycles scheduled for task processing on the edge server should not be greater than the available CPU cycles on the edge server.

[0116] Step B24: determining the second queue length in the task buffer of each edge server according to the processing task volume of each edge server and the time slot unloading task volume of each device.

[0117] It should be noted that the second queue length refers to the queue length of the task buffer of the maintenance device n starting from time slot t on the edge server. The second queue length T n (t) can be determined based on the processing task volume of the edge server and the time slot unloading task volume of each device. The dynamic update formula of the second queue length is:

[0118] Step B25, determining task constraints according to the network stability definition, the first queue length, and the second queue length.

[0119] It should be noted that the task constraint conditions can be determined according to the network stability definition, the first queue length, and the second queue length, which are:

[0120] Step S03, constructing a problem according to the task constraints, the server energy consumption of each edge server, and the device energy consumption of each device, and determining the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

[0121] It should be noted that the server energy consumption of the edge server is the actual computing energy consumption allocated to the device offloading task by the edge server, where the server energy consumption allocated to device n by edge server m is The device energy consumption includes the local computing energy consumption and transmission power energy consumption of the device. The device energy consumption of device n is e n (t).

[0122] It is understandable that for the device energy consumption of the device: in the digital twin system, the energy consumption of executing tasks locally Where κ is a constant representing the effective switching capacitance of the local device CPU, which is related to the chip structure. Assuming that the deviation between the IoT device and its digital twin can be obtained in advance, the computational energy consumption gap between the estimated value and the actual value of the digital twin is The actual computing energy consumption of device n in local computing tasks It can be obtained that the transmission power consumption of local device n is: The energy consumed by the local device n includes local computing energy consumption and transmission power energy consumption:

[0123] It should be noted that for edge servers, the energy consumption of the server is calculated based on how busy the edge server is, which can be calculated by the available CPU frequency. The energy consumption of executing tasks on edge server m is Where ζ is a constant representing the effective switching capacitance of the edge server m CPU, which is related to the chip structure. Assuming that the deviation between the edge server and its digital twin can be obtained in advance, the estimation of the digital twin

[0124] It can be understood that by constructing the problem based on the task constraints, the server energy consumption of each edge server, and the device energy consumption of each device, the original long-term average energy consumption problem corresponding to the heterogeneous edge network can be obtained:

[0125]

[0126] Constraint a represents the maximum computing resources owned by base station m. Constraint b represents the resource upper limit of IoT device n. Constraint c is the maximum value constraint of the transmission power of any IoT device n. Constraint d ensures the stability of the device task queue. The mandatory task buffer is average rate stable, which ensures that all arriving computing tasks can be completed within a bounded delay.

[0127] In a feasible implementation, step S03 may include steps C21 to C23:

[0128] Step C21, energy consumption is defined according to the energy consumption weight of each edge server, the energy consumption weight of each device, the server energy consumption of each edge server, and the device energy consumption of each device, to determine the long-term average energy consumption of the heterogeneous edge network.

[0129] It should be noted that this embodiment focuses on the energy consumption of the task execution process in the device and edge server CPU, as well as the transmission energy consumption for computing offload. The energy consumed for other purposes, such as powering the screen of the mobile device and supporting basic operations in the MEC system, is ignored for simplicity. Therefore, the long-term average energy consumption of the heterogeneous cellular network is used as the performance indicator, and the long-term average energy consumption of the heterogeneous edge network is defined as follows: Among them, w 1 ,w 2 are the energy consumption weights of the edge server and the local device, respectively. These parameters can be adjusted to solve the power consumption cost of different nodes in the MEC system and balance the power consumption of mobile devices and MEC servers. In this embodiment, these parameters are assumed to be constants.

[0130] It is understandable that for convenience, it is expressed as

[0131] Step C22: construct resource constraint conditions according to the server resource information of each edge server and the device resource information of each device.

[0132] It should be noted that the server resource information of the edge server includes but is not limited to the maximum computing resources owned by the edge server / base station, and the device resource information of the device includes but is not limited to the resource upper limit of the device. The resource constraint conditions are constructed through the server resource information of each edge server, the device resource information of each device, and the stability conditions of the device task queue, specifically: Constraint a represents the maximum computing resources owned by base station m. Constraint b represents the resource upper limit of IoT device n. Constraint c is the maximum value constraint of the transmission power of any IoT device n.

[0133] Step C23, constructing a problem according to the resource optimization goal, the long-term average energy consumption, the resource constraints and the task constraints, and determining the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

[0134] It should be noted that based on the task constraints, constraint d can be obtained: Based on resource constraints, constraint d, and the average energy consumption of the previous period, combined with the resource optimization goal of minimizing the system's long-term total energy consumption, the original long-term average energy consumption problem is determined:

[0135]

[0136] This embodiment defines the task volume through the digital twin network corresponding to the heterogeneous edge network, determines the actual task volume of each device and the time slot unloading task volume of each device; derives the problem conditions based on the actual task volume of each device and the time slot unloading task volume of each device, and determines the task constraints; constructs the problem based on the task constraints, the server energy consumption of each edge server, and the device energy consumption of each device, and determines the original long-term average energy consumption problem corresponding to the heterogeneous edge network. In the above manner, the accuracy of the problem construction is guaranteed, laying an accurate foundation for subsequent long-term resource optimization.

[0137] Based on the above embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the above embodiment 1 / embodiment 2 can be referred to the above introduction, and will not be described in detail later. Figure 5 , step S10, the long-term resource allocation method for the heterogeneous edge network further includes steps S11 to S14:

[0138] Step S11, defining a function according to the first queue length in the task buffer of each device and the second queue length in the task buffer of each edge server, and determining a target Lyapunov function.

[0139] It should be noted that the Lyapunov optimization algorithm is a powerful method for solving optimization problems with long-term goals and constraints, which requires less prior information about task arrival, channel state information, and task buffers. The main idea behind Lyapunov optimization is to transform the optimization problem with long-term goals into a series of sub-problems with short-term goals, and transform the long-term constraints into constraints with queue stability. In addition, Lyapunov optimization optimizes sub-problems through an online algorithm with low computational complexity.

[0140] It can be understood that the length of the first queue is Q n (t), the length of the second queue is T n (t). In order to construct the Lyapunov optimization framework, the Lyapunov function is defined as The target Lyapunov function is where Θ(t)≡[Q n (t),T n (t)] represents the length of the task queue maintained by the device locally and on the edge server.

[0141] Step S12: determining a conditional Lyapunov drift based on the target Lyapunov function.

[0142] It should be noted that the conditional Lyapunov drift can be obtained through the target Lyapunov function, specifically ΔL(Θ(t))=E[L(Θ(t+1))-L(Θ(t))|Θ(t)]. By minimizing L(Θ(t)), the length of the task queue can be shortened to a smaller value.

[0143] Step S13, determining a target Lyapunov optimization function according to the conditional Lyapunov drift, the stability control parameter, and the long-term network energy consumption of the heterogeneous edge network.

[0144] It should be noted that the stability control parameter V refers to a control parameter in the algorithm, which is used to balance the optimal objective function and the stability of the system queue. At the same time, it actually reflects the "emphasis" of decision makers on the system queue backlog and cost expenditure. The average weighted energy consumption of the heterogeneous edge network refers to The target Lyapunov optimization function refers to the Lyapunov drift plus penalty function.

[0145] It can be understood that based on the conditional Lyapunov drift, stability control parameters, and the long-term network energy consumption of the heterogeneous edge network, the target Lyapunov optimization function Δ V L(Θ(t))=ΔL(Θ(t))+V·E[e Σ (t)|Θ(t)].

[0146] Step S14, optimizing the original long-term average energy consumption problem corresponding to the heterogeneous edge network according to the target Lyapunov optimization function, and determining a single time slot optimization problem.

[0147] It should be noted that by minimizing Δ V L(Θ(t)) can ensure network stability while minimizing the system energy consumption of the network. First, find Δ V Upper bound of (Θ(t)):

[0148] Where C is a constant that is independent of the variable Θ(t). C is expressed as in

[0149] It is understandable that through the Lyapunov optimization framework, the original long-term average energy consumption problem corresponding to the heterogeneous edge network is decoupled into a simpler single-time slot optimization problem, which only needs to be solved in each time slot:

[0150]

[0151] In the specific implementation, the main idea of ​​the subsequent algorithm solution is to greedily minimize Upper limit on the right.

[0152] This embodiment determines the target Lyapunov function by defining a function based on the first queue length in the task buffer of each device and the second queue length in the task buffer of each edge server; determines the conditional Lyapunov drift based on the target Lyapunov function; determines the target Lyapunov optimization function based on the conditional Lyapunov drift, stability control parameters and the long-term network energy consumption of the heterogeneous edge network; optimizes the original long-term average energy consumption problem corresponding to the heterogeneous edge network according to the target Lyapunov optimization function to determine the single time slot optimization problem. In the above manner, the complex non-convex optimization problem of long-term energy minimization can be decomposed into a low computational complexity problem with a deterministic upper bound for each time slot, which is convenient for subsequent resource allocation solutions.

[0153] Based on the above embodiments of the present application, in the fourth embodiment of the present application, the same or similar contents as those in the above embodiments 1 / 2 / 3 can be referred to the above introduction, and will not be described in detail later. Figure 5 , step S30, the long-term resource allocation method for the heterogeneous edge network further includes steps S31 to S33:

[0154] Step S31, solving the computing resource allocation problem of each device in each time slot in a closed form, and obtaining the target computing resources of each device in each time slot.

[0155] It should be noted that in order to solve the single time slot optimization problem, the single time slot optimization problem is decomposed into three sub-problems: the computing resource allocation problem of each device in each time slot, the transmission power allocation problem of each device in each time slot, and the computing resource allocation problem of each edge server in each time slot. By alternately solving the sub-problems, the algorithm can reach a convergence state.

[0156] It can be understood that, in this embodiment, using To calibrate the CPU calculation frequency error, the influence of digital twin deviation on problem optimization is eliminated. To solve the problem of computing resource allocation for each device in each time slot: First, fix the CPU frequency variable allocated by the edge server and the transmission power variable of the device, and obtain the computing resource allocation problem for each device in each time slot as Only keep The related item is the CPU resource scheduling problem of local device n. In this embodiment, the deviation values ​​α and β between the physical real value and the CPU computing frequency of the IoT device and server estimated by the corresponding digital twin are obtained, and the formula Calibrate the CPU calculation frequency error.

[0157] In the specific implementation, the form of the computing resource allocation problem of each device in each time slot is mainly an affine function plus a cubic power function (the variable is greater than or equal to 0), and the constraint is a convex set. Therefore, it is easy to conclude that this subproblem is a convex optimization problem, because the constraints are decoupled. Therefore, the computing resource allocation problem of each device in each time slot can be decoupled into the optimization problem of each device n: By using the Karush-Kuhn-Tucker (KKT) condition, the closed-form solution is obtained as n∈N. That is, the target computing resource of device n in time slot t.

[0158] Step S32, solving the transmission power allocation problem of each device in each time slot by target fractional programming, and obtaining the target transmission power of each device in each time slot.

[0159] It should be noted that the transmission power allocation problem of each device in each time slot is: Only keep the tx (t) The related coupling term, i.e., the allocation sub-problem of wireless communication resources. The target fractional programming method refers to a direct iterative algorithm of fractional programming based on quadratic transformation.

[0160] It is understandable that the key to solving the problem of allocating transmit power to each device in each time slot is However, due to the existence of interference, the transmission power p of each IoT device n tx,n (t) are mutually coupled. Even if each transmit power constraint of device n is decoupled. And for Q n (t) and T n The queue state of (t) is unknown, which will directly affect the convexity of the function. Therefore, the above problem is a complex non-convex problem. However, the transformation method based on the fractional programming of quadratic transformation makes the problem solvable. For Q n (t)≤T n (t), for the case that satisfies Q n (t)≤T n The device set in state (t) is represented as: when For p tx,n (t) is non-decreasing, that is For p tx,n (t) is a non-decreasing function. The objective function can be kept at the minimum value. For a collection of IoT devices, the transmission power allocation problem for each device in each time slot is simplified to: .in

[0161] In this embodiment, due to the interference term, the variables of the device transmission power are coupled to each other. This is a complex non-convex problem, and the fractional programming problem of quadratic transformation is selected for solution. Proposition 1: The transmission power allocation problem of each device in each time slot is transformed into a concave-convex fractional programming form through quadratic transformation, and is directly solved using an iterative method. The following problem P TX Equivalent to the formula It can be expressed as: Among them, the objective function is: Due to R n The signal to interference plus noise ratio (SINR) in (t) is in the form of a fraction, so the power control problem is a typical sum problem of multiple proportional functions. The partial objective function is a non-decreasing concave logarithmic function. The numerator and denominator of SINR are both convex and concave functions, and can be solved using the concave-convex fractional programming method.

[0162] In the specific implementation, from the above proposition 1, the transmission power allocation problem of each device in each time slot is transformed into the form of concave and convex fractional programming through quadratic transformation, and an iterative method is required to solve the convex optimization problem. After applying the quadratic transformation to each SINR item, the following restatement is obtained: Where y is the set y n ∈y is introduced by the secondary transformation of each device n uplink. And the expression is shown in the following formula.

[0163] where ω n =[Q n (t)-T n (t)]τW>0. Optimize p iteratively tx (t) and y. Fixed p tx (t), the optimal The closed-form solution is Then, find a fixed y n The optimal p tx (t) is a convex problem.

[0164] It should be noted that in order to obtain the optimal After that, the optimal transmission power allocation of device n in time slot t can be calculated Optimal transmit power allocation for device n This is the target transmit power of the device.

[0165] It is understandable that the direct iterative algorithm of fractional programming needs to solve the convex problem in each iteration. It is worth mentioning that the direct iterative algorithm of fractional programming can be initialized with a simple but reasonable heuristic method. For example, the initial power can be set to half of the maximum power. In the simulation results of this embodiment, in order to ensure the fairness of the comparison, a random starting point is used and then the results are averaged. In addition, some small constants δ>0 are set in this embodiment, and the convergence criterion is used where t is the iteration index.

[0166] In the specific implementation, the detailed process of the direct iterative algorithm of fractional programming based on quadratic transformation is as follows:

[0167] 1. Initialization: Get the device and edge server task queue status, location and channel gain, and assign an initial value to the device n transmission power. n (t) and T n (t) Task queue status classification. Get the set of devices that meet the problem Set the number of iterations I MAX and algorithm accuracy δ.

[0168]

[0169] Step S33, solving the computing resource allocation problem of each edge server in each time slot by using the Lagrangian method of target binary search to obtain the target computing resources of each edge server in each time slot.

[0170] It should be noted that the Lagrangian method of target binary search refers to the Lagrangian method based on binary search. The computing resource allocation problem of each edge server in each time slot is Only keep The related item is the CPU resource scheduling problem of the edge server. Using the formula To calibrate the error of the CPU calculation frequency estimated by the digital twin, and eliminate the impact of the digital twin deviation on the problem optimization. β is the error percentage value regularly predicted by the digital twin system, which maintains synchronization with the physical network data. Obviously, according to the modeling problem of this embodiment, when a nm = 0, that is, there is no service relationship between device n and server m, take and Restrict device n to only be associated with one edge server m.

[0171] Understandably, a nm= 1, the form of the computing resource allocation problem of each edge server in each time slot is mainly an affine function plus a cubic power function (the variable is greater than or equal to 0), and the constraint is a convex set. Therefore, it is easy to conclude that this subproblem is a convex optimization problem. However, due to The cumulative constraints make it impossible to decouple the constraints. This is more challenging because the resource allocation of server m CPU is coupled between different associated devices. However, the Lagrangian method provides an effective solution to the problem of computing resource allocation for each edge server in each time slot. Specifically, the partial Lagrangian function of the problem of computing resource allocation for each edge server in each time slot can be written as

[0172] become in The corresponding Lagrange multiplier λ m (t)≥0. According to the KKT condition, in The only valid root. In this embodiment, the following expression Depend on have to It can be seen that λ m (t) About Monotonically decreasing, where Therefore, we can use the binary search method to find L (t),λ U (t)], find the optimal λ L (t) and λ U (t) respectively satisfy In range Perform a binary search. The search process is terminated. In each iteration, the greatest complexity comes from the Lagrangian method of optimal CPU frequency allocation, which * (t) uses a binary search and divides the Terminate within iteration.

[0173] In the specific implementation, the optimal CPU frequency allocated to the edge server m can be obtained through the above method: The optimal CPU frequency assigned to edge server m That is the target computing resource of the edge server.

[0174] In this embodiment, the detailed process of the Lagrangian method based on binary search is as follows:

[0175] 1. Initialize relevant parameters, algorithm accuracy ε = 10 -3 ,λ U (t) = λL (t), maximum number of iterations Ι MAX =100, eta=1.5, m=0, t=0, n∈N,m∈M.

[0176]

[0177]

[0178] This embodiment solves the computing resource allocation problem of each device in each time slot in a closed form to obtain the target computing resource of each device in each time slot; solves the transmission power allocation problem of each device in each time slot in a target fractional programming manner to obtain the target transmission power of each device in each time slot; solves the computing resource allocation problem of each edge server in each time slot in a Lagrangian method of target binary search to obtain the target computing resource of each edge server in each time slot. Through the above methods, the optimal resource configuration of the heterogeneous edge network in each time slot can be accurately obtained.

[0179] For example, in order to help understand the implementation process of the long-term resource allocation method for heterogeneous edge networks obtained by combining the above-mentioned embodiments 1, 2 and 3, please refer to Figure 6 , Figure 6 A brief flow chart of a long-term resource allocation method for a heterogeneous edge network is provided, specifically:

[0180] The long-term resource allocation method for heterogeneous edge networks proposed in this embodiment. Under the constraint of long-term stability of the network, a long-term queue-aware energy minimization scheme for joint communication and computing resource management is proposed for the problem of minimizing the long-term total energy consumption of heterogeneous edge network systems. The steps of the method are as follows: 1. First, the original problem P1 of the long-term time slot is decoupled using the Lyapunov optimization framework. The complex non-convex problem of minimizing long-term energy consumption is decomposed into a problem P2 with a low computational complexity of the deterministic upper bound of each time slot. 2. Then, according to the optimization variables and constraints, problem P2 is decoupled into three independent sub-problems, each of which is solved separately. Specifically, P2 is decomposed into three sub-problems (i.e., problem P3: computing resource allocation problem of the device; problem P4: transmission power allocation problem of the device; problem P5: computing resource allocation problem of the edge server). 3. In each time slot, the device CPU computing resource variable of problem P3 is determined in the form of a closed solution. 4. In each time slot, for the device transmission power variable of problem P4, a fractional programming algorithm based on quadratic transformation is proposed. The details are shown in the following Algorithm 2 (ie, a direct iterative algorithm of fractional programming based on quadratic transformation).

[0181] 5. In each time slot, for the CPU computing resource variables of the edge server of problem P5, a Lagrangian algorithm based on binary search is proposed. The details are shown in the following Algorithm 3 (i.e., Lagrangian method based on binary search). 6. After performing steps 3 to 5, the optimal resource configuration for each time slot can be obtained, and the device and edge server task queues are updated through the formula until the optimal resource allocation strategy is obtained.

[0182] In this embodiment, the complete process of the above method is:

[0183] 1. Initialization: At time slot t = 0, obtain the state R of the system in the digital twin system s , including the task queue status of the device and edge server, the physical location and the wireless channel gain, and setting the step size δ.

[0184] For t=0:T

[0185] 2. At time slot t, obtain the device task queue status Q in the digital twin system n (t),T n (t); channel gain Γ nm (t).

[0186] 3. Solve the problem P2 for each time slot by alternately solving the three sub-problems (the computing resource allocation problem P3 of the device, the transmission power allocation problem P4 of the device, and the computing resource allocation problem P5 of the edge server) to determine the optimal CPU frequency allocated to device n Optimal transmit power of device n The optimal CPU frequency assigned to edge server m 4.if t modδ=0then

[0187] Get the deviation values ​​α, β between the physical real value and the CPU computing frequency of the IoT device and server estimated by the corresponding digital twin, and use the formula Calibrate the CPU calculation frequency error.

[0188] end if

[0189] 5. Through formula Q n (t+1)=max{Q n (t)-D n (t),0}+A n (t) Update device and edge server task queues Q n (t),T n (t); Set t=t+1.

[0190] END For

[0191] Output: optimal resource allocation strategy R s .

[0192] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the long-term resource allocation method for heterogeneous edge networks of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0193] This application also provides a long-term resource allocation device for a heterogeneous edge network, please refer to Figure 7 , the long-term resource allocation device of the heterogeneous edge network includes:

[0194] The optimization module 10 is used to optimize the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method to determine the single time slot optimization problem;

[0195] A decoupling module 20 is used to decouple the single time slot optimization problem to obtain a computing resource allocation problem for each edge server, a computing resource allocation problem for each device, and a transmission power allocation problem for each device;

[0196] A solution module 30 is used to solve the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, respectively, to obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device;

[0197] The allocation module 40 is used to perform long-term resource allocation of the heterogeneous edge network based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

[0198] Optionally, the optimization module 10 is further used for:

[0199] The task volume is defined through the digital twin network corresponding to the heterogeneous edge network to determine the actual task volume of each device and the time slot unloading task volume of each device; the problem conditions are derived based on the actual task volume of each device and the time slot unloading task volume of each device to determine the task constraints; the problem is constructed based on the task constraints, the server energy consumption of each edge server and the device energy consumption of each device to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

[0200] Optionally, the optimization module 10 is further used for:

[0201] The estimated processing rate of each device and the processing rate deviation of each device are determined through the digital twin network corresponding to the heterogeneous edge network; the task volume is defined according to the task processing cycle of each device, the time slot length, the estimated processing rate of each device and the processing rate deviation of each device, and the actual task volume of each device is determined; the task volume is defined according to the channel power gain information of each device, the target total bandwidth and the device association information of each device, and the time slot offloading task of each device is determined.

[0202] Optionally, the optimization module 10 is further used for:

[0203] The target task volume of each device is determined based on the actual task volume of each device and the time slot unloading task volume of each device; the first queue length in the task buffer of each device is determined according to the target task volume and the arrived task volume of each device; the task volume is defined according to the device association information of each edge device, the expected processing rate of each edge server, the processing rate deviation of each edge server and the task processing cycle of each edge server, and the processing task volume of each edge server is determined; the second queue length in the task buffer of each edge server is determined according to the processing task volume of each edge server and the time slot unloading task volume of each device; and the task constraint conditions are determined according to the network stability definition, the first queue length and the second queue length.

[0204] Optionally, the optimization module 10 is further used for:

[0205] Energy consumption is defined according to the energy consumption weight of each edge server, the energy consumption weight of each device, the server energy consumption of each edge server and the device energy consumption of each device, and the long-term average energy consumption of the heterogeneous edge network is determined; resource constraints are constructed according to the server resource information of each edge server and the device resource information of each device; and a problem is constructed according to the resource optimization goal, the long-term average energy consumption, the resource constraints and the task constraints, and the original long-term average energy consumption problem corresponding to the heterogeneous edge network is determined.

[0206] Optionally, the optimization module 10 is further used for:

[0207] A function is defined according to the first queue length in the task buffer of each device and the second queue length in the task buffer of each edge server to determine the target Lyapunov function; a conditional Lyapunov drift is determined based on the target Lyapunov function; a target Lyapunov optimization function is determined according to the conditional Lyapunov drift, stability control parameters and the long-term network energy consumption of the heterogeneous edge network; an original long-term average energy consumption problem corresponding to the heterogeneous edge network is optimized according to the target Lyapunov optimization function to determine a single time slot optimization problem.

[0208] Optionally, the solution module 30 is further used for:

[0209] The computing resource allocation problem of each device in each time slot is solved in the form of a closed solution to obtain the target computing resources of each device in each time slot; the transmission power allocation problem of each device in each time slot is solved by the target fractional programming method to obtain the target transmission power of each device in each time slot; the computing resource allocation problem of each edge server in each time slot is solved by the Lagrangian method of target binary search to obtain the target computing resources of each edge server in each time slot.

[0210] The long-term resource allocation device for heterogeneous edge networks provided by the present application adopts the long-term resource allocation method for heterogeneous edge networks in the above-mentioned embodiments, which can solve the technical problem of how to achieve long-term resource optimization of heterogeneous edge networks. Compared with the prior art, the beneficial effects of the long-term resource allocation device for heterogeneous edge networks provided by the present application are the same as the beneficial effects of the long-term resource allocation method for heterogeneous edge networks provided by the above-mentioned embodiments, and other technical features in the long-term resource allocation device for heterogeneous edge networks are the same as the features disclosed in the above-mentioned embodiment methods, which will not be repeated here.

[0211] The present application provides a long-term resource allocation device for a heterogeneous edge network, and the long-term resource allocation device for a heterogeneous edge network includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the long-term resource allocation method for the heterogeneous edge network in the above-mentioned embodiment 1.

[0212] Reference below Figure 8 , which shows a schematic diagram of the structure of a long-term resource allocation device for a heterogeneous edge network suitable for implementing an embodiment of the present application. The long-term resource allocation device for a heterogeneous edge network in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The long-term resource allocation device for the heterogeneous edge network shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0213] like Figure 8As shown, the long-term resource allocation device for heterogeneous edge networks may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the long-term resource allocation device for heterogeneous edge networks are also stored in RAM1004. The processing device 1001, ROM1002, and RAM1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 may allow the long-term resource allocation device of the heterogeneous edge network to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a long-term resource allocation device of a heterogeneous edge network with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0214] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0215] The long-term resource allocation device for heterogeneous edge networks provided by the present application adopts the long-term resource allocation method for heterogeneous edge networks in the above-mentioned embodiment, which can solve the technical problem of how to achieve long-term resource optimization of heterogeneous edge networks. Compared with the prior art, the beneficial effects of the long-term resource allocation device for heterogeneous edge networks provided by the present application are the same as the beneficial effects of the long-term resource allocation method for heterogeneous edge networks provided by the above-mentioned embodiment, and the other technical features in the long-term resource allocation device for heterogeneous edge networks are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0216] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0217] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0218] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the long-term resource allocation method for heterogeneous edge networks in the above-mentioned embodiment.

[0219] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0220] The computer-readable storage medium may be included in the long-term resource allocation device of the heterogeneous edge network; or may exist independently without being assembled into the long-term resource allocation device of the heterogeneous edge network.

[0221] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the long-term resource allocation device of the heterogeneous edge network, the long-term resource allocation device of the heterogeneous edge network: optimizes the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method to determine the single time slot optimization problem; decouples the single time slot optimization problem to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device; solves the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, respectively, to obtain the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device; and performs long-term resource allocation of the heterogeneous edge network based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

[0222] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0223] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0224] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0225] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned long-term resource allocation method for heterogeneous edge networks, and can solve the technical problem of how to achieve long-term resource optimization of heterogeneous edge networks. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the long-term resource allocation method for heterogeneous edge networks provided in the above-mentioned embodiments, and will not be repeated here.

[0226] The present application also provides a computer program product, including a computer program, which implements the steps of the long-term resource allocation method for a heterogeneous edge network as described above when executed by a processor.

[0227] The computer program product provided in this application can solve the technical problem of how to achieve long-term resource optimization of heterogeneous edge networks. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the long-term resource allocation method for heterogeneous edge networks provided in the above embodiments, which will not be repeated here.

[0228] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A long-term resource allocation method for a heterogeneous edge network, characterized in that: The long-term resource allocation method of the heterogeneous edge network includes: Based on the Lyapunov optimization method, the original long-term average energy consumption problem corresponding to the heterogeneous edge network is optimized to determine the single time slot optimization problem; Decoupling the single time slot optimization problem, obtaining the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device; Solve the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device respectively, and obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device; Long-term resource allocation of the heterogeneous edge network is performed based on target computing resources of each edge server, target computing resources of each device, and target transmission power of each device in each time slot.

2. The method according to claim 1, characterized in that The problem optimization of the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method, before determining the single time slot optimization problem, also includes: The task volume is defined through the digital twin network corresponding to the heterogeneous edge network to determine the actual task volume of each device and the time slot unloading task volume of each device; Derivation of problematic conditions is performed based on the actual task load of each device and the time slot unloading task load of each device to determine the task constraint conditions; A problem is constructed according to the task constraints, the server energy consumption of each edge server, and the device energy consumption of each device to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

3. The method according to claim 2, characterized in that The task volume is defined by the digital twin network corresponding to the heterogeneous edge network to determine the actual task volume of each device and the time slot unloading task volume of each device, including: Determine the expected processing rate of each device and the processing rate deviation value of each device through the digital twin network corresponding to the heterogeneous edge network; The task volume is defined according to the task processing cycle of each device, the time slot length, the expected processing rate of each device and the processing rate deviation of each device, and the actual task volume of each device is determined; The task volume is defined according to the channel power gain information of each device, the target total bandwidth and the device association information of each device, and the time slot offloading task of each device is determined.

4. The method according to claim 2, characterized in that Derivation of problematic conditions based on the actual task load of each device and the time slot unloading task load of each device to determine the task constraint conditions includes: Determine the target task volume of each device based on the actual task volume of each device and the time slot unloading task volume of each device; Determine the first queue length in the task buffer of each device according to the target task amount and the arrived task amount of each device; Defining the task volume according to the device association information of each edge device, the estimated processing rate of each edge server, the processing rate deviation of each edge server, and the task processing cycle of each edge server, and determining the processing task volume of each edge server; Determine the second queue length in the task buffer of each edge server according to the processing task amount of each edge server and the time slot unloading task amount of each device; The task constraint condition is determined according to the network stability definition, the first queue length, and the second queue length.

5. The method according to claim 2, characterized in that The problem construction is performed according to the task constraint conditions, the server energy consumption of each edge server, and the device energy consumption of each device to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network, including: Energy consumption is defined based on the energy consumption weight of each edge server, the energy consumption weight of each device, the server energy consumption of each edge server, and the device energy consumption of each device, and the long-term average energy consumption of the heterogeneous edge network is determined; Construct resource constraints based on server resource information of each edge server and device resource information of each device; A problem is constructed according to the resource optimization goal, the long-term average energy consumption, the resource constraints and the task constraints to determine the original long-term average energy consumption problem corresponding to the heterogeneous edge network.

6. The method according to claim 1, characterized in that The Lyapunov optimization method is used to optimize the original long-term average energy consumption problem corresponding to the heterogeneous edge network, and determine the single time slot optimization problem, including: A function is defined according to a first queue length in a task buffer of each device and a second queue length in a task buffer of each edge server to determine a target Lyapunov function; determining a conditional Lyapunov drift based on the target Lyapunov function; determining a target Lyapunov optimization function based on the conditional Lyapunov drift, the stability control parameter, and the long-term network energy consumption of the heterogeneous edge network; The original long-term average energy consumption problem corresponding to the heterogeneous edge network is optimized according to the target Lyapunov optimization function to determine a single time slot optimization problem.

7. The method according to claim 1, characterized in that The computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device are solved respectively to obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device, including: Solve the computing resource allocation problem of each device in each time slot in a closed form to obtain the target computing resource of each device in each time slot; The transmission power allocation problem of each device in each time slot is solved by target fractional programming to obtain the target transmission power of each device in each time slot; The computing resource allocation problem of each edge server in each time slot is solved by the Lagrangian method of target binary search, and the target computing resources of each edge server in each time slot are obtained.

8. A long-term resource allocation device for a heterogeneous edge network, characterized in that: The long-term resource allocation device of the heterogeneous edge network includes: The optimization module is used to optimize the original long-term average energy consumption problem corresponding to the heterogeneous edge network based on the Lyapunov optimization method and determine the single time slot optimization problem; A decoupling module is used to decouple the single time slot optimization problem to obtain the computing resource allocation problem of each edge server, the computing resource allocation problem of each device, and the transmission power allocation problem of each device; A solution module is used to solve the computing resource allocation problem of each edge server in each time slot, the computing resource allocation problem of each device, and the transmission power allocation problem of each device, respectively, to obtain the target computing resource of each edge server in each time slot, the target computing resource of each device, and the target transmission power of each device; The allocation module is used to perform long-term resource allocation of the heterogeneous edge network based on the target computing resources of each edge server in each time slot, the target computing resources of each device, and the target transmission power of each device.

9. A long-term resource allocation device for a heterogeneous edge network, characterized in that: The long-term resource allocation device for the heterogeneous edge network includes: a memory, a processor, and a long-term resource allocation program for the heterogeneous edge network stored in the memory and executable on the processor, wherein the long-term resource allocation program for the heterogeneous edge network is configured to implement a long-term resource allocation method for the heterogeneous edge network as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a long-term resource allocation program for a heterogeneous edge network, and when the long-term resource allocation program for a heterogeneous edge network is executed by a processor, the long-term resource allocation method for a heterogeneous edge network according to any one of claims 1 to 7 is implemented.