Time delay and cost joint optimization method in wireless heterogeneous network with aggregation nodes

Abstract

The invention belongs to the technical field of heterogeneous network resource allocation, and discloses a time delay and cost joint optimization method in a wireless heterogeneous network with aggregation nodes, which comprises the following steps of: S1, establishing a mathematical model of each parameter in MEC task unloading according to a 5G-WiFi heterogeneous network assisted by the aggregation nodes; s2, constructing a mixed integer nonlinear programming optimization problem according to the mathematical models established in the step S1; and S3, solving the mixed integer nonlinear programming optimization problem constructed in the step S2. According to the time delay and cost joint optimization method in the wireless heterogeneous network with the aggregation node, the problems of task unloading and resource management in the 5G-WiFi mobile edge computing network assisted by the aggregation node are solved through the solution of equivalent transformation and centralized solution, and the balance between time delay and cost is realized; the introduction of the aggregation node enables the network to obtain better performance.

Description

Technical Field

[0001] The present invention relates to the technical field of heterogeneous network resource allocation, and in particular to a method for jointly optimizing delay and cost in a wireless heterogeneous network with aggregation nodes. Background Art

[0002] Mobile edge computing provides new opportunities to meet the needs of computationally intensive applications on resource-constrained mobile devices. According to IDC's forecast, by 2025, there will be 55.9 billion connected devices, and the amount of data generated by IoT devices connected to the network alone will be 79.4ZB, not to mention other traffic consumption. At the same time, diversified applications such as ultra-high-definition streaming, augmented reality (AR), virtual reality (VR), and intelligent autonomous driving are becoming more and more popular, and the demand for instant information continues to grow. The fusion of the above two trends has led to a large number of data communications and processing requests.

[0003] The explosive growth of data traffic is an inevitable problem. Although mobile edge computing can be used as a good solution to deal with the above problems, when a large number of users offload computing tasks at the same time, the MEC system will suffer from severe spectrum congestion, resulting in a decrease in cellular network service and quality of experience (QoE). With the deployment of WiFi networks based on IEEE-802.11, using the additional bandwidth resources of wireless local area networks (WLANs) to ensure high transmission efficiency and QoE has also become a solution that can be considered. Therefore, a heterogeneous network with 5G-WiFi coexistence is a suitable solution. Both 5G and WiFi can provide users with better service quality, but their networking conditions are different. 5G has advantages over WiFi in terms of the security of operator-authorized spectrum and service quality. However, due to its commercial nature, 5G usually has a higher cost per bit of data transmission. Compared with 5G base stations, the deployment of WiFi access points is more flexible, and the cost per bit of WiFi communication is lower and more energy-efficient, which is more important for battery-powered wireless devices. How to comprehensively consider the characteristics and differences of the two networks and complete task offloading scheduling and resource management optimization in a two-layer heterogeneous network is a prominent problem that needs to be solved.

[0004] Offloading 5G cellular traffic to WiFi networks plays an important role in alleviating the increasing burden on 5G cellular networks. However, since WiFi networks are based on the CSMA / CA competitive channel access mechanism, a large number of terminal devices (TDs) accessing and excessive traffic offloading will cause serious access conflicts to the WiFi network, and the coverage of WiFi access points (APs) can usually only achieve partial coverage, and there will still be TDs that are not within the coverage of the WiFi network. TDs that are not covered by APs can only use the computing resources of the 5G cellular network, and cannot fully utilize the resources in the heterogeneous network, resulting in waste. Summary of the invention

[0005] The purpose of the present invention is to provide a method for jointly optimizing delay and cost in a wireless heterogeneous network with aggregation nodes, solve the problem of task offloading and resource management in an AN-assisted 5G-WiFi mobile edge computing network through equivalent transformation and centralized solution, and achieve a trade-off between delay and cost. The introduction of AN enables the network to achieve better performance.

[0006] To achieve the above object, the present invention provides a method for jointly optimizing delay and cost in a wireless heterogeneous network with an aggregation node, comprising the following steps:

[0007] S1. Based on the 5G-WiFi heterogeneous network assisted by aggregation nodes, mathematical models of various parameters in MEC task offloading are established, including terminal equipment TD communication model, aggregation node AN communication model, task completion delay model, and energy consumption and commercial cost model;

[0008] S2, introducing weight factors, adjusting the weight between task completion delay and cost according to different requirements of network scenarios, and combining various mathematical models established in step S1 to construct a mixed integer nonlinear programming optimization problem;

[0009] S3. Introduce auxiliary variables to convert the target problem into a smooth optimization problem, and solve the mixed integer nonlinear programming optimization problem constructed in step S2.

[0010] Preferably, in step S1, establishing a TD communication model of a terminal device specifically includes the following steps:

[0011] S111, Class A TD communication model;

[0012] (1) Class A TD n The uplink transmission rate of the offload task through the 5G cellular network is as follows:

[0013]

[0014] Where ω0 is the value assigned by the base station BS to the Class A TD n The bandwidth, p n Indicates TD n The transmission power; Indicates TD n The channel gain between the base station BS and Indicates TD n The distance to the base station BS, ξ is the path loss factor, σ 2 is the background noise power in the 5G cellular network;

[0015] (2) Class A TD nThe uplink transmission rate of the task offloaded through the WiFi network is as follows:

[0016]

[0017] in, Indicates Class A TD n The transmission rate that can be achieved after successfully seizing the WiFi channel; q n ,q m ∈[0,1] represents class A TD n and AN m The weight in channel competition. The higher the weight, the more advantage it has in channel competition. Indicates the terminal devices covered by both 5G cellular network and WiFi network, set Indicates terminal devices that are only covered by 5G cellular networks. Represents the set of all terminal devices, labeled n; Represents the set of all aggregation nodes AN, labeled m; represents the decision variable for users other than the nth user to choose to use the WIFI network;

[0018] S112, Class B TD communication model;

[0019] (1) Class B TD n The uplink transmission rate of the offload task through the 5G cellular network is as follows:

[0020]

[0021] (2) Class B TD n The uplink transmission rate of the task offloaded through the 5G cellular network to the aggregation node AN is as follows:

[0022]

[0023] in, Indicates Class B TD n To AN m The channel gain, Indicates Class B TD n To AN m The distance between them is set, and ξ=4 is set for subsequent simulation; the decision variable k n,m ∈{0,1} indicates whether Class B TDn chooses to access the WiFi network through ANm. n Select AN m Connect to WiFi network n,m =1 otherwise k n,m =0.

[0024] Preferably, in step S1, an aggregation node AN communication model is established, and the specific process is as follows:

[0025] AN m After receiving the task data transmitted by Class B TD, it is unloaded to the edge server near the WiFi access point AP through the WiFi network, and follows the CSMA / CA wireless communication model. The uplink transmission rate is as follows:

[0026]

[0027] in, Indicates AN m The transmission rate that can be achieved after successfully occupying the WiFi channel; x n The decision variable indicating whether the nth user uses the WIFI network;

[0028] To ensure that there is no congestion, for AN m The constraints are as follows:

[0029]

[0030] in, Indicates the ratio of Class B devices to offload tasks to WiFi access points (APs). The task offloading ratio does not exceed 1. n Indicates TD n The total amount of data for the task.

[0031] Preferably, in step S1, a task completion delay model is established, which specifically includes the following steps:

[0032] S131, Class A TD task completion delay model, including:

[0033] (1) Class A TD n The local computing latency of some tasks is as follows:

[0034]

[0035] in, Represented as TD n The CPU cycle frequency represents TD n computing power; Indicates the proportion of local execution of tasks, c n Indicates TD n The total computational effort of the task;

[0036] (2) Class A TD n If part of the task is offloaded to the edge server near the WiFi access point AP or the base station BS for processing, the transmission delay of this part is as follows:

[0037]

[0038] Among them, f A represents the CPU cycle frequency of the edge server near the WiFi access point AP, f B Indicates the CPU cycle frequency of the edge server near the base station BS; therefore, Class A TD n The calculation delays of some tasks at the edge are as follows:

[0039]

[0040] (3) In summary, Class A TD n The total task offloading latency is as follows:

[0041]

[0042] Class A TD n The task offloading and task local calculation are carried out in parallel, and the actual total task processing delay is as follows:

[0043]

[0044] And Class A TD n There are strict task completion delay requirements, and the total task processing delay should not be greater than The constraints are as follows:

[0045]

[0046] S132, Class B TD task completion delay model, including:

[0047] The tasks of Class B TD are processed in three directions in parallel, namely local computing, offloading to the WiFi access point AP through the aggregation node AN, and offloading to the base station BS for processing. n The local computation latency is as follows:

[0048]

[0049] Class B TD n The uplink transmission delays for offloading to the aggregation node AN and the base station BS are as follows:

[0050]

[0051] Since Class B TD n Access to the WiFi access point AP is through the aggregation node AN relay, so the transmission delay offloaded to the WiFi access point AP also includes the transmission delay from the aggregation node AN to the WiFi access point AP, as shown below:

[0052]

[0053] Finally, Class B TD n The calculation delays on different edge servers are as follows:

[0054]

[0055] In order to fully utilize the computing power of edge servers, Class B TD n The task offloading is transmitted to different ends in parallel, so the total completion delay of tasks to different ends is as follows:

[0056]

[0057] In summary, Class B TD n The actual total task processing delay is as follows:

[0058]

[0059] Similarly, Class B TD n There are strict task completion delay requirements, and the total task processing delay meets the following constraints:

[0060]

[0061] Preferably, in step S1, in establishing the energy consumption and business cost model, the A-type TD n and Class B TD n The transmission costs for task offloading are as follows:

[0062]

[0063] Among them, ζ represents the corresponding traffic cost generated by using 5G cellular network for task offloading; no additional cost is incurred when using WiFi network transmission.

[0064] Preferably, in step S2, a mixed integer nonlinear programming optimization problem is constructed according to the various mathematical models established in step S1, and the specific process is as follows:

[0065] A weight factor 0<β<1 is introduced to adjust the weight between task completion delay and cost according to different requirements of network scenarios. The objective function is the weighted sum of total task completion delay and total cost, as shown below:

[0066]

[0067] Among them, α is a parameter used to balance the different scales between task completion delay and cost;

[0068] The optimization goal is to jointly optimize the task offloading decision and the task offloading rate to minimize the weighted sum of delay and cost, while meeting the task completion delay requirements of TD and the uplink transmission rate requirements of the aggregation node AN. The optimization problem can be expressed as follows:

[0069]

[0070] Among them, constraints C1 and C2 indicate that the task offloading decisions of the two types of TDs are 0-1 discrete variables; constraint C3 represents the one-select constraint of the task offloading decision variable of class B TD, ensuring that each class B TD will only select one aggregation node AN for task offloading; constraints C4 and C5 represent the strict delay constraints of class A TD and class B TD respectively; constraints C6 and C7 indicate that the task offloading rate cannot exceed the upper limit; constraint C8 indicates that the uplink transmission rate of the aggregation node AN should be greater than the incoming traffic to ensure stable operation.

[0071] Preferably, in step S3, solving the mixed integer nonlinear programming optimization problem constructed in step S2 comprises the following steps:

[0072] S31. Introducing auxiliary variables and The target problem is equivalently transformed into a smooth optimization problem, then and The following conditions must be met:

[0073]

[0074] S32. After introducing auxiliary variables, ensure that the transformed problem is equivalent to the original problem. For type A TD, introduce the following complementary constraints:

[0075]

[0076] Similarly, the following complementary constraints are introduced for Class B TD to ensure the equivalence of the problem:

[0077]

[0078] S33, the new equivalent smooth optimization problem is as follows:

[0079]

[0080] Therefore, the present invention adopts the above-mentioned joint optimization method of delay and cost in a wireless heterogeneous network with aggregation nodes, and solves the task offloading and resource management problems in the AN-assisted 5G-WiFi mobile edge computing network through equivalent transformation and centralized solution, thereby achieving a trade-off between delay and cost. The introduction of AN enables the network to achieve better performance.

[0081] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a flow chart of a method for joint optimization of delay and cost in a wireless heterogeneous network with aggregation nodes of the present invention;

[0083] Figure 2 This is a 5G-WiFi heterogeneous network scenario diagram based on AN assistance of the present invention. DETAILED DESCRIPTION

[0084] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0085] like Figure 1 As shown, the present invention provides a method for jointly optimizing delay and cost in a wireless heterogeneous network with an aggregation node, comprising the following steps:

[0086] S1. Based on the 5G-WiFi heterogeneous network assisted by aggregation nodes, mathematical models of various parameters in MEC task offloading are established, including terminal equipment TD communication model, aggregation node AN communication model, task completion delay model, and energy consumption and commercial cost model;

[0087] S2, introducing weight factors, adjusting the weight between task completion delay and cost according to different requirements of network scenarios, and combining various mathematical models established in step S1 to construct a mixed integer nonlinear programming optimization problem;

[0088] S3. Introduce auxiliary variables to convert the target problem into a smooth optimization problem, and solve the mixed integer nonlinear programming optimization problem constructed in step S2.

[0089] Example

[0090] S1. A mathematical model is established based on the 5G-WiFi heterogeneous network assisted by aggregation nodes.

[0091] like Figure 2 As shown in the figure, a 5G-WiFi heterogeneous network MEC task offloading scenario with aggregation node (AN) assistance exists in the network. There is one 5G base station (5G BS) and one WiFi access point (WiFi AP). There is no cross-layer interference between the WiFi network and the 5G cellular network. Edge servers with computing capabilities are deployed near the BS and AP. Several ANs that are associated with both the WiFi network and the 5G cellular network are deployed within the coverage of the AP. The BS covers all terminal devices (TDs) in the network. The AP can only cover some TDs due to its own functional limitations. The TDs covered by both the 5G cellular network and the WiFi network are expressed as Class A TDs and recorded as a set TDs that are only covered by 5G cellular networks are represented as Class B TDs and recorded as a set For the convenience of description, the set of all TDs is denoted as The label is n; the set of all ANs is recorded as The label is m. The task profile of TD is composed of (d n ,c n ) indicates that d n Indicates TD n The total amount of data for the task, c n Indicates TD n The total computational effort of the task.

[0092] For all TDs, use Indicates the proportion of tasks executed locally. In order to reduce the task completion delay of Class B TD, its tasks are divided into three categories for processing, namely, local execution, offloading to edge servers near the BS, and offloading to edge servers near the AP. Indicates the ratio of tasks that Class B devices offload to APs. The task offload ratio cannot exceed 1. The constraints are as follows:

[0093]

[0094] For class A TD use the decision variable x n ∈{0,1} to indicate whether the A-type TDn chooses to access the WiFi network. n Select Connect to WiFi network x n =1 otherwise x n = 0. For class B TDn, use decision variable k n,m ∈{0,1} to indicate whether ANm is selected to access the WiFi network. n Select AN m Connect to WiFi network n,m =1 otherwise k n,m = 0. Class B TD n In order to fully utilize all resources in the network, some tasks must be offloaded to the WiFi network, and only one AN can be selected to access the WiFi network. The constraints are as follows:

[0095]

[0096] S11. Establish a TD communication model.

[0097] S111, Class A TD communication model.

[0098] (1) Class A TD is in the common coverage of two networks, that is, it uses 5G cellular network or WiFi network for communication. The access system of 5G cellular network is based on OFDMA. There is no interference between channels and between different BSs. Therefore, Class A TD n The uplink transmission rate of the offload task through the 5G cellular network is as follows:

[0099]

[0100] Where ω0 is the BS assigned to the Class A TD n The bandwidth, p n Indicates TD n The transmission power; Indicates TD n The channel gain between the BS and Indicates TD n The distance to the BS, ξ is the path loss factor, σ 2 is the background noise power in the 5G cellular network.

[0101] (2) If all TDs accessing the WiFi network share a common channel and compete for channel access rights through CSMA / CA, then Class A TD n The uplink transmission rate of the task offloaded through the WiFi network is as follows:

[0102]

[0103] in, Indicates Class A TD n The transmission rate that can be achieved after successfully seizing the WiFi channel; q n ,q m ∈[0,1] represents class A TD n and AN m The weight in channel competition. The higher the weight, the more advantage it has in channel competition. The decision variable that indicates whether other users except the nth user choose to use the WIFI network.

[0104] S112, Class B TD communication model.

[0105] (1) Class B TD n The uplink transmission rate of the offload task through the 5G cellular network is as follows:

[0106]

[0107] (2) Class B TD n If the edge server near the AP is used, it is connected to the AN through the 5G cellular network and then to the AP.n The uplink transmission rate of offloading tasks to AN through the 5G cellular network is as follows:

[0108]

[0109] in, Indicates Class B TD n To AN m The channel gain, Indicates Class B TD n To AN m The distance between them is set, and ξ=4 is set for subsequent simulations.

[0110] S12. Establish an AN communication model.

[0111] AN m After receiving the task data transmitted by Class B TD, it is unloaded to the edge server near the AP through the WiFi network and follows the CSMA / CA wireless communication model. The uplink transmission rate is as follows:

[0112]

[0113] in, Indicates AN m The transmission rate that can be achieved after successfully occupying the WiFi channel; x n The decision variable indicating whether the nth user uses the WIFI network.

[0114] If an AN has a good channel gain for a Class B TD, a large number of Class B TDs will offload tasks through the AN. However, the AN often does not have strong enough load performance. m To ensure that there is no congestion, the constraints are as follows:

[0115]

[0116] S13. Establish a task completion delay model.

[0117] Since the task results returned from the edge server are often very small, the transmission delay of the task result return is ignored in the task processing delay. Since the task parts of type A TD and type B TD are offloaded in different ways, the composition of task processing delay is also different.

[0118] S131, Class A TD task completion delay model.

[0119] (1) Class A TD n The local computing latency of some tasks is as follows:

[0120]

[0121] in, Represented as TD n The CPU cycle frequency represents TD n computing power.

[0122] (2) In addition to local computing tasks, Class A TD n You can choose to offload the remaining part to the edge server near the AP or BS for processing. The transmission delay of this part is as follows:

[0123]

[0124] Among them, f A represents the CPU cycle frequency of the edge server near the AP, f B represents the CPU cycle frequency of the edge server near the BS; therefore, the A-type TD n The calculation delays of some tasks at the edge are as follows:

[0125]

[0126] (3) In summary, Class A TD n The total task offloading latency is as follows:

[0127]

[0128] Class A TD n The task offloading and task local calculation are carried out in parallel, and the actual total task processing delay is as follows:

[0129]

[0130] And Class A TD n There are strict task completion delay requirements, and the total task processing delay should not be greater than The constraints are as follows:

[0131]

[0132] S132. Class B TD task completion delay model.

[0133] The tasks of Class B TD are processed in three directions in parallel, namely local calculation processing, offloading processing to AP through AN, and offloading processing to BS. n The local computation latency is as follows:

[0134]

[0135] Class B TD nThe uplink transmission delays for offloading to AN and BS are as follows:

[0136]

[0137] Since Class B TD n The access to AP is through AN relay, so the transmission delay to AP also includes the transmission delay from AN to AP, as shown below:

[0138]

[0139] Finally, Class B TD n The calculation delays on different edge servers are as follows:

[0140]

[0141] In order to fully utilize the computing power of edge servers, Class B TD n The task offloading is transmitted to different ends in parallel, so the total completion delay of tasks to different ends is as follows:

[0142]

[0143] In summary, Class B TD n The actual total task processing delay is as follows:

[0144]

[0145] Similarly, Class B TD n There are strict task completion delay requirements, and the total task processing delay should not be greater than The constraints are as follows:

[0146]

[0147] S14. Establish energy consumption and business cost model.

[0148] First, the business fee cost is TD n The traffic transmission cost of 5G cellular network is ζ ($ / bit). When using WiFi network for transmission, no additional cost is incurred. n and Class B TD n The transmission costs for task offloading are as follows:

[0149]

[0150] S2. According to the mathematical models established in step S1, a mixed integer nonlinear programming (MINLP) optimization problem is constructed.

[0151] A weight factor 0<β<1 is introduced to adjust the weight between task completion delay and cost according to different requirements of network scenarios. The objective function is the weighted sum of total task completion delay and total cost, as shown below:

[0152]

[0153] Among them, α is a parameter used to balance the different scales between task completion delay and cost.

[0154] The optimization goal is to jointly optimize the task offloading decision and the task offloading rate to minimize the weighted sum of delay and cost, while meeting the task completion delay requirements of TD and the uplink transmission rate requirements of AN. The optimization problem can be expressed as follows:

[0155]

[0156] Among them, constraints C1 and C2 indicate that the task offloading decisions of the two types of TDs are 0-1 discrete variables; constraint C3 indicates the one-select constraint of the task offloading decision variable of class B TD, ensuring that each class B TD will only select one AN for task offloading; constraints C4 and C5 indicate the strict delay constraints of class A TD and class B TD respectively; constraints C6 and C7 indicate that the task offloading rate cannot exceed the upper limit; constraint C8 indicates that the uplink transmission rate of AN should be greater than the incoming traffic to ensure stable operation. There are highly coupled non-convex terms in both the objective function and the constraint terms, and they are non-smooth, so the optimization problem is a non-smooth mixed integer nonlinear programming (MINLP) problem.

[0157] S3. Solve the MINLP joint optimization problem constructed in step S2.

[0158] S31. Due to the non-smoothness of the task completion delay model in the objective function, the problem cannot be solved directly, so an auxiliary variable is introduced and The target problem is equivalently transformed into a smooth optimization problem, and the introduced and The following conditions must be met:

[0159]

[0160] S32. Since TD processes tasks locally and remotely in parallel, the task completion delay described depends on the larger one of the local and edge ends. After introducing auxiliary variables, in order to ensure that the transformed problem is equivalent to the original problem, the following complementary constraints are introduced for Class A TD:

[0161]

[0162] Similarly, the following complementary constraints are introduced for Class B TD to ensure the equivalence of the problem:

[0163]

[0164] S33, the new equivalent smooth optimization problem is as follows:

[0165]

[0166] Based on the above process, the BARON solver is used to obtain the global optimal solution to evaluate the performance and advantages of the proposed scheme in MEC task offloading.

[0167] Therefore, the present invention adopts the above-mentioned joint optimization method of delay and cost in a wireless heterogeneous network with aggregation nodes, and solves the task offloading and resource management problems in the AN-assisted 5G-WiFi mobile edge computing network through equivalent transformation and centralized solution, thereby achieving a trade-off between delay and cost. The introduction of AN enables the network to achieve better performance.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for joint optimization of delay and cost in wireless heterogeneous networks with aggregation nodes, characterized in that: The following steps are involved: S1. Based on the 5G-WiFi heterogeneous network assisted by aggregation nodes, mathematical models of various parameters in MEC task offloading are established, including terminal equipment TD communication model, aggregation node AN communication model, task completion delay model, and energy consumption and commercial cost model; S2, introducing weight factors, adjusting the weight between task completion delay and cost according to different requirements of network scenarios, and combining various mathematical models established in step S1 to construct a mixed integer nonlinear programming optimization problem; S3. Introduce auxiliary variables to convert the target problem into a smooth optimization problem, and solve the mixed integer nonlinear programming optimization problem constructed in step S2.

2. The method for joint optimization of delay and cost in a wireless heterogeneous network with an aggregation node according to claim 1, characterized in that: In step S1, a TD communication model of a terminal device is established, which specifically includes the following steps: S111, Class A TD communication model; (1) Class A TD n The uplink transmission rate of the offload task through the 5G cellular network is as follows: Where ω0 is the value assigned by the base station BS to the Class A TD n The bandwidth, p n Indicates TD n The transmission power; Indicates TD n The channel gain between the base station BS and Indicates TD n The distance to the base station BS, ξ is the path loss factor, σ 2 is the background noise power in the 5G cellular network; (2) Class A TD n The uplink transmission rate of the task offloaded through the WiFi network is as follows: in, Indicates Class A TD n The transmission rate that can be achieved after successfully seizing the WiFi channel; q n ,q m ∈[0,1] represents class A TD n and AN m The weight in channel competition. The higher the weight, the more advantage it has in channel competition. Indicates the terminal devices covered by both 5G cellular network and WiFi network, set Indicates terminal devices that are only covered by 5G cellular networks. Represents the set of all terminal devices, labeled n; Represents the set of all aggregation nodes AN, labeled m; represents the decision variable for users other than the nth user to choose to use the WIFI network; S112, Class B TD communication model; (1) Class B TD n The uplink transmission rate of the offload task through the 5G cellular network is as follows: (2) Class B TD n The uplink transmission rate of the task offloaded through the 5G cellular network to the aggregation node AN is as follows: in, Indicates Class B TD n To AN m The channel gain, Indicates Class B TD n To AN m The distance between them is set, and ξ=4 is set for subsequent simulation; the decision variable k n,m ∈{0,1} indicates whether Class B TDn chooses to access the WiFi network through ANm. n Select AN m Connect to WiFi network n,m =1 otherwise k n,m =0.

3. The method for joint optimization of delay and cost in a wireless heterogeneous network with aggregation nodes according to claim 2, characterized in that: In step S1, the aggregation node AN communication model is established, and the specific process is as follows: AN m After receiving the task data transmitted by Class B TD, it is unloaded to the edge server near the WiFi access point AP through the WiFi network, and follows the CSMA / CA wireless communication model. The uplink transmission rate is as follows: in, Indicates AN m The transmission rate that can be achieved after successfully occupying the WiFi channel; x n The decision variable indicating whether the nth user uses the WIFI network; To ensure that there is no congestion, for AN m The constraints are as follows: in, Indicates the ratio of Class B devices to offload tasks to WiFi access points (APs). The task offloading ratio does not exceed 1. n Indicates TD n The total amount of data for the task.

4. The method for joint optimization of delay and cost in a wireless heterogeneous network with aggregation nodes according to claim 1, characterized in that: In step S1, a task completion delay model is established, which specifically includes the following steps: S131, Class A TD task completion delay model, including: (1) Class A TD n The local computing latency of some tasks is as follows: in, Represented as TD n The CPU cycle frequency represents TD n computing power; Indicates the proportion of local execution of tasks, c n Indicates TD n The total computational effort of the task; (2) Class A TD n If part of the tasks are offloaded to the edge server near the WiFi access point AP or the base station BS for processing, the transmission delay of this part is as follows: Among them, f A represents the CPU cycle frequency of the edge server near the WiFi access point AP, f B Indicates the CPU cycle frequency of the edge server near the base station BS; therefore, Class A TD n The calculation delays of some tasks at the edge are as follows: (3) In summary, Class A TD n The total task offloading latency is as follows: Class A TD n The task offloading and task local calculation are carried out in parallel, and the actual total task processing delay is as follows: And Class A TD n There are strict task completion delay requirements, and the total task processing delay should not be greater than The constraints are as follows: S132, Class B TD task completion delay model, including: The tasks of Class B TD are processed in three directions in parallel, namely local computing, offloading to the WiFi access point AP through the aggregation node AN, and offloading to the base station BS for processing. n The local computation latency is as follows: Class B TD n The uplink transmission delays for offloading to the aggregation node AN and the base station BS are as follows: Since Class B TD n Access to the WiFi access point AP is through the aggregation node AN relay, so the transmission delay offloaded to the WiFi access point AP also includes the transmission delay from the aggregation node AN to the WiFi access point AP, as shown below: Finally, Class B TD n The calculation delays on different edge servers are as follows: In order to fully utilize the computing power of edge servers, Class B TD n The task offloading is transmitted to different ends in parallel, so the total completion delay of tasks to different ends is as follows: In summary, Class B TD n The actual total task processing delay is as follows: Similarly, Class B TD n There are strict task completion delay requirements, and the total task processing delay meets the following constraints:

5. The method for joint optimization of delay and cost in a wireless heterogeneous network with aggregation nodes according to claim 1, characterized in that: In step S1, the energy consumption and business cost model is established. n and Class B TD n The transmission costs for task offloading are as follows: Among them, ζ represents the corresponding traffic cost generated by using 5G cellular network for task offloading; no additional cost is incurred when using WiFi network transmission.

6. The method for joint optimization of delay and cost in a wireless heterogeneous network with aggregation nodes according to claim 1, characterized in that: In step S2, a mixed integer nonlinear programming optimization problem is constructed according to the mathematical models established in step S1. The specific process is as follows: A weight factor 0<β<1 is introduced to adjust the weight between task completion delay and cost according to different requirements of network scenarios. The objective function is the weighted sum of total task completion delay and total cost, as shown below: Among them, α is a parameter used to balance the different scales between task completion delay and cost; The optimization goal is to jointly optimize the task offloading decision and the task offloading rate to minimize the weighted sum of delay and cost, while meeting the task completion delay requirements of TD and the uplink transmission rate requirements of the aggregation node AN. The optimization problem can be expressed as follows: st Among them, constraints C1 and C2 indicate that the task offloading decisions of the two types of TDs are 0-1 discrete variables; constraint C3 represents the one-select constraint of the task offloading decision variable of class B TD, ensuring that each class B TD will only select one aggregation node AN for task offloading; constraints C4 and C5 represent the strict delay constraints of class A TD and class B TD respectively; constraints C6 and C7 indicate that the task offloading rate cannot exceed the upper limit; constraint C8 indicates that the uplink transmission rate of the aggregation node AN should be greater than the incoming traffic to ensure stable operation.

7. The method for joint optimization of delay and cost in a wireless heterogeneous network with aggregation nodes according to claim 1, characterized in that: In step S3, solving the mixed integer nonlinear programming optimization problem constructed in step S2 includes the following steps: S31. Introducing auxiliary variables and The target problem is equivalently transformed into a smooth optimization problem, then and The following conditions must be met: S32. After introducing auxiliary variables, ensure that the transformed problem is equivalent to the original problem. For type A TD, introduce the following complementary constraints: Similarly, the following complementary constraints are introduced for Class B TD to ensure the equivalence of the problem: S33, the new equivalent smooth optimization problem is as follows:

Images