A method for minimizing latency of drone-assisted edge computing

Abstract

A method for minimizing the latency of drone-assisted edge computing, comprising the following steps: setting up K Internet of Things nodes, N drones, and M ground base station servers in the system; each Internet of Things device generates a computing task W k , each task can be offloaded to a drone for computing, or can be offloaded to a base station server using the drone as a relay for computing; taking the minimization of the computing latency of the task of the node with the most time-consuming as the optimization objective, and taking the position of the drone, the offloading decision of the node, the computing resource allocation strategy of the drone and the server, and the communication resource allocation strategy of the drone and the server as the optimization variables, a mathematical model of the optimization problem P is established; the problem P is solved by using a method combining the K-Means clustering algorithm and cross-iteration. The obtained solution is deployed in the system to achieve the purpose of minimizing the latency. The present invention can achieve the minimization of the computing offloading latency.

Description

Technical Field

[0001] The present invention belongs to the technical field of 5G network computing offloading, and in particular relates to a method for minimizing the time delay of drone-assisted edge computing. Background Art

[0002] Edge computing is one of the core technologies of 5G networks. The deployment location of edge servers in edge computing will seriously affect the performance of edge computing. Applying drones to edge computing scenarios makes services more flexible and more adaptable to the current changing application environment. Due to the energy consumption limitations of drones, their carrying capacity and flight time are limited. It is very necessary to strengthen air-ground collaboration to provide services for IoT nodes. For complex scenarios where ground nodes cannot be directly connected to base stations, drones with limited computing resources are used. They can act as servers and relays to transmit tasks to base station servers with more powerful computing capabilities. It is of great significance to study the node offloading strategy, bandwidth and computing resource allocation under this model. Summary of the invention

[0003] The present invention aims to overcome the above-mentioned shortcomings of the prior art and provide a method for minimizing the time delay of edge computing assisted by a drone.

[0004] The present invention takes advantage of the flexible deployment of drones and designs a drone-assisted MEC system computing offloading model. Considering drones at fixed altitudes, the present invention jointly optimizes the drone position deployment, node offloading decisions, and resource allocation of drones and servers to achieve the goal of minimizing the maximum computing offloading delay.

[0005] In order to achieve the above object, a method for minimizing the delay of drone-assisted edge computing includes the following steps:

[0006] Step 1: Set up a specific application scenario. Set up K IoT nodes, N drones, and M ground base station servers in the system, where D = {d 1 ,d 2 ,…,d K} represents the set of IoT nodes, U = {u 1 ,u 2 ,…,u N} represents the set of drones, A={a 1 ,a 2 ,…,a M} represents a set of M base stations. All IoT nodes and base station servers are deployed on the ground, and the location of IoT nodes is represented by Indicates that the location of the base station server is All drones are deployed at the same height H, and the positions of the drones are represented by express.

[0007] Step 2: Each IoT device generates a computing task W k =(C k ,F k ), the two parameters represent the task data size and the number of CPU cycles required to process one bit of task data. Each task can be offloaded to the drone for calculation, or unloaded to the base station server for calculation using the drone as a relay. k,n ∈{0,1} represents node d k Do you choose drone? n Uninstall, 1 for selection, 0 for non-selection, set each IoT node to only select one drone, that is Use β k,m ∈{0,1} represents node d k The task is finally completed on which ground server or drone, β k,0 =1 means the task is calculated on the drone, β k,m =1,m∈{1,2,…,M} represents the ground server a m Calculate on,

[0008] Step 3: Minimize the node d that takes the longest time k The calculation delay of the task is optimized, and the position of the UAV is Node offloading decision α k,n and β k,m , computing resource allocation strategy for drones and servers and And the communication resource allocation strategy between drones and servers and To optimize the variables, a mathematical model of the optimization problem P is established;

[0009] Step 4: Use the K-Means clustering algorithm and cross iteration to solve the problem P in step 3 and obtain the value of the optimized variable in the above step 3.

[0010] Furthermore, in step 3, node d k The computational latency of the above task is:

[0011]

[0012] in, Represents node d k To UAV n The uplink transmission delay is is the uplink transmission rate, Indicates drone u n Assigned to node d kbandwidth resources, is the transmission power, is the uplink channel gain, ρ 0 It indicates the receiving power at a transmission power of 1W and a reference distance of 1m. represents the Euclidean distance of the uplink, σ 2 is the noise power; Represents task W k In drone u n The computational delay of Indicates drone u n Assigned to node d k computing resources; Indicates drone u n To ground base station a m The downlink transmission delay between and Respectively represent drone u n The downlink bandwidth and the transmission power of the UAV, is the downlink channel gain; is the computational delay of the ground base station, Indicates base station a m Assigned to node d k The computing resources for the task.

[0013] Furthermore, in step 3, the constraint condition of problem P is:

[0014]

[0015]

[0016]

[0017]

[0018]

[0019]

[0020] here represents the maximum computing resource of the UAV, B UL represents the maximum bandwidth that can be allocated to the UAV, B DL Indicates the maximum bandwidth that can be allocated to each base station server. Indicates the maximum computing resources of each base station server.

[0021] Furthermore, in step 4, the steps of solving the problem P in step 3 by combining the K-Means clustering algorithm and cross iteration are as follows:

[0022] Step 4.1: Use the K-Means clustering algorithm to cluster the coordinates of the ground IoT nodes using Euclidean distance to obtain the location of the drone

[0023] Step 4.2: Each node selects the nearest drone to unload and initializes α k,n ;

[0024] Step 4.3: Traverse different unloading schemes for each node α k,n ,β k,m ;

[0025] Step 4.4: Allocate communication bandwidth resources based on the principle of equal communication delay when multiple nodes transmit tasks to the same drone Allocate communication bandwidth resources based on the principle of equal latency when offloading tasks from multiple drones to the same ground server Analogous computing resources and

[0026] Step 4.5: Using the above α k,n , β k,m , value, calculate the computational completion delay of each node task, and take the value T of the node with the longest computational time k is the target value;

[0027] Step 4.6: Repeat steps 4.3-4.5, take the smallest target value, until the target value no longer changes, and write down the corresponding α k,n , β k,m , The value is the solution to the problem P.

[0028] The beneficial effects of the present invention are mainly manifested in: the present invention is applicable to drone-assisted edge computing scenarios, and achieves the goal of minimizing computing offloading delay by jointly optimizing drone location deployment, node offloading strategy, and bandwidth and resource allocation strategy between drones and servers. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Schematic diagram of the drone-assisted edge computing model of the present invention.

[0030] Figure 2 It is the algorithm flow chart of K-Means of the present invention. DETAILED DESCRIPTION

[0031] The present invention will be further described below in conjunction with the accompanying drawings.

[0032] Combination Figure 1, Figure 2 , a method for minimizing time delay of edge computing assisted by a drone, the method comprising the following steps:

[0033] Step 1: Set up specific application scenarios, such as Figure 1 The system is equipped with K IoT nodes, N drones, and M ground base station servers, where D = {d 1 ,d 2 ,…,d K} represents the set of IoT nodes, U = {u 1 ,u 2 ,…,u N} represents the set of drones, A={a 1 ,a 2 ,…,a M} represents a set of M base stations. All IoT nodes and base station servers are deployed on the ground, and the location of IoT nodes is represented by Indicates that the location of the base station server is All drones are deployed at the same height H, and the positions of the drones are represented by express.

[0034] Step 2: Each IoT device generates a computing task W k =(C k ,F k ), the two parameters represent the task data size and the number of CPU cycles required to process one bit of task data. Each task can be offloaded to the drone for calculation, or unloaded to the base station server for calculation using the drone as a relay. k,n ∈{0,1} represents node d k Do you choose drone? n Uninstall, 1 for selection, 0 for non-selection, set each IoT node to only select one drone, that is Use β k,m ∈{0,1} represents node d k The task is finally completed on which ground server or drone, β k,0 =1 means the task is calculated on the drone, β k,m =1,m∈{1,2,…,M} represents the ground server a m Calculate on,

[0035] Step 3: Minimize the node d that takes the longest time k The calculation delay of the task is optimized, and the position of the UAV is Node offloading decision α k,n and β k,m , computing resource allocation strategy for drones and servers and And the communication resource allocation strategy between drones and servers and To optimize the variables, a mathematical model of the optimization problem P is established;

[0036] Step 4: Use the K-Means clustering algorithm and cross iteration to solve the problem P in step 3 and obtain the value of the optimized variable in the above step 3.

[0037] Furthermore, in step 3, node d k The computational latency of the above task is:

[0038] in, Represents node d k To UAV n The uplink transmission delay is is the uplink transmission rate, Indicates drone u n Assigned to node d k bandwidth resources, is the transmission power, is the uplink channel gain, ρ 0 It indicates the receiving power at a transmission power of 1W and a reference distance of 1m. represents the Euclidean distance of the uplink, σ 2 is the noise power; Represents task W k In drone u n The computational delay of Indicates drone u n Assigned to node d k computing resources; Indicates drone u n To ground base station a m The downlink transmission delay between and Respectively represent drone u n The downlink bandwidth and the transmission power of the UAV, is the downlink channel gain; is the computational delay of the ground base station, Indicates base station a m Assigned to node d k The computing resources for the task.

[0039] Furthermore, in step 3, the constraint condition of problem P is:

[0040]

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] here represents the maximum computing resource of the UAV, B UL Indicates the maximum bandwidth that can be allocated to the drone, the maximum bandwidth that can be allocated to each base station server, and the maximum computing resources of each base station server. C1 indicates that each IoT node can only select one drone, C2 indicates that the computing task can only be offloaded to the drone or one of the ground base station servers for computing, C3 indicates that the resources allocated to the node task by the drone cannot exceed its own maximum computing resources, C4 indicates that the bandwidth allocated to each node by the drone cannot exceed its own maximum bandwidth, C5 indicates that the resources allocated to the node task by the base station server cannot exceed its own maximum computing resources, and C6 indicates that the downlink bandwidth of the drone cannot exceed the maximum bandwidth allocated to it by the server.

[0047] Furthermore, in step 4, the steps of solving the problem P in step 3 by combining the K-Means clustering algorithm and cross iteration are as follows:

[0048] Step 4.1: Figure 2 As shown in the figure, the K-Means clustering algorithm is used to cluster the coordinates of the ground IoT nodes with Euclidean distance to obtain the location of the drone.

[0049] Step 4.2: Each node selects the nearest drone to unload and initializes α k,n ;

[0050] Step 4.3: Traverse different unloading schemes for each node α k,n ,β k,m ;

[0051] Step 4.4: Allocate communication bandwidth resources based on the principle of equal communication delay when multiple nodes transmit tasks to the same drone Allocate communication bandwidth resources based on the principle of equal latency when offloading tasks from multiple drones to the same ground server Analogous computing resources and

[0052] Step 4.5: Using the above α k,n , β k,m , value, calculate the computational completion delay of each node task, and take the value T of the node with the longest computational time k is the target value;

[0053] Step 4.6: Repeat steps 4.3-4.5, take the smallest target value, until the target value no longer changes, and write down the corresponding α k,n , β k,m , The value is the solution to the problem P.

[0054] For example Figure 1 The application scenario of K IoT nodes, N drones, and M ground base station servers is used to illustrate the specific implementation scheme of the present invention.

[0055] First, set up K IoT nodes, N drones, and M ground base station servers. All IoT nodes and base station servers are deployed on the ground, and drones are deployed at the same height H.

[0056] Second, each IoT device generates a computing task W k ,Each task can be offloaded to the drone for calculation, or it can be offloaded to the base station server for calculation using the drone as a relay;

[0057] Then, the mathematical model of the optimization problem P is established with minimizing the computational delay of the task that takes the most time as the optimization goal, and the location of the UAV, the node offloading decision, the computing resource allocation strategy of the UAV and the server, and the communication resource allocation strategy of the UAV and the server as the optimization variables;

[0058] Then, the problem P in step 3 is solved by combining the K-Means clustering algorithm and cross iteration.

[0059] Finally, the best solution is deployed in the system.

[0060] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for minimizing the time delay of edge computing assisted by a drone, characterized in that: The steps include: Step 1: Set up a specific application scenario; set up K IoT nodes, N drones, and M ground base station servers in the system, where D = {d1, d2, …, d K } represents the set of IoT nodes, U = {u1,u2,…,u N } represents the set of drones, A={a1,a2,…,a M } represents a set of M base stations; all IoT nodes and base station servers are deployed on the ground, and the location of IoT nodes is represented by Indicates that the location of the base station server is Indicates; all drones are deployed at the same height H, and the positions of the drones are represented by express; Step 2: Each IoT device generates a computing task W k =(C k ,F k ), the two parameters represent the task data size and the number of CPU cycles required to process one bit of task data respectively; each task can be offloaded to the drone for calculation, or unloaded to the base station server for calculation using the drone as a relay; α k,n ∈{0,1} represents node d k Do you choose drone? n Uninstall, 1 for selection, 0 for non-selection, set each IoT node to only select one drone, that is Use β k,m ∈{0,1} represents node d k The task is finally completed on which ground server or drone, β k,0 =1 means the task is calculated on the drone, β k,m =1,m∈{1,2,…,M} represents the ground server a m Calculate on, Step 3: Minimize the node d that takes the longest time k The calculation delay of the task is optimized, and the position of the UAV is Node offloading decision α k,n and β k,m , computing resource allocation strategy for drones and servers and And the communication resource allocation strategy between drones and servers and To optimize the variables, a mathematical model of the optimization problem P is established; Node d k The computational latency of the above task is: in, Represents node d k To UAV n The uplink transmission delay is is the uplink transmission rate, Indicates drone u n Assigned to node d k bandwidth resources, is the transmission power, is the uplink channel gain, ρ0 represents the received power at a reference distance of 1m when the transmission power is 1W, represents the Euclidean distance of the uplink, σ 2 is the noise power; Represents task W k In drone u n The computational delay of Indicates drone u n Assigned to node d k computing resources; Indicates drone u n To ground base station a m The downlink transmission delay between and Respectively represent drone u n The downlink bandwidth and the transmission power of the UAV, is the downlink channel gain; is the computational delay of the ground base station, Indicates base station a m Assigned to node d k The computing resources for the task; The constraints of problem P are: here represents the maximum computing resource of the UAV, B UL represents the maximum bandwidth that can be allocated to the UAV, B DL Indicates the maximum bandwidth that can be allocated to each base station server. Indicates the maximum computing resources of each base station server; Step 4: Use the K-Means clustering algorithm and the cross-iteration method to solve the problem P in step 3 and obtain the value of the optimized variable in step 3. The specific solution steps are: Step 4.1: Use the K-Means clustering algorithm to cluster the coordinates of the ground IoT nodes using Euclidean distance to obtain the location of the drone Step 4.2: Each node selects the nearest drone to unload and initializes α k,n ; Step 4.3: Traverse different unloading schemes for each node α k,n ,β k,m ; Step 4.4: Allocate communication bandwidth resources based on the principle of equal communication delay when multiple nodes transmit tasks to the same drone Allocate communication bandwidth resources based on the principle of equal latency when offloading tasks from multiple drones to the same ground server Analogous computing resources and Step 4.5: Using the above α k,n , β k,m , value, calculate the computational completion delay of each node task, and take the value T of the node with the longest computational time k is the target value; Step 4.6: Repeat steps 4.3-4.5, take the smallest target value, until the target value no longer changes, and write down the corresponding α k,n , β k,m , The value is the solution to the problem P.

Images