A multi-server MEC-D2D system joint task offloading and resource allocation method

Abstract

The application provides a multi-server MEC-D2D system joint task offloading and resource allocation method, belongs to the field of combination of wireless communication and calculation, solves the problems of a large number of computing-intensive task offloading and resource allocation, and constructs an optimization problem with the minimum task execution cost as the target in the case that the task is not divisible. The optimization problem is a mixed integer nonlinear problem, which is decomposed into two problems of resource allocation and task offloading decision to solve. Through derivation, a closed-form expression of the computing resource is obtained, the optimal computing resource is solved, the optimization of the user transmission power is solved by using the three-point method according to the characteristics of the target function, and then the task offloading decision is obtained according to the joint task offloading and resource allocation algorithm, so that the time delay and energy consumption of task execution are reduced. Compared with other benchmark schemes, the joint task offloading and resource allocation algorithm of the multi-server MEC-D2D system considered in the application can effectively reduce the time delay and energy consumption of task execution.

Description

Technical Field

[0001] This invention relates to the fields of wireless communication technology and edge computing, and in particular to a method for joint task offloading and resource allocation in a multi-server MEC-D2D system. Background Technology

[0002] With the development of internet and wireless communication technologies, modern information society has gradually entered the era of the Internet of Things (IoT). A large number of emerging mobile internet services have sprung up, such as AR and autonomous driving. The GSMA predicts that the number of licensed cellular internet connections worldwide will reach 4 billion by 2025. Statistics from the China Internet Network Information Center (CNNIC) show that by the end of 2022, domestic mobile internet access traffic reached 165.6 billion GB. The computational and storage operations required for this massive amount of data undoubtedly place higher demands on end-user capabilities. The implementation of these emerging applications also faces challenges such as limited computing power, resource constraints, and difficulty in meeting latency and energy consumption requirements. To address these issues, existing technologies mainly focus on latency, energy consumption, and optimizing resource allocation.

[0003] Mobile Edge Computing (MEC) is a new computing model that performs computations at the network edge, closer to the source of things or data. The edge layer integrates core capabilities of network, computing, and storage applications, enabling low-latency, low-cost big data and artificial intelligence services and applications, effectively alleviating network congestion and improving service reliability.

[0004] D2D (Device-to-Device) communication technology refers to a communication method in which two peer user nodes communicate directly, enabling direct communication between two mobile users without going through a base station or core network. The main advantages of D2D communication are: 1) Proximity gain: D2D direct communication can further reduce transmission distance, saving base station transmission resources, thereby reducing energy consumption and transmission latency; 2) Multiplexing gain: D2D devices share the same spectrum resources with traditional cellular users, effectively improving the spectral efficiency of system transmission; 3) Single-hop gain: D2D communication has single-hop and multi-hop modes, which is more flexible than the two-hop mode of traditional cellular mode, and its single-hop D2D communication can improve energy efficiency.

[0005] Research shows that combining MEC with D2D communication, through resource allocation and task offloading optimization, can effectively reduce latency, energy consumption, and total task execution costs. Current research on cloud computing and edge computing includes different architectures such as single-server, multi-server, cloud-edge collaboration, and multi-user collaboration. However, traditional cloud computing also faces challenges such as insufficient network bandwidth to support massive data transmission, end-to-end latency failing to meet real-time application requirements, and a high risk of user privacy data leakage. Edge computing can better reduce end-to-end latency and alleviate network congestion, while D2D communication can promote sharing between devices and reduce the traffic load on MEC. Therefore, combining MEC with D2D communication and researching effective task offloading and resource allocation strategies is of great significance. Summary of the Invention

[0006] To address the limitations of traditional cloud computing, such as limited computing power, resource constraints, and difficulty in meeting latency and energy consumption requirements, this paper proposes a joint task offloading and resource allocation method for a multi-server MEC-D2D system. This method considers a multi-server MEC-D2D system, which differs from traditional edge computing systems by adding D2D auxiliary devices to assist users in processing tasks. It includes one macrocell and multiple microcells, multiple users, and auxiliary devices. The optimization objective simultaneously considers the allocation of computing resources to the edge servers, user transmit power, task execution latency, and energy consumption. Furthermore, a joint task offloading and resource allocation algorithm for the multi-server MEC-D2D system is proposed to jointly optimize task offloading and resource allocation, minimizing the total cost of task execution (a weighted sum of energy consumption and latency). Compared with other benchmark schemes (which do not consider task offloading, only optimize power, or only optimize computing resources), the proposed joint task offloading and resource allocation algorithm for the multi-server MEC-D2D system effectively reduces latency and energy consumption, thereby significantly reducing the total system cost.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for joint task offloading and resource allocation in a multi-server MEC-D2D system is proposed. This method is applicable to multi-server MEC-D2D systems, which include multiple base stations (BSs) and multiple users. Multiple auxiliary devices and multiple edge servers Users can choose to execute the task locally, offload it to an auxiliary device for execution, or offload it to an edge server for execution.

[0009] The method includes the following steps:

[0010] Step 1: Macro base station acquires information for each user and auxiliary equipment Ideal channel state information and calculation of communication rate, calculation task information, and auxiliary equipment and edge servers Computational power;

[0011] Step 2: Based on the obtained task and equipment information, the macro base station initializes the user transmit power, the task offloading decision matrix L, V, X, and calculates the time delay and energy consumption weighted sum of the current task execution, ETC. pre User's task preference list for other devices Other devices' user task preference lists

[0012] Step 3: Calculate the optimal computing resources allocated to each user by each edge server, and calculate the current task execution latency and energy consumption weighted sum ETC based on the optimization results;

[0013] Step 4: Determine if ETC > ETC pre Check if the condition is met. If it is met, proceed to step 5; otherwise, proceed to step 9.

[0014] Step 5: Calculate the current optimal user transmit power and update the task offload decision matrix L, V, X and the task preference list.

[0015] Step 6: All users who have not yet made a task uninstallation decision will be categorized according to their task preference list. To their favorite or Send a task request;

[0016] Step 7: For each request received or Based on task preference list Accept user requests within its computing capabilities;

[0017] Step 8: After all users have completed the task unloading decision, calculate the current task unloading decision L, V, X and update the task execution latency and energy consumption weighted sum ETC. pre =ETC, return to step 4;

[0018] Step 9: Output the task unloading decision L, V, X and the optimal task execution latency and energy consumption weighted sum U.

[0019] Specifically, the task unloading decision matrix includes These represent task processing locally, task offloading to an auxiliary device for processing, and task offloading to an edge server for processing, respectively, and each user... Only one method can be selected to handle the task; where l i ∈{0,1}, formula li =1 indicates user Its tasks are computed locally, while l i =0 means Its task is not computed locally; where v ij ∈{0,1}, formula v ij =1 represents Offload the task to the D2D link. Execute the task above, otherwise v ij =0 indicates that the user did not uninstall the task to Execute on, each Only one task can be processed at a time; where x ij ∈{0,1}, formula x ij =1 represents Offload the task to via cellular link Execute the task above, otherwise x ij =0 indicates that the user did not offload the task to the MEC server for execution, and the number of tasks that each MEC server can handle at the same time cannot exceed its maximum computing capacity.

[0020] Specifically, the i-th user To the m-th edge server The achievable data transmission rate is in, for and The transmission signal-to-noise ratio between them, where B is the transmission bandwidth, and σ is the signal-to-noise ratio. 2 The power of additive white Gaussian noise, For users To the edge server The transmission power, for and Channel gain between; i-th user up to the dth auxiliary device The achievable data transmission rate is in, for and The transmission signal-to-noise ratio between them For users Launch to auxiliary equipment The transmission power, for and Channel gain between.

[0021] Specifically, the i-th user The latency and energy consumption for executing tasks locally are as follows: Among them, fi L For the computing power of the i-th user, k L c is the energy consumption factor for locally executed tasks. i The CPU's computing power required to perform the task.

[0022] Specifically, the i-th user Offload the task to the d-th auxiliary device. The latency on is Including the transmission delay T for task unloading i d,trans and the latency T required for task execution i d,exe , For users To auxiliary equipment The achievable data transmission rate is, for The computing power of b i c is the size of the input data when the i-th task is executed. i The CPU computing power required to execute the i-th task; the i-th user Offload the task to the d-th auxiliary device. The energy consumption is Including the energy consumption of task transmission and task execution energy consumption For users Launched to The transmit power, k D For the task in auxiliary equipment The energy consumption coefficient of the above execution, For the d-th auxiliary device Its computing power.

[0023] Specifically, the i-th user Offload the task to the m-th edge server. The latency and energy consumption are respectively Among them, T i m,trans , T represents the latency and energy consumption of the i-th task transmission, respectively. i m,exe , Let k be the latency and energy consumption of the i-th task execution. M f represents the energy consumption factor for executing tasks on edge servers. i m For the m-th edge server Assigned to the i-th user computing power for arrive The achievable data transmission rate is, for arrive Transmission power, b i c is the size of the input data when the i-th task is executed. i The CPU computing power required to execute the i-th task.

[0024] The beneficial effects of this invention are:

[0025] 1. This method aims to minimize the total cost of task execution (weighted sum of energy consumption and latency), decomposes and optimizes the problem, and effectively reduces the total cost of task execution compared to other benchmark solutions.

[0026] 2. The combination of edge computing technology and D2D technology can further improve the utilization of computing resources. When there are multiple mobile users, multiple D2D auxiliary devices and multiple edge servers in the system, the present invention can optimize the total cost and rationally allocate resources by reasonably offloading tasks.

[0027] 3. The multi-server MEC-D2D system joint task offloading and resource allocation algorithm provided by this invention is applicable to multi-server joint optimization of task offloading decision and resource allocation when tasks are indivisible, effectively reducing the total cost of task execution, and can be extended to multi-server task offloading systems where tasks are divisible. Attached Figure Description

[0028] Figure 1 It is a system model diagram;

[0029] Figure 2 This is a flowchart illustrating the joint task unloading and resource allocation process of a multi-server MEC-D2D system;

[0030] Figure 3 This is a comparison chart of the impact of latency and energy consumption influencing factor δ on system cost under different algorithms;

[0031] Figure 4 This is a comparison of the task execution costs of the joint optimization algorithm proposed in this paper with other benchmark algorithms. Detailed Implementation

[0032] To more clearly describe the technical content of the present invention, the present invention will be further described below with reference to the accompanying drawings.

[0033] This invention proposes a joint task offloading and resource allocation method for a multi-server MEC-D2D system. It considers the task offloading and resource allocation problem in a system containing one macrocell and multiple microcells, multiple mobile users, and auxiliary equipment. To meet the latency and energy consumption requirements of computationally intensive tasks, an optimization problem is constructed with the objective of minimizing task execution cost. This optimization problem is a mixed-integer nonlinear problem, which is decomposed into two problems: resource allocation and task offloading decision-making. A closed-form expression for computational resources is derived to obtain the optimal computational resources. For optimizing user transmit power, a ternary search is used based on the characteristics of the objective function. Then, the proposed joint task offloading and resource allocation algorithm for a multi-server MEC-D2D system is used to obtain the task offloading decision, reducing the total task execution cost. Compared with other benchmark schemes, the proposed joint task offloading and resource allocation algorithm for a multi-server MEC-D2D system effectively reduces the total task execution cost.

[0034] Figure 1 The system model diagram illustrates a multi-server MEC-D2D system within a communication scenario comprising one macrocell and multiple microcells. The system includes multiple users, multiple auxiliary devices, and multiple edge servers. When users perform multiple computationally intensive tasks, the total system cost will vary depending on whether the computation is performed locally, offloaded to an edge server, or offloaded to an auxiliary device. Simultaneously, the allocation of computing resources and the optimization of transmit power for the computing devices will also affect the total system cost. Note that the total system cost is a weighted sum of energy consumption and latency for task processing.

[0035] Figure 2 The flowchart of this method mainly consists of the following steps:

[0036] (1) The macro base station initializes the user transmit power p based on the acquired information. i Task unloading decision matrix L, V, X, total task execution cost ETC pre Based on the current task information, calculate and sort each user's list of preferences for other devices. And the sorting of the user's preference list for each edge server and auxiliary device.

[0037] (2) Construct an optimization function with the objective of minimizing the total cost of task execution, and describe the optimization problem as follows:

[0038]

[0039]

[0040]

[0041]

[0042]

[0043]

[0044] Formula 1 is the objective function, where,

[0045]

[0046] The execution cost is defined as either executing the task locally or offloading it to an auxiliary device or edge server. δ∈[0,1] represents the latency and energy consumption impact factor. Constraints (1.a) and (1.b) ensure that the task is indivisible and can only be processed on one device. (1.c) and (1.d) represent the maximum and minimum constraints on user transmit power and edge server computing power. (1.e) represents the constraint on the number of tasks that the edge server can process simultaneously.

[0047] (3) Through derivation, OP1 is decomposed into three subproblems, and the computational resource allocation optimization problem is described as follows:

[0048]

[0049]

[0050] δ∈[0,1] (2.b)

[0051] The function is convex in the defined domain, and its closed expression can be obtained by differentiation:

[0052]

[0053] Where δ∈[0,1] is the time delay energy consumption influence factor, k M f represents the energy consumption factor for executing tasks on edge servers. i m The computing power allocated to the i-th user by the m-th edge server.

[0054] (4) The power optimization problem for each user to offload tasks to auxiliary devices is described as follows:

[0055]

[0056]

[0057] δ∈[0,1] (4.b)

[0058] (5) The power optimization problem of offloading tasks to edge servers for each mobile user is described as follows:

[0059]

[0060]

[0061] δ∈[0,1] (5.b)

[0062] (6) According to the closed-form solution formula 3 of the optimization problem OP2, obtain the optimal computing resource f of the edge server i m , and calculate the current total task execution cost ETC.

[0063] (7) Compare the current total task execution cost ETC with the total task execution cost ETC obtained in the previous iteration pre .

[0064] (8) If ETC > ETC pre does not hold, for OP3 and OP4, the property of the function f is a convex function. Solve the minimum value by the ternary search method. For each iteration, trisect the feasible interval [a, b] to obtain the trisection points l and r. If f(l) < f(r), it means that the minimum value must be obtained within [a, r]. Assign b to a to update the feasible interval, and continuously iterate until the feasible interval reaches the set precision value to obtain the minimum function value. According to formula 4 and formula 5, obtain the optimal transmission power of each mobile user for the auxiliary device and the edge server respectively and update the task offloading decisions L, V, X and the task preference list For each user who has not made a task decision, optimize the task offloading decision according to the task preference list and the computing capabilities of the AD and MEC servers, calculate L, V, X, and set ETC pre = ETC, and return to step (7) for iteration.

[0065] (9) If ETC > ETC pre holds, output the current task offloading decision matrix and the optimal total task execution cost ETC.

[0066] Figure 3 is the system joint optimization algorithm proposed based on this patent. The figure shows the comparison of the influence of the simulation delay - energy consumption impact factor on the total task execution cost of the multi - server MEC - D2D system joint task offloading and resource allocation algorithm proposed in this paper and three other benchmark schemes. It can be seen from the simulation results that by adjusting the delay - energy consumption impact factor δ, the total task execution cost can be affected, so as to balance the relationship between delay and energy consumption.

[0067] Figure 4Based on the system joint optimization method proposed in this patent, the total execution cost of the simulation task calculation is calculated, where δ is taken as 0.5. The results are compared with the schemes of executing all tasks locally, optimizing only power, and optimizing only computing resources. The simulation results show that the algorithm proposed in this paper can effectively reduce the system cost, and power and computing resources also have a certain impact on the system cost.

Claims

1. A method for joint task offloading and resource allocation in a multi-server MEC-D2D system, applicable to a multi-server MEC-D2D system including multiple base stations (BSs) and multiple users. Multiple auxiliary devices and multiple edge servers Users can choose to execute the task locally, offload it to an auxiliary device for execution, or offload it to an edge server for execution. The method includes the following steps: Step 1: Macro base station acquires information for each user and auxiliary equipment Ideal channel state information and calculation of communication rate, calculation task information, and auxiliary equipment and edge servers Computational power; Step 2: Based on the obtained task information and equipment information, the macro base station initializes the user transmit power and the task offloading decision matrix. Calculate the weighted sum of latency and energy consumption for the current task execution. User's task preference list for other devices Other devices' user task preference lists ; Step 3: Calculate the optimal computing resources allocated to each user by each edge server, and calculate the weighted sum of current task execution latency and energy consumption based on the optimization results. The specific steps are as follows: (1) Construct an optimization function with the objective of minimizing the total cost of task execution, and describe the optimization problem as follows: Official 1 (1.a) (1.b) (1.c) (1.d) (1.e) Formula 1 is the objective function, where, The execution cost of executing the task locally or offloading it to an auxiliary device or edge server. For latency and energy consumption, constraints (1.a) and (1.b) ensure that the task is indivisible and can only be processed on one device. (1.c) and (1.d) are the maximum and minimum constraints on user transmit power and edge server computing power. (1.e) is the constraint on the number of tasks that the edge server can process at the same time. (2) Through derivation, OP1 is decomposed into three sub-problems, and the computational resource allocation optimization problem is described as follows: Official 2 (2.a) (2.b) The function is convex in the defined domain, and its closed expression can be obtained by differentiation: Official 3 in, The time delay energy consumption influencing factor, The energy consumption factor for executing tasks on edge servers. The computing power allocated to the i-th user for the m-th edge server; (3) The power optimization problem for each user offloading tasks to auxiliary devices is described as follows: Official 4 (4.a) (4.b) (4) The power optimization problem of offloading tasks to edge servers for each mobile user is described as follows: Official 5 (5.a) (5.b) (5) Based on Formula 3 of the closed-form solution of optimization problem OP2, the optimal computing resources of the edge server are obtained. Calculate the total cost of the current task execution. ; Step 4: Determine Check if the condition is met. If it is met, proceed to step 5; otherwise, proceed to step 9. Step 5: Calculate the current optimal user transmit power and update the task offload decision matrix. and task preference list ; Step 6: All users who have not yet made a task uninstallation decision will be categorized according to their task preference list. To their favorite or Send a task request; Step 7: For each request received or Based on task preference list Accept user requests within its computing capabilities; Step 8: After all users have completed the task uninstallation decision, calculate the current task uninstallation decision. And update the weighted sum of latency and energy consumption for task execution. Return to step 4; Step 9: Output task unloading decision and the optimal weighted sum of task execution latency and energy consumption .

2. The method for joint task offloading and resource allocation in a multi-server MEC-D2D system according to claim 1, characterized in that: The task unloading decision matrix includes , , These represent task processing locally, task offloading to an auxiliary device for processing, and task offloading to an edge server for processing, respectively, and each user... Only one method can be selected to handle the task; among them ,formula Indicates user Its tasks are computed locally, while express Its tasks are not computed locally; among them ,formula represent Offload the task to the D2D link. Execute the task above, otherwise This indicates that the user did not uninstall the task. Execute on, each Only one task can be processed at a time; among which ,formula represent Offload the task to via cellular link Execute the task above, otherwise This indicates that the user did not offload the task to the MEC server for execution, and the number of tasks that each MEC server can handle at the same time cannot exceed its maximum computing capacity.

3. The method for joint task offloading and resource allocation in a multi-server MEC-D2D system according to claim 1, characterized in that: No. individual users To the m-th edge server The achievable data transmission rate is ,in, for and The transmission signal-to-noise ratio between them, where B is the transmission bandwidth. The power of additive white Gaussian noise, For users To the edge server The transmission power, for and Channel gain between; individual users To the Auxiliary equipment The achievable data transmission rate is ,in, for and The transmission signal-to-noise ratio between them For users Launch to auxiliary equipment The transmission power, for and Channel gain between.

4. The method for joint task offloading and resource allocation in a multi-server MEC-D2D system according to claim 1, characterized in that: No. individual users The latency and energy consumption for executing tasks locally are as follows: , ,in, For the first The computing power of each user The energy consumption coefficient for locally executed tasks. The CPU's computing power required to perform the task.

5. The method for joint task offloading and resource allocation in a multi-server MEC-D2D system according to claim 1, characterized in that: No. individual users Unload the task to the Auxiliary equipment The latency on is This includes the transmission latency of task unloading. and the latency required for task execution , For users To auxiliary equipment The achievable data transmission rate is, for computing power For the first The size of the input data when each task is executed. To execute the first The CPU computing power required for the first task; individual users Unload the task to the Auxiliary equipment The energy consumption is This includes the energy consumption of task transmission. and task execution energy consumption , For users Launched to The transmission power, For the task in auxiliary equipment The energy consumption coefficient of the above execution, For the first Auxiliary equipment Its computing power.

6. The method for joint task offloading and resource allocation in a multi-server MEC-D2D system according to claim 1, characterized in that: No. individual users Unload the task to the Edge servers The latency and energy consumption are respectively , ,in, Let be the latency and energy consumption of the i-th task transmission, respectively. For the first The latency and energy consumption of each task execution. The energy consumption factor for executing tasks on edge servers. For the first Edge servers Assigned to the individual users computing power for arrive The achievable data transmission rate is, for arrive Transmission power, For the first The size of the input data when each task is executed. To execute the first The CPU computing power required for each task.

Images