Latency minimization offloading method and system in multi-user industrial internet of things scenario

Abstract

The application discloses a kind of time delay minimization unloading method and system under the scene of multi-user industrial internet of things, method as follows: S1, initialization;S2, establish joint optimization model;S3, modeling is carried out at user end, and the optimal partial unloading strategy of each user itself is solved;S4, modeling is carried out at cloud end, and the optimal computing resource distribution of MEC server is solved by optimal partial unloading strategy;S5, the optimal partial unloading strategy and the optimal computing resource distribution of MEC server are calculated to the minimum time delay of entire system by solving out.S1, the original optimization model is modeled respectively at user end and server end, and the calculation complexity is greatly reduced.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of mobile edge computing network, and particularly relates to a time delay minimization offloading method and system in a multi-user industrial Internet of Things scene. BACKGROUND

[0002] Industrial Internet of Things is an important network for modern industrial enterprises to move towards automation and digitization in Industry 5.0, bringing value-added services to intelligent manufacturing, smart grids, intelligent digital supply chains and other fields. The proliferation of interconnected devices poses a challenge to industrial Internet of Things networks that rely on frequent communication. To overcome these problems, mobile edge computing is proposed and widely used. Mobile edge computing improves the computing capacity of the network edge by deploying high-performance servers closer to users, solving the problem of limited resources of mobile devices and the problem of excessive network load pressure of cloud computing, and can achieve the demand of ultra-low delay and ultra-low energy consumption. However, the MEC-based offloading system still faces some challenges. The MEC server is closer to the user device in location than the cloud server, but the resources of the MEC server are less than the central cloud. Usually, the edge server provides services for nearby users. In addition, due to the mobility of users, the channel conditions in the MEC system will change, especially in the case of multiple users. Therefore, it is crucial to design an efficient computing offloading strategy.

[0003] In the prior art, the computing task is usually set to support binary computing offloading. This means that the computing task can only be calculated locally on the user device or completely offloaded to the MEC server. However, in some specific scenarios, the computing offloading strategy should support partial offloading according to the characteristics of the computing task to make full use of the limited resources of the MEC server. Based on this, the present application cooperates the edge server and the user local two parties to calculate, takes the total computing resources of the system and the offloading strategy as constraints, offloads the task to the local or MEC server, models the purpose of minimizing the total time delay of the system, and then uses a minimax algorithm to solve it, finally calculates the minimum time delay of the entire system. SUMMARY

[0004] In view of the above status of the prior art, the application discloses a latency minimization offloading method and system in a multi-user industrial Internet of Things scene. In a multi-user single MEC server wireless network, the application takes the user offloading decision and the server computing resource as a constraint, models a system total latency minimization target, and proposes an algorithm for joint computing offloading and resource allocation. The algorithm firstly models the user end to solve the optimal partial offloading strategy of each user, then models the server end to solve the optimal computing resource allocation of the MEC server through the optimal partial offloading strategy, and finally solves the minimum latency of the whole system through the optimal partial offloading strategy and the optimal computing resource allocation of the MEC server.

[0005] In order to achieve the purpose of the application, the application adopts the following technical solutions:

[0006] A latency minimization offloading method in a multi-user industrial Internet of Things scene comprises the following steps:

[0007] Step one, initialization stage: each user node obtains basic configuration information of the network through information interaction;

[0008] Step two, optimization model establishment stage: taking the system total latency minimization as a target, a joint optimization model of offloading decision and resource allocation is established according to the task offloading constraint and the user resource allocation constraint;

[0009] Step three, modeling the user end to solve the optimal offloading strategy of each user itself;

[0010] Step four, modeling the server end to solve the optimal computing resource allocation of the MEC server through the optimal partial offloading strategy;

[0011] Step five, solving the minimum latency of the whole system through the optimal partial offloading strategy and the optimal computing resource allocation of the MEC server.

[0012] Further, in step one, each node obtains basic configuration information of the network through information interaction, and the basic configuration information comprises topology information, transmission power, link distance, task data size of user equipment, local computing resource and MEC computing resource, etc.

[0013] Further, in the step two, the optimization model established is:

[0014]

[0015] θ n ∈[0,1](2)

[0016]

[0017]

[0018] where θ represents the set of partial offloading decisions of users, f represents the set of computing resources allocated by MEC servers to users; formula (2) represents the task offloading decision of users; formula (3) represents that the computing resources allocated by MEC servers are positive and should not exceed the total computing resources of MEC servers; and formula (4) represents that the computing resources allocated by MEC servers to all offloaded tasks should not exceed the total computing resources of MEC servers.

[0019] Further, in step three, first modeling is performed at the user end:

[0020]

[0021] s.t(2)

[0022] where formula (2) represents the task offloading decision of users. When and only when , the optimal offloading strategy minimizing the delay can be solved.

[0023] Proof: first, when , the following can be solved:

[0024]

[0025] When the local computing delay is large, the derivative of the objective function is:

[0026]

[0027] It can be found that the objective function T n monotonically decreases when 0≤θ n ≤θ mec monotonically increases when the offloading execution delay is large. Therefore, it can be proved that when

[0028]

[0029] It can be found that the objective function T n monotonically increases when θ mec ≤θ n ≤1. Therefore, it can be proved that when , there is a minimum value. Therefore, the optimal offloading decision of each user n can be solved as:

[0030]

[0031] Further, in step four, modeling is performed at the server end, and the optimal computing resource allocation of MEC servers is solved through the solved optimal partial offloading strategy, so that the minimized total delay of the entire system is:

[0032]

[0033] s.t(3)(4)

[0034] iff T1=T2=T3=…=T n iff T1=T2=T3=…=T

[0035] Further, in the fifth step, the minimum time delay of the whole system is solved by the optimal partial offloading strategy and the optimal computing resource allocation of the MEC server.

[0036] The application further discloses a multi-user industrial Internet of Things scene-based time delay minimization offloading system for executing the method.

[0037] The initialization module: each user node obtains basic configuration information of the network through information interaction;

[0038] The joint optimization model establishment module: a joint optimization model of offloading decision and resource allocation is established according to task offloading constraints and user resource allocation constraints, with the target of minimizing the total time delay of the system;

[0039] The user end modeling module: the user end is modeled, and the optimal offloading decision of each user is solved;

[0040] The server end modeling module: the server end is modeled, and the optimal computing resource allocation of the MEC server is solved through the optimal offloading strategy;

[0041] The minimum time delay solving module: the minimum time delay is solved through the optimal offloading strategy and the optimal computing resource allocation of the MEC server.

[0042] The application has the following advantages:

[0043] (1) The application realizes a multi-user industrial Internet of Things scene-based time delay minimization offloading method, and experiments show that the method proposed in the application has smaller total time delay of the system than existing algorithms.

[0044] (2) The most prominent advantage of the application is that the original optimization model is modeled at the user end and the server end respectively, an optimal offloading algorithm is proposed to solve the offloading problem at the user end and the computing resource allocation problem at the server end, and the calculation complexity is greatly reduced. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 FIG. 1 is a flowchart of a multi-user industrial Internet of Things scene-based time delay minimization offloading method according to a preferred embodiment of the application.

[0046] Figure 2 This is a system model diagram for a multi-user single MEC server.

[0047] Figure 3 This is a graph showing how the number of users affects the total system latency under different methods.

[0048] Figure 4 This graph shows how the computing power of the MEC server affects the total system latency under different methods.

[0049] Figure 5 This is a block diagram of a latency-minimizing offloading system in a multi-user industrial IoT scenario, according to a preferred embodiment of the present invention. Detailed Implementation

[0050] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0051] like Figures 1-2 As shown in this embodiment, a latency minimization offloading method in a multi-user industrial IoT scenario includes the following steps:

[0052] Step 1, Initialization Phase: Each node obtains basic network configuration information through information exchange. The basic configuration information includes topology information, transmit power, link distance, user equipment task data size, local computing resources, and MEC computing resources.

[0053] This embodiment considers a multi-user industrial IoT mobile edge computing system. The system consists of a base station, multiple user devices (user devices), and a MEC server. The MEC server is deployed alongside the base station. The set of user devices is represented by N = {1, 2, ..., N}. User devices can sense the production environment in an Industry 5.0 scenario, and each user device has an AI-based computationally intensive task requiring computation. However, due to resource limitations, user devices may not be able to meet the computational demands of such computationally intensive tasks. User devices can migrate tasks to the MEC server, thereby reducing computational latency. Compared to traditional cloud computing, the MEC server is closer to the users, which can reduce communication costs between user devices and the MEC server. However, the MEC server has relatively low computing power and storage capacity, and may not be able to simultaneously handle all tasks. It is assumed that each user device has the same task priority. When tasks are offloaded simultaneously, wireless bandwidth resources are evenly allocated to each user device to upload their tasks. The upload data rate of the nth user device is:

[0054]

[0055] Where W is the bandwidth of the wireless channel, K is the number of user devices being offloaded, and p nis the data transmission power of the nth user equipment uploaded, h n is the radio channel gain between the nth user equipment and the base station, and N0 is the noise power spectral density.

[0056] Let R n = (D n , C n ) be the task that the nth user equipment needs to perform. This task can be calculated locally or calculated on the MEC server by offloading calculation. D n represents the amount of data required to complete this task, and C n represents the number of CPU cycles required to complete the task R n . The user equipment task can be offloaded at any ratio. This is a substantial difference between the model of the present application and other existing binary calculation offloading schemes.

[0057] The present application uses θ n ∈ [0, 1] as the calculation offloading strategy of the nth user equipment, that is, the nth user equipment offloads θ n part of the task to the MEC server, and (1-θ n ) of the task is calculated locally. The calculation offloading strategy θ of the user equipment set N is composed of all user equipments, denoted as θ = [θ1, θ2,..., θ n ]. When θ n = 0, it means that the nth task R n is calculated locally. When θ n = 1, it means that R n is completely offloaded to the MEC server. When 0 < θ n < 1, it means that θ n part of the task R n is offloaded to the MEC server, and (1-θ n ) part of the task is calculated locally.

[0058] Step two, the optimization model stage: taking the minimization of the total system delay as the goal, according to the task offloading constraints and user resource allocation constraints, a joint optimization model of offloading decision and resource allocation is established.

[0059] When the user equipment performs the task locally, the task completion process is irrelevant to the MEC server, and the local calculation resource is completely used. The delay T n of the task R n in local calculation can be represented as: l

[0060]

[0061] where f n lrepresents the computing capability of the nth user equipment (the number of CPU cycles per second of the nth user equipment).

[0062] When the user equipment requests to offload computing, the task R n θ n part is offloaded to the MEC server. The total delay of the task R n θ n part includes data upload delay, MEC computing delay and data download delay. The (1-θ n ) part of the task R n is local computing. Therefore, the upload delay when the θ n part of the task R n is offloaded to the MEC server can be represented as:

[0063]

[0064] The computing delay when the θ n part of the task R n is offloaded to the MEC server can be represented as:

[0065]

[0066] wherein f n represents the computing resource allocated by the MEC to the task R n . F is defined as the computing capability of the MEC server (the number of CPU cycles per second of the MEC server). The computing resource allocated by the MEC server to all offloaded tasks should not exceed the total computing resource, which means that the condition

[0067] The data download delay when the θ n part of the task R n is offloaded to the MEC server can be represented as:

[0068]

[0069] wherein B n represents the data size of the execution result returned by the MEC server, and r' n represents the download rate of the nth user equipment from the MEC server to download the execution result. Since the size D n of the uploaded data is much larger than B n , the download rate is much larger than the upload rate. Therefore, the present application does not consider the data download delay problem.

[0070] For the (1-θ n ) part of the task R nThe ) part is computed locally, and this process does not require consideration of the MEC server. Therefore, R can be obtained. n (1-θ) n The computational delay of part ) will be reduced in task R. n (1-θ) n ) computation delay It can be represented as:

[0071]

[0072] Due to task R n Due to the parallelism, the total time delay for the nth user device to execute the partial computation of the offloading strategy should be that of task R. n θ n Partial delays and task R n (1-θ) n The larger of the delays in the () part. According to formulas (3)(4)(5)(6), define T. n When performing partial computation of the offload strategy for the nth user equipment, task R... n Total delay, T n The calculation method is as follows:

[0073]

[0074] In equation (7), when θ n When = 0, we can obtain When θ n When = 1, we can obtain Therefore, equation (7) can be used to comprehensively consider the time delays of three strategies: local computation, partial computation offloading, and complete computation offloading. In this invention, the user equipment side is considered first, aiming to minimize the latency of each user equipment. Secondly, this invention minimizes the task execution time of all user equipment in the entire industrial IoT mobile edge system as its optimization objective. Based on the above network model, computation task model, local computation model, and edge computation model, the optimization problem can be expressed as:

[0075]

[0076] θ n ∈[0,1](9)

[0077]

[0078] Where θ represents the set of partial unloading decisions of the user, and f represents the set of computing resources allocated to the user by the MEC server; Equation (2) represents the user's task unloading decision; Equation (3) indicates that the computing resources allocated by the MEC server are positive and should not exceed the total computing resources of the MEC server; Equation (4) indicates that the computing resources allocated by the MEC server to all unloading tasks should not exceed the total computing resources of the MEC server.

[0079] The original optimization model will be modeled on both the user end and the cloud.

[0080] Step 3: First, model the user side, specifically as follows:

[0081]

[0082] st(9)

[0083] Equation (9) represents the user's task unloading decision. The decision is made if and only if... When the time is minimized, the optimal unloading strategy that minimizes latency can be obtained.

[0084] Proof: First, when When, we can solve for:

[0085]

[0086] When the local computation latency is large, the derivative of the objective function is:

[0087]

[0088] It can be found that the objective function T n In 0≤θ n ≤θ mec The time-varying frequency decreases monotonically. When the unloading execution delay is large, the derivative of the objective function is:

[0089]

[0090] It can be found that the objective function T n In θ mec ≤θ n It monotonically increases when ≤1. Therefore, it can be proven that when When the minimum value is reached, the optimal uninstallation decision for each user n can be determined as follows:

[0091]

[0092] Step 4: Model the cloud environment and use the optimal partial offloading strategy θ to solve for the optimal allocation of computing resources to the MEC server. The minimum total latency for the entire system is then:

[0093]

[0094] s.t(10)(11)

[0095] iff T1=T2=T3=…=T n The solution of (17) can be obtained. The proof is the same as above.

[0096] Step five, the minimum time delay of the whole system is solved by substituting the optimal partial offloading strategy θ and the optimal computing resource allocation f of the MEC server into (8).

[0097] In Figure 3 , the influence of the number of user devices on the total delay of the system is studied. The abscissa represents the number of user devices, and the ordinate represents the total delay of the completion of the tasks of all user devices. It can be seen from the figure that the total delay of full offloading increases with the increase of the number of user devices. This is because as the number of user devices increases, the computing resources allocated by the server to each device will decrease, resulting in an increase in the time delay of each user device, and the total time delay of the whole system will increase. The total delay of partial offloading of the present application will also increase slowly, and when the number of user devices is sufficient, the total delay of partial offloading will tend to the total delay of local computing, because when the number of users is sufficient, the computing resources allocated by the server will be particularly small, and it can be considered that the computing tasks of the users are all locally computed, so it will tend to the total delay of local computing.

[0098] In Figure 4 , the influence of the computing capacity of the MEC server on the total delay of the system is studied. The abscissa represents the computing capacity of the MEC server, and the ordinate represents the total delay of the completion of the tasks of all user devices. It can be seen from the figure that the total delay of full offloading decreases with the increase of the computing capacity of the MEC server. This is because as the computing capacity of the MEC server increases, the computing resources allocated by the server to each device will be more, resulting in a decrease in the time delay of each user device, and the total time delay of the whole system will decrease. At the same time, it can be found that the total delay of partial offloading is always less than the total delay of full offloading, and when the computing capacity of the MEC server is infinite, the curves will infinitely approach but never intersect. This is because the partial offloading of the present application is to offload locally and simultaneously offload to the server, while full offloading is to execute the task only by the server. When the computing capacity of the MEC server is infinite, most of the tasks in the partial offloading will be completed by the server, and only a few tasks will be calculated locally, so it will infinitely approach but never intersect.

[0099] As Figure 5 shown, the embodiment discloses a time delay minimization offloading system in a multi-user industrial Internet of Things scene, which is used for executing the above method, and comprises the following modules:

[0100] Initialization module: each user node obtains the basic configuration information of the network through information interaction;

[0101] Joint optimization model establishment module: a joint optimization model of unloading decision and resource allocation is established according to the task unloading constraint and the user resource allocation constraint, with the minimum system total delay as the target;

[0102] User end modeling module: the user end is modeled, and the optimal unloading decision of each user is solved;

[0103] Server end modeling module: the server end is modeled, and the optimal computing resource allocation of the MEC server is solved through the optimal unloading strategy;

[0104] Minimum delay solving module: the minimum delay is solved through the optimal unloading strategy and the optimal computing resource allocation of the MEC server.

[0105] Other contents of the embodiment can refer to the above method embodiment.

[0106] The preferred embodiments and principles of the application are described in detail above, and for those skilled in the art, the specific implementation manner can be changed according to the idea provided by the application, and these changes should be regarded as the protection scope of the application.

Claims

1. A latency-minimizing offloading method for multi-user industrial IoT scenarios, characterized in that, Specifically, the following steps are included: Step 1: Each user node obtains basic network configuration information through information exchange; Step 2: With the goal of minimizing the total system latency, establish a joint optimization model for offloading decisions and resource allocation based on task offloading constraints and user resource allocation constraints; Step 3: Model the user side and solve for the optimal uninstallation decision for each user; Step 4: Model the server side and solve for the optimal allocation of computing resources for the MEC server through the optimal offloading strategy; Step 5: Solve for the minimum latency using the optimal offloading strategy and the optimal allocation of computing resources for the MEC server; In step two, the joint optimization model takes the following form: Where F is defined as the computing power of the MEC server, and f represents the set of computing resources allocated by the MEC server to users; f n This indicates that task R is assigned by the MEC server. n The computing resources; Equation (2) represents the user's task unloading decision; Equation (3) represents that the computing resources allocated to the MEC server are positive and do not exceed the total computing resources of the MEC server; Equation (4) represents that the computing resources allocated by the MEC server to all unloading tasks do not exceed the total computing resources of the MEC server; In step three, the user end is modeled, and the minimum latency for each user n is: when When the time is minimized, the optimal unloading strategy is obtained. This represents the computation offloading strategy for the user equipment set N; Indicates task R n of( The computational delay of this part; Indicates task R n of Upload latency when partially unloaded to the MEC server; Indicates task R n of Computational latency when partially offloaded to the MEC server; The computation offloading strategy for the nth user device; Step four: Model the server side and solve for the optimal allocation of computing resources for the MEC server using the optimal partial offloading strategy. The minimum total latency of the entire system is: when When, the solution to equation (10) is obtained, where T n When performing partial calculation of the offload strategy for the nth user equipment, task R... n Total delay.

2. The latency minimization offloading method in a multi-user industrial IoT scenario according to claim 1, characterized in that, In step one, the basic configuration information includes topology information, transmit power, link distance, user equipment task data size, local computing resources, and MEC server computing resources.

3. A latency-minimizing offloading system for multi-user industrial IoT scenarios, used to execute the method as described in any one of claims 1-2, characterized in that, The system includes the following modules: Initialization module: Each user node obtains basic network configuration information through information exchange; Joint optimization model building module: With the goal of minimizing the total system latency, a joint optimization model for offloading decisions and resource allocation is established based on task offloading constraints and user resource allocation constraints; User-side modeling module: Models the user-side application and solves the optimal uninstallation decision for each user; Server-side modeling module: Models the server side and solves the optimal allocation of computing resources for the MEC server through the optimal offloading strategy; Minimum latency solution module: Solve for the minimum latency by using the optimal offloading strategy and the optimal allocation of computing resources for the MEC server.

Images