Divisible task offloading decision-making method based on evolutionary game theory

Abstract

The present invention discloses a divisible task offloading decision method based on evolutionary game theory, comprising: 1) constructing divisible tasks for mobile terminal users and dividing the divisible tasks into multiple identical subtasks; 2) constructing offloading decisions about the subtasks and establishing a cost function for the user; 3) constructing an evolutionary game model based on the cost function and initializing the user's policy state; 4) using a dynamic replicator method to iteratively solve the policy state until the state no longer changes, thereby obtaining the final offloading strategy for each user. The present invention makes offloading decisions for divisible tasks based on evolutionary game theory, and can reduce the cost of completing tasks under the constraints of limited time, limited resources, user task attributes, and different geographical locations, while improving resource utilization, further realizing offloading decision applications under dense base station deployment.

Description

Technical Field

[0001] The present invention relates to the technical field of game theory and computation offloading decision-making, and in particular to a divisible task offloading decision-making method based on evolutionary game theory. Background Art

[0002] In recent years, with the continuous development of emerging technologies such as Mobile Edge Computing (MEC), a growing number of computation offloading applications have emerged. Computation offloading involves transferring computing tasks from terminal devices to edge servers, which then perform the tasks and return the results to the device. By leveraging the idle resources of edge servers deployed at base stations, researchers can design and implement a variety of everyday application scenarios, including facial recognition on mobile phones, complex computations for augmented reality devices, and navigation and road condition awareness for in-vehicle terminals. However, offloading computation tasks to terminal devices relies on abundant computing resources on edge servers and advanced communications technology. According to a white paper, base station coverage has reached 10 per kilometer, meaning multiple base stations can offload user tasks. Furthermore, advancements in communications technology are gradually reducing interference caused by channel sharing and improving the utilization of shared channels. Only when these two elements are effectively combined can computation offloading applications operate effectively and bring convenience to people's lives.

[0003] To ensure high-quality service for these applications, a key factor lies in determining which base station, among numerous candidate base stations, to offload tasks to. A base station's computing resources are limited in the amount of tasks they can handle. When the total amount of tasks offloaded to a base station exceeds this capacity, task completion takes longer, reducing user satisfaction. For example, if users are unaware of each other's offloading decisions, it's possible that some base stations' total workloads far exceed their capacity, while others might not. This significantly reduces resource utilization. Furthermore, when users communicate their decisions to all other users, this increases transmission pressure on communication links and wastes bandwidth. Therefore, it's crucial to find the right balance between maximizing the utilization of all base station computing resources and reducing the transmission of irrelevant information on the links. Ultimately, this ensures that edge servers can provide users with high-quality computing offload services.

[0004] In recent research, many researchers have done extensive work, designing various game models to coordinate offloading decisions among users and maximize the utilization of base station computing resources. Among these traditional methods, most only consider user offloading decisions within a single base station and propose a variety of game models based on this. Designing a utility function based on potential games is a convenient way to coordinate the decisions of all users. However, it assumes a global control center to collect and arrange the decision choices of all users. In this case, if traditional mechanisms are adopted and an attempt is made to designate a base station as its global control center, the communication cost between the base station and the control center will obviously increase, which is clearly not suitable for multi-user computing offloading in multi-base station scenarios.

[0005] Considering the overlapping coverage of base stations and the variability in users' offloading decisions—for example, different users may offload to different sets of base stations—traditional mechanisms can fall into local optima in decision coordination, resulting in underutilization of computing resources at some base stations. In these new circumstances, traditional approaches are no longer effective, leading to the design of a new decision-making mechanism to address this situation. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings and deficiencies of the existing technology and provide a divisible task offloading decision method based on evolutionary game theory, which can reduce the cost of task completion under the constraints of limited time, limited resources, user task attributes and different geographical locations, while improving resource utilization, and further realize the offloading decision application under dense base station deployment.

[0007] To achieve the above objectives, the present invention provides a technical solution: a divisible task offloading decision-making method based on evolutionary game theory, comprising the following steps:

[0008] 1) Construct divisible tasks for mobile terminal users and divide the divisible tasks into multiple identical subtasks;

[0009] 2) Construct offloading decisions about subtasks and establish a cost function for users;

[0010] 3) Based on the cost function, an evolutionary game model is constructed to initialize the user's strategy state;

[0011] 4) Using the dynamic replicator method, the policy state is iteratively solved until the state no longer changes, and the final uninstallation policy for each user is obtained.

[0012] In step 1), consider the user set N, for user i, where i=1,2,...,N, it has splittable tasks T i , the CPU cycles required to complete the entire task are Z i; Define subtask T0, the CPU cycle required to complete a single subtask is Z0, where Z i ≥Z0; then for the divisible task T i , the number of subtasks a user has is expressed as:

[0013] i=1,2,...,N

[0014] Where M i is the number of subtasks owned by user i, Indicates rounding up.

[0015] In step 2), considering the base station set K, according to the coverage overlap range of the base stations, user i obtains the candidate base station set K according to the location attribute i , user i will m ik Subtasks are unloaded to base station k, which accounts for the proportion of the total number of subtasks s ik for:

[0016] i=1,2,...,N,k=1,2,...,K

[0017] Among them, s ik ∈[0,1], when s ik =0;M i is the number of subtasks owned by user i, N is the set of users; user i’s uninstallation strategy The uninstallation policy of all users combined s = {s i ,i∈N}.

[0018] The total amount of resources owned by base station k is C k , combined with the offloading strategies of all users, we can get the total resource request L of base station k k (s) is:

[0019] k=1,2,...,K

[0020] Among them, L k is calculated based on the offloading policy of all users, so it is a variable related to s; Z0 is the CPU cycle required to complete a single subtask; and the load rate of base station k is l k (s) is expressed as:

[0021] k=1,2,...,K

[0022] According to the load rate returned by the base station, establish user i to m ik The cost function u of offloading the subtask to base station k ik (s) is:

[0023] u ik (s)=s ik M i Z0l k (s).

[0024] In step 3), an evolutionary game model is constructed based on the cost function. Among the three elements of the game model, the player set, the strategy set, and the utility function correspond to the mobile terminal user set N, the strategy set Λ, and the cost function. i and the cost function U i , where Λ i Indicates s i The value space of U i represents a set of cost functions for user i; in the evolutionary game, all subtasks owned by the user form a population, the user specifies an offloading base station for all subtasks, and the number of subtasks offloaded to each base station is counted to obtain the user's offloading decision s i , which is called a strategic state in the game below. The user's initial strategic state is:

[0025]

[0026] In the formula, ||·|| represents the number of elements in the set; s ik For user i, m ik The proportion of subtasks unloaded to base station k to the total number of subtasks; i Get a set of candidate base stations for user i based on his location attributes; define the average cost function for user i The following formula is used to measure the quality of the strategy status:

[0027]

[0028] Among them, K is the base station set, u ik (s) is user i's m ik The cost function for offloading the subtask to base station k.

[0029] In step 4), the dynamic replicator method is used to define the dynamic equation of the policy state according to the cost function and the average cost:

[0030]

[0031] Where β represents the rate of change factor, is the average cost function of user i, u ik (s) is user i's m ik The cost function for offloading subtasks to base station k is: Indicates s ik The changing trend of sik For user i, m ik The proportion of subtasks unloaded to base station k accounts for the total number of subtasks; under the current s, s is the unloading strategy of all users combined, when s ik <0, that is The cost of offloading to base station k is higher than the average cost. The user can reduce u by reducing the number of subtasks offloaded to this base station. ik (s); On the contrary, when s ik When it is greater than 0, the number of subtasks will be increased, ultimately achieving the goal of reducing the average cost; The value of determines the ratio of the number of subtasks to be increased or decreased. Therefore, the adjusted policy state is expressed as:

[0032]

[0033] Introduce t to mark the value of each variable at different times, T represents the preset maximum time; then at time t, s ik (t) is the policy state of user i, is the dynamic equation of user i, the strategy state s at the next moment ik (t+1) is expressed as:

[0034] t=1,2,...,T

[0035] After multiple calculations, when the user's policy status no longer changes, Use t e Mark the current moment, i.e. K i Get the candidate base station set for user i according to the location attribute, At time t e The dynamic equation, at this time s ik (t e +1)=s ik (t e ), s ik (t e ) and s ik (t e +1) indicates that at time t e and t e +1 policy state, at this time the user's policy state no longer changes, then the policy state is the uninstall decision finally solved by the user.

[0036] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0037] 1. This invention uses evolutionary game theory for the first time to make offloading decisions for divisible tasks, breaking through the problem of difficulty in solving problems caused by the huge decision space.

[0038] 2. Compared with other computation offloading decision methods, the present invention improves the utilization rate of base station resources and reduces task completion time.

[0039] 3. The method of the present invention has a wide range of applications in computing offloading decision-making tasks, is simple to operate, has strong adaptability, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the logic flow of the method of the present invention.

[0041] Figure 2 Schematic diagram of subtask examples used in the present invention.

[0042] Figure 3 This is the decision diagram used in the present invention. DETAILED DESCRIPTION

[0043] The present invention will be described in further detail below with reference to the embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0044] like Figures 1 to 3 As shown, the divisible task offloading decision method based on evolutionary game theory provided in this embodiment uses a dynamic replicator method, which includes the following steps:

[0045] 1) Construct divisible tasks for mobile terminal users and divide the divisible tasks into multiple identical subtasks; Consider the user set N, for user i, where i=1,2,...,N, it has divisible tasks T i , the CPU cycles required to complete the entire task are Z i ; Define subtask T0, the CPU cycle required to complete a single subtask is Z0, where Z i ≥Z0; then for the divisible task T i , the number of subtasks a user has is expressed as:

[0046] i=1,2,...,N

[0047] Where M i is the number of subtasks owned by user i, Indicates rounding up. Figure 2 As shown, M=10 at this time.

[0048] 2) Construct the offloading decision about the subtask and establish the user's cost function; Consider the base station set K, according to the coverage overlap range of the base stations, user i obtains the candidate base station set K according to the location attribute i , user i will m ikThe proportion of subtasks unloaded to base station k is:

[0049] i=1,2,...,N,k=1,2,...,K

[0050] Among them, s ik ∈[0,1], when s ik =0; user i’s uninstall policy is The uninstallation policy of all users is s={s i ,i∈N}. Figure 3 As shown, for users in the area, their base station set K = {1, 2, 3}, where Figure 2 As shown in Figure 2, the number of subtasks that user i offloads to the base station are 3, 5, and 2 respectively, the ratio is s1 = 0.3, s2 = 0.5, s3 = 0.2, and the offloading strategy is s = {0.3, 0.5, 0.2}.

[0051] The total amount of resources owned by base station k is C k , the value range is 5Mhz~10Mhz. Combined with the offloading strategies of all users, the total resource request of base station k is:

[0052] k=1,2,...,K

[0053] Among them, L k It is calculated based on the offloading policy of all users, so it is a variable related to s; and the load rate of base station k is expressed as:

[0054] k=1,2,...,K

[0055] According to the load rate returned by the base station, establish user i to m ik The cost function u of offloading the subtask to base station k ik (s) is:

[0056] u ik (s)=s ik M i Z0l k (s)

[0057] 3) Based on the cost function, an evolutionary game model is constructed to initialize the user's strategy state; in the three elements of the game model, the player set, strategy set, and utility function correspond to the mobile terminal user set N, the strategy set Λ, and the user's strategy state. i and the cost function U i , where Λ i Indicates s i The value space of U irepresents a set of cost functions for user i; in the evolutionary game, all subtasks owned by the user form a population, the user specifies an offloading base station for all subtasks, and the number of subtasks offloaded to each base station is counted to obtain the user's offloading decision s i , which is called a strategic state in the game below. The user's initial strategic state is:

[0058]

[0059] In the formula, |·| represents the number of elements in the set; the average cost function of user i is defined as follows, which is used to measure the quality of the strategy state and is expressed as:

[0060]

[0061] 4) Using the dynamic replicator method, the policy state is iteratively solved until the state no longer changes, and the final uninstallation policy for each user is obtained. Based on the cost function and the average cost, the dynamic equation defining the policy state is:

[0062]

[0063] In the formula, β represents the rate of change factor, which is set to 1. Indicates s ik The changing trend of ; Under the current s, when s ik <0, that is The cost of offloading to base station k is higher than the average cost. The user can reduce u by reducing the number of subtasks offloaded to this base station. ik (s); On the contrary, when s ik When it is greater than 0, the number of subtasks will be increased, ultimately achieving the goal of reducing the average cost; The value of determines the ratio of the number of subtasks to be increased or decreased. Therefore, the adjusted policy state is expressed as:

[0064]

[0065] Introduce t to mark the value of each variable at different times, T represents the preset maximum time; then at time t, s ik (t) is the policy state of user i, is the dynamic equation of user i, the strategy state s at the next moment ik (t+1) is expressed as:

[0066] t=1,2,...,T

[0067] After multiple calculations, when the user's policy status no longer changes, Use t e Mark the current moment, i.e. K i Get the candidate base station set for user i according to the location attribute, At time t e The dynamic equation, at this time s ik (t e +1)=s ik (t e ), s ik (t e ) and s ik (t e +1) indicates that at time t e and t e +1 policy state, at this time the user's policy state no longer changes, then the policy state is the uninstall decision finally solved by the user.

[0068] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A splittable task offloading decision-making method based on evolutionary game theory, characterized by: The following steps are involved: 1) Construct divisible tasks for mobile terminal users and divide the divisible tasks into multiple identical subtasks; 2) Construct the offloading decision about the subtask and establish the user's cost function as follows: Considering the base station set K, according to the coverage overlap range of the base stations, user i obtains the candidate base station set K according to the location attribute i , user i will m ik Subtasks are unloaded to base station k, which accounts for the proportion of the total number of subtasks s ik for: Among them, s ik ∈[0,1], when s ik =0;M i is the number of subtasks owned by user i, N is the set of users; user i’s uninstallation strategy The uninstallation policy of all users combined s = {s i ,i∈N}; The total amount of resources owned by base station k is C k , combined with the offloading strategies of all users, we can get the total resource request L of base station k k (s) is: Among them, L k is calculated based on the offloading policy of all users, so it is a variable related to s; Z0 is the CPU cycle required to complete a single subtask; and the load rate of base station k is l k (s) is expressed as: According to the load rate returned by the base station, establish user i to m ik The cost function u of offloading the subtask to base station k ik (s) is: u ik (s)=s ik M i Z0l k (s) 3) Based on the cost function, an evolutionary game model is constructed to initialize the user's strategy state; 4) Using the dynamic replicator method, the policy state is iteratively solved until the state no longer changes, and the final uninstallation policy for each user is obtained, as follows: Using the dynamic replicator method, the dynamic equation defining the policy state is: Where β represents the rate of change factor, is the average cost function of user i, u ik (s) is user i's m ik The cost function for offloading subtasks to base station k is: Indicates s ik The changing trend of s ik For user i, m ik The proportion of subtasks unloaded to base station k accounts for the total number of subtasks; under the current s, s is the unloading strategy of all users combined, when s ik <0, that is The cost of offloading to base station k is higher than the average cost. The user can reduce u by reducing the number of subtasks offloaded to this base station. ik (s); On the contrary, when s ik When it is greater than 0, the number of subtasks will be increased, ultimately achieving the goal of reducing the average cost; The value of determines the ratio of the number of subtasks to be increased or decreased. Therefore, the adjusted policy state is expressed as: Introduce t to mark the value of each variable at different times, T represents the preset maximum time; then at time t, s ik (t) is the policy state of user i, is the dynamic equation of user i, the strategy state s at the next moment ik (t+1) is expressed as: After multiple calculations, when the user's policy status no longer changes, Use t e Mark the current moment, i.e. K i Get the candidate base station set for user i according to the location attribute, At time t e The dynamic equation, at this time s ik (t e +1)=s ik (t e ), s ik (t e ) and s ik (t e +1) indicates that at time t e and t e +1 policy state, at this time the user's policy state no longer changes, then the policy state is the uninstall decision finally solved by the user.

2. The divisible task offloading decision-making method based on evolutionary game theory according to claim 1 is characterized in that: In step 1), consider the user set N, for user i, where i = 1, 2, ..., N, it has splittable tasks T i , the CPU cycles required to complete the entire task are Z i ; Define subtask T0, the CPU cycle required to complete a single subtask is Z0, where Z i ≥Z0; then for the divisible task T i , the number of subtasks a user has is expressed as: Where M i is the number of subtasks owned by user i, Indicates rounding up.

3. The divisible task offloading decision-making method based on evolutionary game theory according to claim 1 is characterized in that: In step 3), an evolutionary game model is constructed based on the cost function. Among the three elements of the game model, the player set, the strategy set, and the utility function correspond to the mobile terminal user set N, the strategy set Λ, and the cost function. i and the cost function U i , where Λ i Indicates s i The value space of U i represents a set of cost functions for user i; in the evolutionary game, all subtasks owned by the user form a population, the user specifies an offloading base station for all subtasks, and the number of subtasks offloaded to each base station is counted to obtain the user's offloading decision s i , which is called a strategic state in the game below. The user's initial strategic state is: In the formula, |·| represents the number of elements in the set; s ik For user i, m ik The proportion of subtasks unloaded to base station k to the total number of subtasks; i Get a set of candidate base stations for user i based on his location attributes; define the average cost function for user i The following formula is used to measure the quality of the strategy status: Among them, K is the base station set, u ik (s) is user i's m ik The cost function for offloading the subtask to base station k.

Images