Mobile edge computing resource allocation method based on limited codebook

Abstract

The invention discloses a mobile edge computing resource allocation method based on a limited codebook. The method comprises the following steps: constructing an edge computing part unloading model oriented to a time delay sensitive task; according to input and output mutual information estimation, a reachable rate and a reachable rate approximate value based on a limited codebook are obtained; based on the reachable rate, the transmission delay and the calculation delay of the task under the edge calculation part unloading model are calculated, and the local calculation delay of the task equipment is combined to obtain the task execution delay; defining an optimization target as minimizing system time delay, and introducing a penalty term aiming at a task timeout condition; based on the approximate value of the reachable rate, jointly optimizing a task unloading proportion and edge cloud computing resources to complete resource allocation; compared with a traditional optimization method based on a capacity formula, the method can further improve the time delay performance of the system.

Description

Technical Field

[0001] The present invention belongs to the technical field of mobile edge computing, and in particular relates to a mobile edge computing resource allocation method based on a finite codebook. Background Art

[0002] In the current mobile edge computing scenarios, the calculation of latency and the optimization of computing resources are mainly analyzed based on the capacity formula. However, the capacity formula can only reflect the upper limit of the unloading rate when the input signal obeys the Gaussian distribution. In practical applications, due to the complexity and hardware limitations of the task equipment, the communication rate between task equipment is often affected by the specific modulation method (such as phase shift keying or orthogonal amplitude modulation). Therefore, the actual system usually uses the achievable rate based on a finite codebook for analysis. It is worth noting that as the transmission power increases, the gap between the achievable rate based on the finite codebook and the capacity will become larger and larger. Therefore, the existing delay calculation based on the capacity formula is difficult to accurately characterize the delay performance of the actual system, and the corresponding resource optimization scheme is also difficult to further improve the system delay performance.

[0003] With the surge in the use of mobile mission equipment, the continuous increase in mission data volume and the rapid development of communication technology, the application scope of mobile edge computing is rapidly expanding. Compared with the traditional cloud computing model, the edge cloud provides computing resources at the edge of the network closer to the mobile mission equipment. This configuration effectively reduces the data transmission delay and shortens the system response time, thereby providing more effective support for delay-sensitive applications. Therefore, MEC has been widely used in the fields of Internet of Things (IoT), Internet of Vehicles (IoV), smart networks, telemedicine, etc.

[0004] In the latency calculation and optimization process of existing MEC scenarios, capacity is mainly used as the offloading rate. Specifically, [5] jointly considers partial offloading and resource allocation, and uses data segmentation adjustment iteration to reduce parallel processing delay, thereby solving the problem of minimizing the total delay in the MEC system. In, the selection of offloading mode is expressed as an exact potential game, and a joint optimization algorithm based on game theory is proposed to reduce system latency. In order to compare the system latency performance of partial offloading (PO) and full offloading (FO), the task scheduling and resource allocation problems under the two modes are studied. The results show that compared with FO, PO allows more flexible task scheduling, thereby reducing task processing delay.

[0005] However, the premise for the offloading rate to reach the capacity is the assumption that the input signal is Gaussian distributed. Due to the limitation of the complexity of the task equipment, the actual system usually adopts a finite modulation codebook, such as phase shift keying (PSK) and quadrature amplitude modulation (QAM). This leads to a large gap between capacity and achievable rate, making it challenging to accurately describe the task offloading delay based on capacity. Based on the achievable rate of the finite codebook, the present invention proposes an optimization problem of delay minimization for delay-sensitive MEC scenarios and designs an optimization scheme for resource allocation and task ratio division. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a mobile edge computing resource allocation method based on a finite codebook, which estimates the achievable rate of the system by input-output mutual information. Since the expression of the achievable rate is a non-closed expression, the Jensen inequality is used to obtain the approximate value of the achievable rate, thereby simplifying the parameter optimization process. In addition, in order to cope with the situation of device task timeout, a penalty term is introduced in the objective function. Compared with the traditional optimization method based on the capacity formula, this method can further improve the delay performance of the system.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A mobile edge computing resource allocation method based on a finite codebook, comprising:

[0009] Build an edge computing partial offloading model for latency-sensitive tasks;

[0010] The achievable rate and the approximate achievable rate based on the finite codebook are estimated according to the input-output mutual information;

[0011] Based on the achievable rate, the transmission delay and computing delay of the task under the edge computing partial offloading model are calculated, and the task execution delay is obtained by combining the local computing delay of the task device.

[0012] The optimization goal is defined as minimizing system delay, and a penalty term is introduced for task timeouts.

[0013] Based on the approximate value of the achievable rate, the task offloading ratio and edge cloud computing resources are jointly optimized to complete resource allocation.

[0014] Furthermore, a method for allocating mobile edge computing resources based on a finite codebook further includes: the edge computing partial offloading model is composed of a base station equipped with an edge cloud and a plurality of task devices;

[0015] Each task device has a delay-sensitive computing task, and all task devices constitute a set

[0016] The maximum tolerable delay of all task devices is T max ;

[0017] The task description of task device k is Among them, D k (bits) indicates the data size of the task, C k Indicates the number of CPU cycles required to calculate each bit of data, f k Represents the computing resources of the task device;

[0018] F MEC Represents the overall computing resources of the edge cloud, which are used to allocate to all task devices for parallel edge computing, so that the task can be completed within the maximum tolerable delay T max Completed within;

[0019] Set the binary variable u k,0 Indicates whether the task device is uninstalled. k,0 =1, the task device k performs edge offloading through the cellular link, otherwise it does not perform offloading.

[0020] Furthermore, the achievable rate and the approximate achievable rate based on the finite codebook are estimated according to the input-output mutual information, including:

[0021] Assume that both the transmitter and receiver in the task offloading link are equipped with a single antenna, and obtain the receiving signal y of the base station of the kth offloading link k :

[0022]

[0023] Among them, h k represents the amplitude gain of the kth offloading link, p k represents the transmission power of device k, n k represents complex additive Gaussian white noise, and n~CN(0,σ 2 ), x k Represents the sending signal of the task device, which is taken from M with equal probability k Constellation chart and

[0024] When the transmitted signal is a limited codebook input, the task device sends a signal x k and the received signal y k The mutual information between them is expressed as:

[0025] R k =I(x k ;y k )=h(y k )-h(y k |x k)

[0026] Among them, h(y k |x k ) indicates sending signal x k and the received signal y k The conditional differential entropy between k ) represents the received signal y k The differential entropy of

[0027] Among them, h(y k |x k ) based on the received signal y k The conditional probability density function is calculated as follows:

[0028]

[0029] h(y k ) based on the received signal y k The marginal probability density function is calculated as:

[0030]

[0031] h(y k |x k ) and h(y k ) Substitute the task device to send signal x k and the received signal y k In the mutual information expression between k As the achievable rate of the offload link:

[0032]

[0033] in,

[0034] The achievable rate R of the offload link is calculated by Jensen’s inequality: k Approximate value of :

[0035]

[0036] in, The achievable rate R k The approximate achievable rate.

[0037] Further, based on the received signal y k The conditional probability density function is calculated to obtain h(y k |x k ) process includes:

[0038] Get the received signal y k The conditional probability density function of is:

[0039]

[0040] According to the received signal y k The conditional probability density function of the transmitted signal x is calculated k and the received signal y k The conditional differential entropy between:

[0041]

[0042] Based on the received signal y k The marginal probability density function is calculated to obtain h(y k ) process includes:

[0043] Get the received signal y k The marginal probability density function of is:

[0044]

[0045] According to the received signal y k The marginal probability density function is calculated to obtain the received signal y k The differential entropy of :

[0046]

[0047] Among them, h(y k ) represents the received signal y k The differential entropy of .

[0048] Furthermore, a mobile edge computing resource allocation method based on a limited codebook also includes:

[0049] When the signal at the transmitting end adopts Gaussian codebook, the capacity formula is expressed as:

[0050]

[0051] Among them, C k Indicates the number of CPU cycles required to calculate each bit of data.

[0052] Furthermore, based on the achievable rate, the transmission delay and computing delay of the task under the edge computing partial offloading model are calculated, and the task execution delay is obtained by combining the local computing delay of the task device, including:

[0053] Get the local computing latency of task device k:

[0054]

[0055] in, represents the local computation delay of task device k, β k Indicates the offloading ratio of task device k for task offloading;

[0056] Get the transmission delay and computing delay of the task when the task device k selects edge computing offloading:

[0057]

[0058] in, Indicates the transmission delay, represents the computation delay, represents the computing resources allocated by the edge cloud to the task device, and B represents the bandwidth of the cellular link;

[0059] according to and The calculated task execution delay of task device k performing edge computing offloading is:

[0060]

[0061] Among them, t k Indicates the task execution delay of task device k performing edge computing offloading.

[0062] Furthermore, the optimization goal is defined as minimizing system delay, and penalty items are introduced for task timeouts, including:

[0063] In the optimization process of the objective function, the approximate achievable rate is substituted into The calculation formula of is used to obtain the approximate delay of each task device.

[0064] Based on the delay approximation And the optimization goal determines the optimization problem:

[0065]

[0066] Among them, P1 represents the optimization problem, st represents the constraint condition, Indicates the maximum delay limit for task execution of task equipment. Indicates that the total computing resources allocated to the task devices do not exceed the total computing resources of the edge cloud. Indicates the range of unloading ratio;

[0067] Since P1 is a mixed integer nonlinear programming problem, for the constraints There are situations where there is no feasible solution due to task equipment timeout;

[0068] A penalty term is introduced for the case of task timeout, and the optimization problem is redefined based on the penalty term:

[0069]

[0070] in, represents the time when task device k times out, and ρ represents the violation of the delay constraint The penalty factor.

[0071] Furthermore, based on the approximate value of the achievable rate, the task offloading ratio and edge cloud computing resources are jointly optimized to complete resource allocation, including:

[0072] Based on the achievable rate approximation Calculate the offloading ratio β of task device k under edge offloading k for:

[0073]

[0074] Substitute β k To the optimization target, regenerate the optimization problem P3:

[0075]

[0076] in, The objective function in the optimization problem P3 is a convex function, and the constraints are convex constraints. The Lagrangian method is used to solve the optimization problem P3, and the corresponding Lagrangian function is:

[0077]

[0078] Among them, λ and ω k is the Lagrange multiplier, and the KKT condition is:

[0079]

[0080] Among them, when When η k =1, otherwise η k =0;

[0081] Solve the optimization problem P3 according to the KKT condition and obtain the optimal solution for the computing resources allocated by the edge cloud to the task device k:

[0082]

[0083] Among them, λ * The solution is:

[0084]

[0085] in, indicates that task device k performs edge offloading, and Correspondingly It means that task device k is not allocated edge computing resources.

[0086] The beneficial effects of the present invention are:

[0087] The present invention provides a mobile edge computing resource allocation method based on a finite codebook. The achievable rate of the system is estimated by input-output mutual information. Since the expression of the achievable rate is a non-closed expression, the approximate value of the achievable rate is obtained by using the Jensen inequality, thereby simplifying the parameter optimization process. In addition, in order to cope with the situation of device task timeout, a penalty term is introduced in the objective function. Compared with the traditional optimization method based on the capacity formula, the method can further improve the delay performance of the system.

[0088] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or they may be taught from the practice of the present invention. The purposes and other advantages of the present invention may be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0089] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0091] Figure 1 A method flow chart of a mobile edge computing resource allocation method based on a limited codebook in an embodiment of the present invention;

[0092] Figure 2 A method flow of a mobile edge computing resource allocation method based on a limited codebook in an embodiment of the present invention Figure 2 ;

[0093] Figure 3 A verification data diagram of the accuracy of the achievable rate approximation in a mobile edge computing resource allocation method based on a finite codebook in an embodiment of the present invention;

[0094] Figure 4 The invention relates to a mobile edge computing resource allocation method based on a limited codebook in an embodiment of the present invention. k System delay performance graph;

[0095] Figure 5 The edge computing resource F based on the mobile edge computing resource allocation method based on the limited codebook in the embodiment of the present invention MEC Delay performance diagram. DETAILED DESCRIPTION

[0096] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0097] like Figure 1 As shown, the present invention proposes a mobile edge computing resource allocation method based on a limited codebook, comprising:

[0098] S101. Build an edge computing partial offloading model for latency-sensitive tasks;

[0099] S102, estimating a achievable rate and an approximate achievable rate based on a finite codebook according to input-output mutual information;

[0100] S103, based on the achievable rate, calculate the transmission delay and computing delay of the task under the edge computing partial offloading model, and combine the local computing delay of the task device to obtain the task execution delay;

[0101] S104, defining the optimization goal as minimizing system delay, and introducing a penalty term for task timeout;

[0102] S105, based on the approximate value of the achievable rate, jointly optimize the task offloading ratio and edge cloud computing resources to complete resource allocation;

[0103] Specifically, 1. Figure 2 As shown in Figure 1, we consider an edge computing partial offloading system model, which consists of a base station equipped with an edge cloud and multiple task devices. Each task device has a latency-sensitive computing task, and these task devices form a set Assume that T max is the maximum tolerable delay of all task devices, that is, the delay tolerance of the task device calculation task. The task of the task device can be described as Among them, D k (bits) indicates the data size of the task, C k (CPU cycles per bit) is the number of CPU cycles required to calculate each bit of data; the computing resources of the task device are represented by f k (CPU cycles per second), F MEC It represents the overall computing resources of the edge cloud, which can be allocated to all task devices for parallel edge computing, so that the task can be completed within the maximum delay tolerance T max At the same time, define the binary variable u k,0 Indicates whether the task device is uninstalled. k,0 =1, the task device k performs edge offloading via the cellular link, otherwise it is 0. And because the size of the calculation result is usually small enough, the delay of the result download return is negligible.

[0104] 2. Signal transmission based on finite codebook: Consider that both the transmitter and receiver in the task offloading link are equipped with a single antenna. Therefore, the received signal y of the base station of the kth offloading link is k It can be expressed as:

[0105]

[0106] Among them, h k is the amplitude gain of the kth unloading link, p k represents the transmission power of device k, n k is additive white Gaussian noise (AWGN), and n~CN(0,σ 2 ), x k represents the transmission symbol of task device k, which is taken from M with equal probability k Constellation chart And E‖a‖ 2 = 1, when the signal at the transmitting end adopts Gaussian codebook, the capacity formula is:

[0107]

[0108] When the transmitted signal is a limited codebook input, the task device transmits a signal x k and the received signal y k The mutual information between can be expressed as:

[0109] R k =I(x k ;y k )=h(y k )-h(y k |x k )

[0110] Receive signal y k The conditional probability density function of is:

[0111]

[0112] and k The marginal probability density function expression is:

[0113]

[0114] According to the conditional probability density function, the signal x is sent k and the received signal y k The conditional differential entropy h(y k |x k ) can be calculated as:

[0115]

[0116] The received signal y calculated by the marginal probability density function k The differential entropy h(y k )for:

[0117]

[0118] The differential entropy h(y k ) and conditional differential entropy h(y k |x k ) into R k =I(x k ;y k )=h(y k )-h(y k |x k ), then the achievable rate of the offload link is:

[0119]

[0120] in, However, due to R k The complexity of integration makes it difficult to obtain a closed-form expression for the above formula, which makes parameter optimization complicated. Then, the achievable rate R is obtained by the Jensen inequality: k The approximate value of is:

[0121]

[0122] 3. Task offloading and computation latency characterization

[0123] 1) The local computing delay of task device k is

[0124]

[0125] Among them, f k represents the local computing power of task device k (CPU cycles per second), β k Indicates the offloading ratio of task device k.

[0126] 2) When task device k chooses edge computing offloading, the transmission and computing delays of the task are:

[0127]

[0128] in, represents the computing resources (CPU cycles per second) allocated by the edge cloud to task device k, and B represents the bandwidth of the cellular link. Since local computing and remote offloading are parallel processes, the latency of task device k to perform edge computing offloading is expressed as:

[0129]

[0130] Optimization goal: Unloading ratio β through joint optimization k and edge cloud computing resources Minimize the total system delay. In order to reduce the computational complexity, in the optimization process of the objective function, the approximate value of the achievable rate is substituted into The approximate delay of each task device is obtained The optimization problem is expressed as:

[0131]

[0132] Constraints Indicates the maximum delay limit for task execution of task equipment. Indicates that the total computing resources allocated to the task devices do not exceed the total computing resources of the edge cloud. Indicates the range of uninstall ratio.

[0133] 5. Joint optimization algorithm for offloading ratio partitioning and computing resource allocation

[0134] 1) Since this is a mixed integer nonlinear programming problem, for the constraints There may be situations where there is no feasible solution due to task device timeout, so a penalty term for the timed task device k is introduced into the objective function. Then the optimization objective can be re-expressed as:

[0135]

[0136] in, represents the time when task device k times out, and ρ represents the violation of the delay constraint The penalty factor of

[0137] 2) Unloading ratio division: Due to and The uninstall ratio is β k The linear function of , when the time of local execution and remote offloading plus calculation is equal, the delay of task device k is the smallest. The offloading ratio of task device k under edge offloading can be calculated as:

[0138]

[0139] 3) Allocation of edge cloud computing resources: Substitute the offloading ratio of task device k into the optimization target. The optimization problem can be reformulated as follows:

[0140]

[0141] in, The objective function in the optimization problem P3 is a convex function, and the constraints are convex constraints. Therefore, the Lagrangian method is used to solve it, and the corresponding Lagrangian function is:

[0142]

[0143] Among them, λ and ω k is the Lagrange multiplier, and the Karush-Kuhn-Tucker (KKT) condition is:

[0144]

[0145] Among them, when When η k =1, otherwise η k = 0. The above conditions can be used to obtain the optimal solution for the computing resources allocated by the edge cloud to the task device k:

[0146]

[0147] where λ * The solution is:

[0148]

[0149] also, indicates that task device k performs edge offloading; and Correspondingly This indicates that task device k is not allocated edge computing resources, that is, task device k does not offload the task to the edge cloud;

[0150] The present invention adopts the achievable rate as the unloading rate, and further optimizes the unloading ratio and computing resources jointly. Figure 3 The accuracy of the achievable rate approximation is verified. Figure 3 It can be seen that the achievable rate approximation With the exact value R k The agreement is very good, especially in the high and low SNR regions. At the same time, due to the inherent complexity of the mission equipment, the gap between the achievable rate and capacity becomes larger as the transmission power increases.

[0151] Figure 4 Given the transmission power p k The delay performance of the system. It can be seen that the system delay obtained by the achievable rate-based optimization decreases monotonically with the increase of transmission power and converges to a stable value. However, the system delay obtained by the capacity-based optimization scheme decreases initially and then increases with the transmission power p. k This is because as power increases, the gap between capacity and achievable rate becomes larger (see Figure 3), which ultimately reduces the effectiveness of system delay optimization.

[0152] Figure 5 Based on edge cloud computing resources F MEC The system latency based on achievable rate optimization proposed in the present invention decreases monotonically with the increase of edge cloud computing resources and is lower than the system latency based on capacity optimization. In addition, due to the achievable rate R k and capacity C k The performance gap between the two optimization schemes becomes more obvious as the modulation order decreases.

[0153] The above simulation results show that compared with the capacity-based optimization scheme, the achievable rate-based optimization scheme significantly reduces system latency and emphasizes the importance of finite codebook-based achievable rate analysis in delay-sensitive scenarios.

[0154] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A mobile edge computing resource allocation method based on a finite codebook, characterized in that: include: Build an edge computing partial offloading model for latency-sensitive tasks; The achievable rate and the approximate achievable rate based on the finite codebook are estimated according to the input-output mutual information; Based on the achievable rate, the transmission delay and computing delay of the task under the edge computing partial offloading model are calculated, and the task execution delay is obtained by combining the local computing delay of the task device. The optimization goal is defined as minimizing system delay, and a penalty term is introduced for task timeouts. Based on the approximate value of the achievable rate, the task offloading ratio and edge cloud computing resources are jointly optimized to complete resource allocation.

2. A mobile edge computing resource allocation method based on a limited codebook according to claim 1, characterized in that: Also includes: The edge computing partial offloading model consists of a base station equipped with an edge cloud and multiple task devices; Each task device has a delay-sensitive computing task, and all task devices constitute a set The maximum tolerable delay of all task devices is T max ; The task description of task device k is Among them, D k (bits) indicates the data size of the task, C k Indicates the number of CPU cycles required to calculate each bit of data, f k Represents the computing resources of the task device; F MEC Represents the overall computing resources of the edge cloud, which are used to allocate to all task devices for parallel edge computing, so that the task can be completed within the maximum tolerable delay T max Completed within; Set the binary variable u k,0 Indicates whether the task device is uninstalled. k,0 =1, the task device k performs edge offloading through the cellular link, otherwise it does not perform offloading.

3. The method for allocating mobile edge computing resources based on a limited codebook according to claim 1, characterized in that: The achievable rate and achievable rate approximation based on the finite codebook are estimated based on the input and output mutual information, including: Assume that both the transmitter and receiver in the task offloading link are equipped with a single antenna, and obtain the receiving signal y of the base station of the kth offloading link k : Among them, h k represents the amplitude gain of the kth offloading link, p k represents the transmission power of device k, n k represents complex additive Gaussian white noise, and n~CN(0,σ 2 ), x k Represents the sending signal of the task device, which is taken from the number of symbols M with equal probability k The finite codebook And Ε‖a‖ 2 = 1; When the transmitted signal is a limited codebook input, the task device sends a signal x k and the received signal y k The mutual information between them is expressed as: R k =I(x k ;y k )=h(y k )-h(y k |x k ) Among them, h(y k |x k ) indicates sending signal x k and the received signal y k The conditional differential entropy between k ) represents the received signal y k The differential entropy of Among them, h(y k |x k ) based on the received signal y k The conditional probability density function is calculated as follows: h(y k ) based on the received signal y k The marginal probability density function is calculated as: h(y k |x k ) and h(y k ) Substitute the task device to send signal x k and the received signal y k In the mutual information expression between k As the achievable rate of the offload link: in, The achievable rate R of the offload link is calculated by Jensen’s inequality: k Approximate value of : in, The achievable rate R k The approximate achievable rate.

4. A mobile edge computing resource allocation method based on a limited codebook according to claim 3, characterized in that: Based on the received signal y k The conditional probability density function is calculated to obtain h(y k |x k ) process includes: Get the received signal y k The conditional probability density function of is: According to the received signal y k The conditional probability density function of the transmitted signal x is calculated k and the received signal y k The conditional differential entropy between: Based on the received signal y k The marginal probability density function is calculated to obtain h(y k ) process includes: Get the received signal y k The marginal probability density function of is: According to the received signal y k The marginal probability density function is calculated to obtain the received signal y k The differential entropy of : Among them, h(y k ) represents the received signal y k The differential entropy of .

5. A mobile edge computing resource allocation method based on a limited codebook according to claim 2 or 3, characterized in that: Also includes: When the signal at the transmitting end adopts Gaussian codebook, the capacity formula is expressed as: Among them, C k Indicates the number of CPU cycles required to calculate each bit of data.

6. A mobile edge computing resource allocation method based on a limited codebook according to claim 1, characterized in that: Based on the achievable rate, the transmission delay and computing delay of the task under the edge computing partial offloading model are calculated, and the task execution delay is obtained by combining the local computing delay of the task device, including: Get the local computing latency of task device k: in, represents the local computation delay of task device k, β k Indicates the offloading ratio of task device k for task offloading; Get the transmission delay and computing delay of the task when the task device k selects edge computing offloading: in, Indicates the transmission delay, represents the computation delay, represents the computing resources allocated by the edge cloud to the task device, and B represents the bandwidth of the cellular link; according to and The calculated task execution delay of task device k performing edge computing offloading is: Among them, t k Indicates the task execution delay of task device k performing edge computing offloading.

7. The method for allocating mobile edge computing resources based on a limited codebook according to claim 1, characterized in that: The optimization goal is defined as minimizing system latency, and penalty items are introduced for task timeouts, including: In the optimization process of the objective function, the approximate achievable rate is substituted into The calculation formula of is used to obtain the approximate delay of each task device. Based on the delay approximation And the optimization goal determines the optimization problem: Among them, P1 represents the optimization problem, st represents the constraint condition, Indicates the maximum delay limit for task execution of task equipment. Indicates that the total computing resources allocated to the task devices do not exceed the total computing resources of the edge cloud. Indicates the range of unloading ratio; Since P1 is a mixed integer nonlinear programming problem, for the constraints There are situations where there is no feasible solution due to task equipment timeout; A penalty term is introduced for the case of task timeout, and the optimization problem is redefined based on the penalty term: in, represents the time when task device k times out, and ρ represents the violation of the delay constraint The penalty factor.

8. The method for allocating mobile edge computing resources based on a limited codebook according to claim 1, characterized in that: Based on the approximate achievable rate, the task offloading ratio and edge cloud computing resources are jointly optimized to complete resource allocation, including: Based on the achievable rate approximation Calculate the offloading ratio β of task device k under edge offloading k for: Substitute β k To the optimization target, regenerate the optimization problem P3: in, The objective function in the optimization problem P3 is a convex function, and the constraints are convex constraints; The Lagrangian method is used to solve the optimization problem P3, and the corresponding Lagrangian function is: Among them, λ and ω k is the Lagrange multiplier, and the KKT condition is: Among them, when When η k =1, otherwise η k =0; Solve the optimization problem P3 according to the KKT condition and obtain the optimal solution for the computing resources allocated by the edge cloud to the task device k: Among them, λ * The solution is: in, indicates that task device k performs edge offloading, and Correspondingly It means that task device k is not allocated edge computing resources.

Images