A computation offloading optimization method based on D2D communication
Through D2D communication and graph theory algorithm, the task offloading strategy of user equipment is optimized, the problem of limited computing resources of MEC edge servers is solved, the delay and energy consumption are reduced, and the task processing efficiency is improved.
Patent Information
- Application Number
- CN202210610433.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-31
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2042-05-31
AI Technical Summary
In the case of insufficient computing power of user terminal equipment and excessive number of users, the computing resources of MEC edge servers are limited, resulting in low processing efficiency of delay-sensitive tasks and energy-consuming-sensitive tasks.
Using the calculation and offload optimization method based on D2D communication, by establishing a classification model of user equipment, MEC server and MCC remote cloud server, using graph theory principle to build a time delay weight graph and energy consumption edge weight graph, and using the maximum matching minimum cost algorithm, Select the appropriate uninstall target to achieve task uninstallation.
It effectively reduces equipment energy consumption and task execution delay, optimizes the processing efficiency of delay-sensitive tasks and energy-consuming-sensitive tasks, and makes full use of the computing resources of edge servers and nearby users.
Smart Images

Figure CN115103405B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of communications, and in particular to a method for optimizing computational offloading based on D2D communications. Background Art
[0002] D2D communication and mobile edge computing, as two important technologies of 5G, have received widespread attention. Through D2D communication, terminals can use the computing resources of nearby idle devices to provide computing for devices with limited computing power. At the same time, users far away from the base station can communicate with nearby devices, enjoying a high signal-to-noise ratio and communicating at a high rate.
[0003] D2D communication can enable computing collaboration between user terminals. Reference [Li Zishu, Xie Renchao, Sun Li, et al. A review of mobile edge computing [J]. Telecommunications Science, 2018, 34(1): 87-101.] For the D2D collaborative mobile edge computing offloading scenario, user devices can perform D2D communication to offload tasks and use orthogonal frequency division multiple access for communication. By considering the required energy consumption, partial offloading and resource allocation constraints, the total delay minimization is formulated as a mixed integer nonlinear programming problem. A strategy for combining partial offloading and resource allocation is proposed. Simulation results show that compared with other schemes, the algorithm effectively reduces the task processing delay. The paper [Pu L, Chen X, Xu J, et al. D2D fogging: An energy-efficient and incentive-aware task offloading framework via networkassisted D2D collaboration [J]. IEEE Journal on Selected Areas in Communications, 2016, 34 (12): 3887-3901.] proposed a task offloading method that combines D2D with MEC. The proposed framework shares the computing resources of mobile devices to achieve energy consumption during task execution. Since D2D communication does not require the base station and communicates directly between users, it can effectively reduce transmission delay and the transmission power of user devices. In the paper [Diao X, Zheng J, Wu Y, et al. Joint computing resource, power, and channel allocations for D2D-assisted and NOMA-based mobile edge computing [J]. IEEE Access, 2019, 7: 9243-9257.], in the MEC computing offloading scenario, the authors consider computing resources, transmission power and channel allocation to fully utilize the advantages of D2D communication.In the literature [Orakzai FA, Iqbal M, Naeem M, et al. Energy efficient joint radio resource management in D2D assisted cellular communication [J]. Telecommunication Systems, 2018, 69 (4): 505-517.], the authors proposed a joint task allocation and resource allocation offloading system in the MEC environment, in which devices can use nearby devices to assist in communicating with MEC, thereby reducing the energy consumption of executing tasks. Summary of the invention
[0004] The technical problem to be solved by the present invention is to overcome the problems in the prior art of insufficient computing power of user terminal devices and limited computing resources of MEC edge servers when the number of users is too large, and to provide a computing offloading optimization method for D2D communication. With the assistance of D2D communication, the computing resources of nearby users are fully utilized, which can not only reduce the transmission delay of users, but also reduce the computing pressure of edge servers, thereby achieving the goal of optimizing the delay of delay-sensitive tasks and optimizing the energy consumption of energy-sensitive tasks.
[0005] The present invention specifically adopts the following technical solutions to solve the above technical problems:
[0006] A method for optimizing computing offloading based on D2D communication, based on a base station, and an MEC server and an MCC remote cloud server respectively connected to the base station via high-speed optical fiber communication; and each user device respectively connected to the base station and the MEC server via wireless communication; a user device with a task request to be offloaded and an idle user device are connected via D2D communication; for each user device with a task request to be offloaded, steps A to B are executed to classify the task request to be offloaded of the user device, and a classification category of the task request to be offloaded is obtained; and then steps C to D are executed to realize task offloading of the task request to be offloaded of the user device;
[0007] Step A: For the user device i with the task request to be offloaded, establish the communication model of the user device and the MEC server computing model respectively; and define the task request to be offloaded of the user device as Among them, I i represents the amount of task request data to be offloaded by user device i, O i represents the amount of data after the calculation of the task request to be offloaded of user device i, Z i represents the CPU computing cycles required to execute the task request to be offloaded of user device i, and They represent the uplink and downlink information fed back by the user equipment i to the base station, T i req represents the execution delay limit of the task to be offloaded request of user device i;
[0008] Step B: Calculate the execution delay T of the task to be offloaded requesting user equipment i i l , User equipment execution energy consumption Compare the execution delay limit T of the task request to be offloaded i req The value of the execution delay T of user equipment i i l The value of the user device i is compared with the execution energy consumption The value of and the preset energy consumption threshold for the user equipment to complete the task to be offloaded;
[0009] Condition 1: Execution delay limit T of the task request to be offloaded i req Less than the user equipment execution delay T i l The value of
[0010] Condition 2: User equipment execution energy consumption The value of is lower than the energy consumption threshold of the user equipment;
[0011] If only condition 1 is met, the task request to be offloaded is a delay-sensitive task;
[0012] If only condition 2 is met, the task request to be offloaded is an energy-sensitive task;
[0013] If both conditions 1 and 2 are met, it is considered that the task request to be offloaded does not need to be offloaded;
[0014] Step C: Divide the user equipment with task requests to be offloaded, idle user equipment, MEC server, and MCC remote cloud server into the following four types of nodes: user nodes with task requests to be offloaded, idle user equipment nodes, MEC server resource pre-allocation nodes, and MCC remote cloud server resource pre-allocation nodes; and establish a delay edge weight graph and an energy consumption edge weight graph for the four types of nodes according to the principles of graph theory;
[0015] Step D: Solve the delay edge weight graph and the energy consumption edge weight graph in the order of delay edge weight graph first and energy consumption edge weight graph later. According to the results of solving the delay weight graph and the energy consumption weight graph, choose one of the following three different offloading schemes: offloading to idle user equipment, offloading to MEC server, and offloading to MCC remote cloud server to realize the offloading of delay-sensitive tasks and energy-sensitive tasks to be offloaded.
[0016] Furthermore, in the aforementioned step A, establishing the communication model of the user equipment is specifically as follows:
[0017] According to the following formula:
[0018] f i l =(1-θ)f i 0 ,
[0019] Calculate the computing power of the current device i,
[0020] According to the following formula:
[0021]
[0022] Calculate the transmission rate of the D2D communication connection between the user equipment with task requests and the idle user equipment; and the transmission rate of the communication rate data between the user and the base station, where C represents the maximum transmission rate, B is the bandwidth, S is the signal power, N is the noise power, and θ is the percentage of the user equipment computing resources occupied.
[0023] Furthermore, in the aforementioned step A, the calculation model of the MEC server is established as follows:
[0024] Calculated as follows:
[0025]
[0026] Calculate the remaining computing resources of the MEC server, where f mec Indicates the remaining computing resources of the MEC server. represents the raw computing power of the MEC server, μ i is a binary data used to indicate whether user i occupies the computing resources of the MEC server. i mec Indicates the minimum computing resources required by user i.
[0027] Furthermore, in the aforementioned step B, the following formula is used:
[0028] T i l =Z i / f i l ,
[0029] Calculate the current device execution delay T with task request i l ;
[0030] According to the following formula:
[0031]
[0032] Calculate the current device execution energy consumption Where T i l Indicates the execution delay of the current device. Indicates the current device execution energy consumption, Z i Indicates the CPU computing cycles required to execute the task, f i l represents the computing power of user device i, ρ i It represents the energy consumption of each CPU cycle executed by device i;
[0033] According to the following formula:
[0034]
[0035] Calculate user equipment energy consumption Energy consumption of user equipment and compare the execution delay T of user equipment i i l The execution delay limit T of the task request to be offloaded i req , if the user equipment consumes Energy consumption of user equipment The difference is less than the energy consumption threshold of the user equipment. And user equipment i performs delay T i l Less than the execution delay limit T of the task to be offloaded i req , then it is considered that the task request to be uninstalled does not need to be uninstalled; among them, represents the energy consumption of user device i, represents the execution energy consumption of user device i, Indicates that user equipment i has completed the energy consumption threshold, T i l represents the execution delay of user equipment i, T i req Indicates the execution delay limit of the task request to be offloaded.
[0036] Furthermore, in the aforementioned step C, according to the principle of graph theory, the delay edge weight graph and the energy consumption edge weight graph are established for the four types of nodes; specifically, according to the following formula:
[0037]
[0038] Calculate the minimum computing resources required by user device i;
[0039] According to the following formula:
[0040]
[0041] Calculate the delay of offloading the task to be offloaded to the idle device.
[0042] According to the following formula:
[0043]
[0044] Calculate the delay of offloading the task to be offloaded to the MEC server; where f i mec represents the minimum computing resources required by user i, represents the computing capability of device j, and Respectively represent the uplink and downlink data transmission rates; if User device j can be used as an alternative offloading target for user device i;
[0045] According to the following formula:
[0046]
[0047] Calculate the energy consumption value of offloading the task to be offloaded to the idle device;
[0048] According to the following formula:
[0049]
[0050] Compute the offloaded tasks to MEC
[0051] The energy consumption value of the server; where P i ul With P i dl represents the transmit and receive power of device i, P i d and Represent the transmit power of device i and j respectively.
[0052] Furthermore, in the aforementioned step D, according to the following formula:
[0053]
[0054] The minimum execution time of latency-sensitive tasks and the minimum energy consumption of energy-sensitive tasks are calculated; where μ i,t ∈{0,1}, indicating whether the task to be offloaded is a delay-sensitive task, μ i ∈{0,1} indicates whether the task to be offloaded is offloaded to the MEC server.
[0055] Furthermore, in the aforementioned step D, a maximum matching and minimum cost graph theory algorithm is used to solve the time delay edge weight graph and the energy consumption edge weight graph.
[0056] The present invention adopts the above technical solution to produce the following technical effects:
[0057] (1) The method of the present invention proposes an offloading scheme for reducing device energy consumption and task execution delay for heterogeneous cellular networks based on D2D communication. In the scenario of the present invention, mobile devices can offload their tasks to MEC servers or idle mobile devices near them, and an optimization problem is defined. The entire process of reasonably allocating offloading targets is designed in the MEC server. The offloading requests are classified according to the deadline and energy constraint of the requesting device, and then the delay sensitivity map and energy consumption sensitivity map are established respectively, and the maximum matching minimum cost map algorithm is used to find the appropriate offloading target.
[0058] (2) The edge computing offloading method based on D2D communication proposed in the present invention makes full use of the computing resources of the edge server and nearby users. Compared with local execution, the present invention has achieved effective reduction in both latency and energy consumption for latency-sensitive tasks and energy-sensitive tasks. At the same time, the present invention can utilize limited MEC server computing resources to perform more task offloading. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] Figure 1 It is a schematic diagram of the process of the present invention;
[0060] Figure 2 It is a schematic diagram of the system model of the present invention;
[0061] Figure 3 offloading the adjacency graph for users of the present invention;
[0062] Figure 4 It is the energy consumption diagram under different numbers of user equipments of the present invention;
[0063] Figure 5 It is the energy consumption diagram under different idle device proportions of the present invention;
[0064] Figure 6 It is a delay cost diagram under different numbers of user equipments of the present invention;
[0065] Figure 7 This is a delay cost diagram under different idle device ratios of the present invention. DETAILED DESCRIPTION
[0066] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.
[0067] Various aspects of the invention are described herein with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the invention are not limited to those described in the accompanying drawings. It should be understood that the invention is implemented by any of the various concepts and embodiments described above, as well as the concepts and embodiments described in detail below, because the concepts and embodiments disclosed in the invention are not limited to any implementation. In addition, some aspects disclosed in the invention may be used alone or in any appropriate combination with other aspects disclosed in the invention.
[0068] like Figure 1 As shown, the present invention relates to a method for optimizing calculation offloading based on D2D communication. The process of the method is as follows Figure 1 First, for the user device i with a task request to be offloaded, the communication model of the user device and the MEC server computing model are established respectively; and the task request to be offloaded of the user device is defined as Among them, I i represents the amount of task request data to be offloaded by user device i, O i represents the amount of data after the calculation of the task request to be offloaded of user device i, Z i represents the CPU computing cycles required to execute the task request to be offloaded of user device i, and They represent the uplink and downlink information fed back by the user equipment i to the base station, T i req represents the execution delay limit of the task to be offloaded request of user device i;
[0069] Then, the task requests to be offloaded are classified according to latency and energy consumption: the execution latency T of the task request to be offloaded by user device i is calculated respectively. i l , User equipment execution energy consumption Compare the execution delay limit T of the task request to be offloaded i req The value of the execution delay T of user equipment i i l The value of the user device i is compared with the execution energy consumption The value of and the preset energy consumption threshold for the user equipment to complete the task to be offloaded;
[0070] Condition 1: Execution delay limit T of the task request to be offloaded i req Less than the user equipment execution delay T i l The value of
[0071] Condition 2: User equipment execution energy consumption The value of is lower than the energy consumption threshold of the user equipment;
[0072] If only condition 1 is met, the task request to be offloaded is a delay-sensitive task;
[0073] If only condition 2 is met, the task request to be offloaded is an energy-sensitive task;
[0074] If both conditions 1 and 2 are met, it is considered that the task request to be offloaded does not need to be offloaded;
[0075] Then, user devices with task requests to be offloaded, idle user devices, MEC servers, and MCC remote cloud servers are divided into the following four types of nodes: user nodes with task requests to be offloaded, idle user device nodes, MEC server resource pre-allocation nodes, and MCC remote cloud server resource pre-allocation nodes; and according to the principles of graph theory, the four types of nodes are established with a delay edge weight graph and an energy consumption edge weight graph.
[0076] Then, the graph theory algorithm of maximum matching and minimum cost is used to solve the unloading decision of delay-sensitive tasks. If the idle device has been selected, the idle device is removed from the candidate node. Otherwise, it continues to determine whether the MEC server is selected. If the MEC server is selected, the MEC server is removed. Otherwise, after solving the delay-sensitive tasks, the MEC server is used to calculate the weights of energy-sensitive tasks, and the graph theory algorithm of maximum matching and minimum cost is used to solve the unloading decision of energy-sensitive tasks. The MEC server will make the final unloading decision: unloading to idle user devices, unloading to MEC servers, and unloading to MCC remote cloud servers; realizing the unloading of delay-sensitive tasks to be unloaded and energy-sensitive tasks to be unloaded.
[0077] The method of the present invention is applied to the scenario of D2D communication edge network such as Figure 2 As shown, it is mainly divided into: user equipment that supports D2D communication, MEC server and MCC remote cloud server. Among them, user equipment includes devices with tasks to be offloaded and idle devices, which establish a connection with the MEC server on the base station side in real time, and send information such as the user's location and channel noise. User tasks are divided into delay-sensitive tasks and energy-sensitive tasks according to the required delay requirements and energy consumption requirements. User equipment can offload tasks to the nearest user equipment or server for execution, so that some delay-sensitive tasks can be successfully completed. At the same time, if the user device lacks sufficient energy, its tasks can be offloaded to three other devices: neighboring user devices, MEC servers or MCC remote cloud servers. The MEC server makes decisions on the task offloading strategy by obtaining information from user devices. It mainly includes the following steps:
[0078] The first step is to establish the communication model of the user device and the MEC server computing model for the user device i that has the task request to be offloaded;
[0079] (1) User Equipment Communication Model
[0080] In the model of the present invention, user devices can only communicate with MEC servers. The devices continuously provide feedback to the base station to obtain information such as the device's transmission power, location, and channel noise. The MEC server then makes user unloading decisions based on this information and its own computing resources. If the MEC server believes that D2D collaboration is the most efficient way to execute tasks, then mobile devices can communicate directly without the need for assistance and intervention from the base station. Background services and computing task resources that cannot be unloaded by the device occupy the device's CPU through θ. Therefore, the computing power of the current device i can be expressed as:
[0081] f i l =(1-θ)f i 0 (1)
[0082] where f i 0 represents the initial computing capability of device i, f i l Represents the available computing power of device i. Each device has its own unique signal to interference noise ratio (SINR) at a given time. Each device in the cellular network periodically reports its status to the base station in the form of a channel quality indicator (CQI). The required transmission delay is expressed as:
[0083]
[0084]
[0085] In the above formula, T i ul With T i dl They represent the uplink and downlink time of feedback information to the base station respectively. and They represent the uplink and downlink information fed back by user equipment i to the base station, and They represent the uplink and downlink transmission rates of user equipment i respectively. However, since the feedback uplink and downlink delays are small, they are ignored in the present invention.
[0086] In the model proposed in the present invention, the SINR value of D2D is reported to the base station, according to the Shannon formula:
[0087]
[0088] The transmission rate between the user equipment with task requests and the idle user equipment via the D2D communication connection can be calculated; as well as the transmission rate of the communication rate data between the user and the base station. In the above formula, C represents the maximum transmission rate, B is the bandwidth, S is the signal power, and N is the noise power.
[0089] (2) MEC Server Model
[0090] The MEC server is close to the user device. The MEC server can communicate directly with the user device on one end and with the remote cloud server on the other end. Therefore, user devices that require real-time response can use low-latency services. The present invention considers integrating traditional cloud computing servers into MEC servers near users to run complex tasks. There are different virtual machines (VMs) in the MEC server to support different services at the same time. Network administrators can statically adjust the computing power of each virtual machine or dynamically allocate computing power. The MEC server allocates a minimum computing power that meets the latency constraints to each task to provide scalability. The remaining computing power of the MEC server is expressed as:
[0091]
[0092] where f mec Indicates the remaining computing resources of the MEC server. Represents the raw computing power of the MEC server, μ i is a binary data used to indicate whether user i occupies the computing resources of the MEC server. i mec Indicates the minimum computing resources required by user i.
[0093] The second step is to define the optimization problem to be solved.
[0094] Define the task request to be uninstalled by the user device as Among them, I i represents the amount of task request data to be offloaded from user device i, O i represents the amount of data after the task request to be offloaded of user device i is calculated, Z i represents the CPU computing cycles required to execute the task request to be offloaded of user device i, and They represent the uplink and downlink information fed back by the user equipment i to the base station, Ti req represents the execution delay limit of the task to be offloaded request of user device i;
[0095] During the local execution process, the user device calculates the task through its own CPU, and the total execution delay is T i l It mainly includes execution delay, which can be expressed as:
[0096] T i l =Z i / f i l (6)
[0097] According to the above expression, the energy required to complete the task can be obtained through equation (7). If T i l >T i req , the user device itself cannot meet the delay constraint, and can send an offload request to the MEC server. Similarly, if the current device energy is lower than the threshold for completing the task, an offload request can also be sent to the edge server. The energy consumption of the task executed locally is:
[0098]
[0099] where ρ i It represents the energy consumption of device i for executing each CPU cycle. If the user device sends an offloading request to the edge server, the edge server can provide the user device with three different offloading solutions: D2D-assisted offloading, offloading to MEC server, and offloading to remote cloud server.
[0100] If device i needs to offload tasks to device j, the cost mainly includes the sending and receiving of tasks, as well as the task execution overhead of device j. In summary, the total delay and energy consumption of D2D offloading can be expressed as:
[0101]
[0102]
[0103] in represents the computing power of device j, P i d and Represent the transmission power of device i and j respectively. If Device j can be used as an alternative D2D offloading target for device i. At the same time, the MEC server can allocate its own resources to delay-sensitive tasks with higher priority to save more computing resources and satisfy more users. Therefore, the MEC server can meet the delay constraint through equation (10). The delay required for MEC offloading can be expressed as formula (11), and the energy required by the device can be expressed as equation (12):
[0104]
[0105]
[0106]
[0107] Where P i ul With P i dl represents the transmit and receive power of device i, and Represents the transmission rate of uplink and downlink data respectively. If the edge server cannot accept the task, and the request to offload the task to the neighboring device is also invalid, then the task will be returned to the cloud for processing. Therefore, in the scenario proposed by the present invention, since the user can only communicate with the MEC server, if the task needs to be offloaded to the remote cloud, the edge server is required to upload it, then the energy consumption of offloading to the remote cloud is the same as the energy consumption of offloading to the edge server:
[0108] The goal of the offloading scenario proposed in the present invention is to minimize the execution time of delay-sensitive tasks in consideration of the delay constraints of the tasks, and to minimize the energy consumption of energy-sensitive tasks, which can be specifically expressed as follows:
[0109]
[0110] where μ i,t ∈{0,1}, indicating whether the task is a delay-sensitive task, μ i ∈{0,1} indicates whether the task is offloaded to the MEC server.
[0111] In order to solve the optimization problem in a less complex way, the present invention will use graph theory to solve the optimal unloading decision. First, by comparing T i req With T i l The value of can classify tasks into delay-sensitive and energy-sensitive types. If the limited delay of the task is less than T i l, then the task is a delay-sensitive task. At the same time, if the energy of the current device is lower than the threshold for completing the task, it is considered an energy-sensitive task. The first stage is for delay-sensitive tasks. This type of task needs to be assigned a higher priority and user satisfaction, so resource allocation can be prioritized. The second stage is for devices with energy consumption requirements. The energy of mobile devices has always been a precious resource. Tasks can be unloaded to save energy costs. At the same time, in order to avoid increasing the communication cost of the cellular network, the present invention allows user devices that satisfy formula (14) to choose to execute tasks locally.
[0112]
[0113] in represents the energy consumption of device i, represents the threshold energy consumption of device i. If all the requirements are met, it means that the device has sufficient energy to perform the task and can meet the delay constraint. Otherwise, in any other case, the user device will send a task offload request to the MEC server.
[0114] Next, we start to build the corresponding edge weight graph. Low-energy devices and devices with delay-sensitive tasks send their offloading requests to the MEC server. The MEC server first divides the devices with tasks to be offloaded, idle devices, MEC server computing resources allocated to the corresponding users, and remote cloud server computing resources allocated to the corresponding users into four types of nodes. Then, the task nodes that need to be offloaded are added to the delay edge weight graph and the energy consumption edge weight graph respectively.
[0115] Through the above node division, the delay value of the delay edge weight graph is calculated by equations 8 and 11, and the energy consumption value of the energy consumption edge weight graph is calculated by equations 9 and 12. Therefore, we can construct the following Figure 3 The first one is a delay-sensitive task graph, which is used to effectively allocate MEC resources to meet the delay constraints, and the second one is an energy-sensitive task graph, which is used to reduce the energy consumption of user devices.
[0116] Figure 3 The offloading decision diagram of the present invention presents a topology of a computation offloading cellular network based on D2D communication. The computing resources in the edge server dedicated to processing tasks are divided into multiple corresponding virtual machine computing resources. If the MEC resources are insufficient, it can forward the tasks to the remote cloud. Figure 3In the example scenario, suppose device 1 chooses to offload the task to idle device 6, while devices 2 and 3 choose to offload the task to the MEC server, and device 7 is not selected. Then in the second stage, for the establishment of the energy-sensitive task graph, device 7 can be used as a candidate device, and the MEC computing resources used in the first stage also need to be deleted. The edge weight of the delay-sensitive graph in the above figure is the delay required for task execution, and the edge weight of the energy-sensitive graph is the energy consumption required by the task device when the task is executed. Define E(a,b) as the edge weight from node a to node b, and the node set in the delay-sensitive graph is S t , the node set in the energy-sensitive graph is S e .
[0117] Based on the scenario of D2D communication, under the network architecture of the combined edge cloud and remote cloud, the present invention discusses the user task scheduling problem of multiple users, aiming to minimize the delay or energy consumption of user devices. The framework proposed in the present invention overcomes the disadvantages of using only D2D communication or MEC servers, and makes full use of limited computing resources to serve user devices. According to the completion deadline and energy constraints of user tasks, user computing tasks are divided into delay-sensitive tasks and energy-sensitive tasks. Based on graph theory, the corresponding delay weight graph and energy weight graph are constructed for idle devices, MEC servers and MCC remote cloud servers, and the optimization objectives are defined as delay and energy consumption. The maximum matching minimum cost graph algorithm is used to solve the unloading strategy. First, the delay weight graph is solved to obtain the unloading decision of the delay-sensitive task and the computing resources required. The remaining computing resources are used for the calculation of energy-sensitive tasks, and the energy weight graph is solved to obtain the unloading decision of the energy-sensitive task.
[0118] Figure 4 and Figure 5 Analyze the average energy consumption between different algorithms. Figure 4 The change of user equipment from 20 to 100 is analyzed in the figure, and four scenarios are compared and analyzed. From the analysis of the figure, it is found that the performance of the algorithm proposed in the present invention is much better than other single offloading scenarios, because the present invention makes full use of the computing resources of the D2D scenario and the MEC server. Compared with the scenario without offloading, the performance of the algorithm of the present invention is 92% better than that of local processing. Figure 5 The changes in task proportion from 30% to 90% were analyzed, and the same trend of change can be observed.
[0119] Figure 6 and Figure 7 The delay changes in different scenarios are analyzed. Figure 6The changes in the number of users between 20 and 100 were analyzed. For the D2D collaboration scenario, the ratio of idle devices to devices with tasks remained unchanged, and the average latency change trend was relatively stable. However, for the MEC scenario, when the number of user devices exceeded 40, the latency changed as the number of users increased, resulting in insufficient MEC computing resources and increased latency. Figure 7 It shows that when the number of tasks is fixed at 50, the proportion of user tasks varies between 30% and 90%. Compared with the local execution by the user, the latency effect of the present invention is significantly improved.
[0120] The method of the present invention proposes an offloading scheme for reducing device energy consumption and task execution delay for heterogeneous cellular networks based on D2D communication. In the scenario of the present invention, mobile devices can offload their tasks to MEC servers or idle mobile devices near them, and an optimization problem is defined. The entire process of reasonably allocating offloading targets is designed in the MEC server. The offloading requests are classified according to the deadline and energy constraints of the requesting devices, and then the delay-sensitive graph and the energy-sensitive graph are established respectively, and the maximum matching minimum cost graph algorithm is used to find the appropriate offloading target. The edge computing offloading method based on D2D communication proposed in the present invention makes full use of the computing resources of the edge server and nearby users. Compared with local execution, for delay-sensitive tasks and energy-sensitive tasks, the present invention has achieved effective reductions in both latency and energy consumption. At the same time, the present invention can use limited MEC server computing resources to perform more task offloading.
[0121] Although the present invention has been described above with preferred embodiments, it is not intended to limit the present invention. A person skilled in the art of the present invention may make various modifications and improvements without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the definition of the claims.
Claims
1. A computing offloading optimization method based on D2D communication, based on a base station, and a MEC server and an MCC remote cloud server connected to the base station via high-speed optical fiber communication; and each user equipment connected to the base station and the MEC server by wireless communication respectively; the user equipment with task requests to be offloaded and the idle user equipment are connected via D2D communication; characterized in that, For each user device having a task request to be uninstalled, perform steps A and B respectively to classify the task request to be uninstalled of the user device and obtain the classification category of the task request to be uninstalled; then perform steps C and D to realize task uninstallation of the task request to be uninstalled of the user device; Step A: For the user device i with the task request to be offloaded, establish the communication model of the user device and the MEC server computing model respectively; and define the task request to be offloaded of the user device as Among them, I i represents the amount of task request data to be offloaded by user device i, O i represents the amount of data after the task request to be offloaded of user device i is calculated, Z i represents the CPU computing cycles required to execute the task request to be offloaded of user device i, and They respectively represent the uplink and downlink information fed back by the user equipment i to the base station, represents the execution delay limit of the task to be offloaded request of user device i; Step B: Calculate the execution delay T of the task to be offloaded requesting user device i i l , User equipment execution energy consumption Compare the execution delay limit T of the task request to be offloaded i req The value of the execution delay T of user equipment i i l The value of the user device i is compared with the execution energy consumption The value of and the preset energy consumption threshold for the user equipment to complete the task to be offloaded; Condition 1: Execution delay limit T of the task request to be offloaded i req Less than the user equipment execution delay T i l The value of Condition 2: User equipment execution energy consumption The value of is lower than the energy consumption threshold of the user equipment; If only condition 1 is met, the task request to be offloaded is a delay-sensitive task; If only condition 2 is met, the task request to be offloaded is an energy-sensitive task; If both conditions 1 and 2 are met, it is considered that the task request to be offloaded does not need to be offloaded; Step C: Divide the user equipment with task requests to be offloaded, idle user equipment, MEC server, and MCC remote cloud server into the following four types of nodes: user nodes with task requests to be offloaded, idle user equipment nodes, MEC server resource pre-allocation nodes, and MCC remote cloud server resource pre-allocation nodes; and establish a delay edge weight graph and an energy consumption edge weight graph for the four types of nodes according to the principles of graph theory; Step D: Solve the delay edge weight graph and the energy consumption edge weight graph in the order of the delay edge weight graph first and the energy consumption edge weight graph later. According to the results of solving the delay edge weight graph and the energy consumption edge weight graph, choose one of the following three different offloading schemes: offloading to idle user devices, offloading to MEC servers, and offloading to MCC remote cloud servers to realize the offloading of delay-sensitive tasks and energy-sensitive tasks to be offloaded.
2. According to claim 1, a method for optimizing computing offloading based on D2D communication is characterized in that: In the step A, establishing the communication model of the user equipment is specifically as follows: According to the following formula: Calculate the computing power of the current device i, According to the following formula: Calculate the transmission rate between the user equipment with task request and the idle user equipment via D2D communication connection; and the transmission rate of the communication rate data between the user and the base station, where C represents the maximum transmission rate, B is the bandwidth, S is the signal power, N is the noise power, and θ is the percentage of the user equipment computing resources occupied.
3. The method for optimizing computing offloading based on D2D communication according to claim 1, characterized in that: In step A, the calculation model of the MEC server is established by calculating according to the following formula: Calculate the remaining computing resources of the MEC server, where f mec Indicates the remaining computing resources of the MEC server. Represents the raw computing power of the MEC server, μ i is a binary data used to indicate whether user i occupies the computing resources of the MEC server. i mec Indicates the minimum computing resources required by user i.
4. The method for optimizing computing offloading based on D2D communication according to claim 1, characterized in that: In step B, follow the formula: Calculate the current device execution delay T with task request i l ; According to the following formula: Calculate the current device execution energy consumption Where T i l Indicates the execution delay of the current device. Indicates the current device execution energy consumption, Z i Indicates the CPU computing cycles required to execute the task, f i l represents the computing power of user device i, ρ i It represents the energy consumption of each CPU cycle executed by device i; According to the following formula: Calculate user equipment energy consumption Energy consumption of user equipment and compare the execution delay T of user equipment i i l The execution delay limit T of the task request to be offloaded i req , if the user equipment consumes Energy consumption of user equipment The difference is less than the energy consumption threshold of the user equipment. And user equipment i performs delay T i l Less than the execution delay limit T of the task to be offloaded i req , then it is considered that the task request to be uninstalled does not need to be uninstalled; in, represents the energy consumption of user device i, represents the execution energy consumption of user device i, Indicates that user equipment i has completed the energy consumption threshold, T i l represents the execution delay of user equipment i, T i req Indicates the execution delay limit of the task request to be offloaded.
5. The method for optimizing computing offloading based on D2D communication according to claim 1, characterized in that: In the step C, the delay edge weight graph and the energy consumption edge weight graph are established for the four types of nodes according to the principle of graph theory; specifically, according to the following formula: Calculate the minimum computing resources required by user device i; According to the following formula: Calculate the delay of offloading the task to be offloaded to the idle device. According to the following formula: Calculate the delay of offloading the task to be offloaded to the MEC server; where f i mec represents the minimum computing resources required by user i, represents the computing capability of device j, and Respectively represent the uplink and downlink data transmission rates; if User device j can be used as an alternative offloading target for user device i; According to the following formula: Calculate the energy consumption value of offloading the task to be offloaded to the idle device; According to the following formula: Calculate the energy consumption of the task to be offloaded to the MEC server; where P i ul With P i dl represents the transmit and receive power of device i, P i d and Represent the transmit power of device i and j respectively.
6. The method for optimizing computing offloading based on D2D communication according to claim 1, characterized in that: In step D, according to the following formula: The minimum execution time of latency-sensitive tasks and the minimum energy consumption of energy-sensitive tasks are calculated; where μ i,t ∈{0,1}, indicating whether the task to be offloaded is a delay-sensitive task, μ i ∈{0,1} indicates whether the task to be offloaded is offloaded to the MEC server.
7. The method for optimizing computing offloading based on D2D communication according to claim 1, characterized in that: In the step D, a maximum matching minimum cost graph theory algorithm is used to solve the time delay edge weight graph and the energy consumption edge weight graph.
