Satellite-ground fusion end-side cloud collaborative computing unloading method

Abstract

The invention discloses a satellite-ground fusion end-side cloud collaborative computing unloading method, and belongs to the technical field of wireless communication. In order to solve the problem of high task completion time delay caused by large ground processing time delay and limited satellite processing capability, the invention provides an end-side cloud collaborative computing unloading method through marginal computing. In the task modeling stage, modeling is carried out on the split calculation sub-task dependency relationship in combination with a linked list and a directed acyclic graph, and a generated task model can effectively reduce task processing delay. In the calculation unloading stage, task completion time delay is defined according to a node unloading strategy and a link transmission rate, and a time delay list based on task unloadable nodes is calculated and generated by adopting a complete unloading strategy; under the constraint of satellite remaining coverage time, node computing power and link unloading bandwidth, a real number coding genetic algorithm based on task completion time delay is provided to find an optimal unloading point of a computing task, and a formed optimal computing unloading strategy can effectively reduce time delay and improve the task completion rate.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology, and specifically relates to a satellite-ground fusion end-edge-cloud collaborative computing offloading method. Background Art

[0002] With the rapid development of information technology, the 6G-oriented satellite-ground integrated end-edge-cloud collaborative network, as a new type of communication architecture, is gradually attracting widespread attention. The satellite-ground integrated end-edge-cloud collaborative network forms an efficient, flexible and intelligent communication system by organically integrating satellite communications, ground networks, and edge-cloud collaboration. Among them: End-edge-cloud collaboration refers to the integration of communication, computing, and storage resources of terminals, edge computing nodes, and cloud data centers, as well as the reasonable allocation of tasks between terminals, edge devices, and the cloud. However, with the rise of IoT technology, the popularization of wireless networks, and the increasing number of user devices, the surge in real-time tasks will lead to increased system latency. Therefore, how to reasonably allocate processing tasks between edge computing nodes and cloud computing centers has become a difficult problem that needs to be solved urgently.

[0003] According to existing research on computing offloading, it can be divided into three categories according to the offloading scenario: (1) single-server single-user scenario, (2) single-server multi-user scenario, and (3) multi-user multi-server scenario. The first two scenarios are only suitable for simple applications, scenarios with high security requirements, or scenarios with relatively low resource requirements but requiring flexible expansion; multi-server multi-user is suitable for large-scale applications or scenarios that require high availability and performance optimization. Most of these studies offload tasks to specific edge servers for execution or only utilize the resources of the edge network, which does not fully utilize the resources in the dense edge network and remote cloud, resulting in low device resource utilization. Therefore, when studying task offloading, it is necessary to consider the computing scenario of cloud-edge collaboration.

[0004] In terms of task models, they can be mainly divided into the following two types: one type mainly considers the coarse-grained scope of tasks, almost without considering the dependencies of tasks, and places the tasks as a whole locally or offloads them to the server for execution; the other type considers the dependencies between tasks. However, neither of the above two types considers the dependencies between subtasks when the task is completely offloaded to a single task node. Therefore, the research on the offloading strategy of this task model needs further exploration.

[0005] In order to improve resource utilization and reduce task completion delay, the present invention proposes a satellite-ground integrated end-edge-cloud collaborative computing offloading method. In the task modeling stage, a functional module subdivision strategy is used to split the tasks, define the sequence, selection and parallel dependencies, and combine the linked list with the directed acyclic graph to model the task dependencies to determine the subtask priority. In the task offloading stage, a complete offloading method is adopted. According to the distance between different nodes, the transmission rate, and the computing power of edge computing, cloud computing and satellite offloading nodes, the propagation, transmission and processing delays of different tasks on different offloading nodes are calculated to generate a task completion delay list. Under the multiple constraints of communication, computing resources and satellite coverage time, a real number coded genetic algorithm is used to solve the optimal offloading strategy. Summary of the invention

[0006] The present invention aims to solve the above problems of the prior art. A satellite-ground fusion end-edge-cloud collaborative computing offloading method is proposed. The method considers the timing and data dependency between tasks, and the joint offloading of satellite, edge computing and cloud computing. The computing offloading decision method is obtained by using a real number coded genetic algorithm. The technical solution of the present invention is as follows:

[0007] A satellite-ground integrated end-edge-cloud collaborative computing offloading method comprises the following steps:

[0008] S1: Build a satellite-ground integrated end-edge-cloud collaborative network model to calculate the transmission rates of ground and satellite-ground links;

[0009] S2: Use functional module segmentation strategy to split the task into several subtasks;

[0010] S3: Define sequential, parallel and selective dependencies, combine linked lists with directed acyclic graphs, and build serial, parallel and hybrid task dependency models;

[0011] S4: Define serial, parallel and hybrid task completion delays based on node offloading strategy and link transmission rate;

[0012] S5: Calculate the task completion delay of edge computing, cloud computing and satellite nodes using a complete offloading strategy, and generate a task completion delay list based on task offloading nodes;

[0013] S6: Calculate the satellite coverage arc length and remaining coverage time according to the distance and pitch angle between the user and the satellite;

[0014] S7: Under the multiple constraints of satellite remaining coverage time, node computing power and link offloading bandwidth, a real-coded genetic algorithm based on task completion delay is proposed to find the optimal offloading point of the computing task, and then form a computing offloading strategy that minimizes the task completion delay.

[0015] Furthermore, the ground and satellite-to-ground link transmission rates calculated in step S1 are

[0016] (1) Ground network link

[0017] The link uses frequency division multiplexing, and the channel is modeled as a Rayleigh channel. User terminal i connects to ground edge computing j 1 and cloud computing servers 2 The link transmission rate between them is:

[0018]

[0019] in, are different link channel bandwidths, is the transmission power of the user equipment in different ground network links, N 0 is the power of the additive white Gaussian noise of the link, is the channel gain of different links, and its calculation formula is:

[0020]

[0021] in, is the channel coefficient, and its calculation formula is:

[0022]

[0023] in, is the Rayleigh random variable, γ is the path loss exponent;

[0024] (2) Satellite-to-ground uplink

[0025] The satellite-to-ground uplink uses the Ka band for communication. Line-of-sight transmission dominates the uplink. The channel is modeled as a Ricean channel. The link transmission rate between user terminal i and satellite m is:

[0026]

[0027] Among them, B i,m is the channel bandwidth of the satellite-to-ground uplink, P i,m is the transmit power of user terminal i in the uplink, W i is the antenna gain of user terminal i, G m is the receiving antenna gain of satellite m, and its calculation formula is:

[0028] G m ≈(4D RA f u / c) 2 (5) Among them, D RA is the diameter of the satellite receiving antenna, f u is the uplink carrier frequency, c is the speed of light;

[0029] Considering the impact of free path loss and rain attenuation on satellite signal transmission quality, the uplink channel coefficient is:

[0030]

[0031] Among them, ρ i,m is a Ricean random variable, As the rain fades, is the free path loss, which is calculated as:

[0032]

[0033] Furthermore, the sequential, parallel and selection dependency relationships in step S3 are:

[0034] (1) Sequential dependency: Only when the current subtask is completed and data is transferred to the next subtask, can the next subtask continue to execute;

[0035] (2) Parallel dependency relationship: A task has multiple parallel subtasks, and the execution of subtasks does not interfere with each other;

[0036] (3) Select dependency row relationship: During the execution of a task, after the previous subtask is completed, the next subtask to be executed may have a choice between two or more subtasks.

[0037] Furthermore, the task dependency model in step S3 is:

[0038] The tasks generated by the user terminal are represented by a set, namely u y is the number of tasks; the subtasks divided by the user terminal are represented by a set, namely u s represents the number of task subtasks; the task model is modeled as a four-tuple to characterize the task requirements, that is, y k =(S k ,ξ k ,w k ,q k );q k represents the amount of data of the kth task itself, w k represents the amount of computation required to process the kth task; at the same time, Represents the dependency relationship between subtasks. , indicating a subtask and There is no dependency between , indicating a subtask and There is a timing, data, or dual timing and data dependency between the subtasks; based on whether there is a timing, data, or dual timing and data dependency between the subtasks, a linked list and directed acyclic graph are combined to construct serial, parallel, and hybrid task dependency models.

[0039] Furthermore, the completion delay of the serial, parallel and hybrid tasks in step S4 is:

[0040] The propagation delay, transmission delay, and processing delay are:

[0041]

[0042] Where f represents the computing power of edge computing, cloud computing, and satellite servers, that is, the number of CPU cycles that the server can provide, R represents the transmission rate in different links, d represents the distance from the user terminal to the server, and q k Represents task y k The amount of data itself, w k Represents processing task y k The amount of calculation required; c represents the propagation speed of the signal in the optical fiber; R represents the transmission rate in different links.

[0043] The completion delay of a single task is:

[0044] T=T tran +T prop +T proc (11)

[0045] The completion delay of the kth serial task is:

[0046]

[0047] The completion delay of the kth parallel task is

[0048]

[0049] in, represents the maximum amount of computation in task k, Indicates the maximum amount of data in task k;

[0050] The completion delay of the kth mixed task should be:

[0051]

[0052] in, Represented as task y k The amount of computation required for the nth subtask is the number of cycles required for the computation task. ∩ represents the intersection. Represents task y kThe subtask with the largest computational complexity and dependency relationship, u s Represents task y k Number of neutron tasks.

[0053] Furthermore, the completion delay of the serial, parallel and hybrid tasks offloaded to edge computing, cloud computing and satellite in step S5 is:

[0054] Task y k Uninstall strategy vector To express, Represents task y k Binary decision variables for offloading to ground edge computing servers, cloud computing servers, satellites, and on-board edge computing servers. The user terminal offloads the task to the ground edge server; if Indicates that the user terminal offloads the task to the cloud computing server; if Indicates that the user terminal offloads the task to the satellite; if Indicates that the user terminal offloads the task to the on-board edge computing server;

[0055] (1) The completion delay of serial, parallel and hybrid tasks offloaded to the ground edge computing server is:

[0056]

[0057] in, The computing power of the ground edge server, is the distance between the user terminal and the ground edge computing server, is the transmission rate between the user terminal and the ground edge computing server;

[0058] (2) The completion delay of serial, parallel and hybrid tasks offloaded to the cloud computing server is:

[0059]

[0060] in, The computing power of cloud computing servers. is the distance between the user terminal and the cloud computing server, is the transmission rate between the user terminal and the cloud computing server;

[0061] (3) The completion delay of serial, parallel and hybrid tasks unloaded to the satellite is:

[0062]

[0063] Among them, f m is the computing power of the satellite, d i,mis the distance between the user terminal and the satellite, R i,m is the transmission rate between the user terminal and the satellite;

[0064] (4) When the satellite cannot process due to limited computing resources, part of the edge computing is carried out on the satellite. Therefore, the completion delay of serial, parallel and hybrid tasks offloaded to the on-board edge server is:

[0065]

[0066] in, The computing power of the on-board edge computing server.

[0067] Furthermore, the remaining satellite coverage time and coverage arc length in step S6 are:

[0068] (1) Based on the relative position of the satellite and the user terminal, the elevation angle θ between the user terminal and the satellite i,m It is expressed as:

[0069]

[0070] Among them, h m Represents the height between the user equipment and the satellite orbit, R e represents the radius of the Earth, d i,m represents the distance between the user terminal and the satellite, It represents the angle between the satellite and the straight line passing through the center of the earth and perpendicular to the tangent line of the user terminal;

[0071] (2) Geocentric angle ζ corresponding to the satellite coverage area 1 It is expressed as:

[0072]

[0073] (3) The longest communication time between user terminal i and the satellite is expressed as:

[0074]

[0075] Among them, v i,m is the speed of the LEO satellite, L is the arc length of the satellite's coverage of the user equipment, and the calculation formula is:

[0076]

[0077] Furthermore, the constraint for minimizing the task completion delay in step S7 should be:

[0078]

[0079] in, They represent the computing power of ground edge computing, cloud computing and satellite edge computing servers, respectively. m represents the computing power of the satellite, represents the offload bandwidth of the links between the user terminal and the ground edge computing, cloud computing and on-board edge computing servers, respectively, and B max is the total offloading bandwidth, C1 is the satellite coverage time constraint, C2, C3, C4 and C5 are the node computing capacity constraints, C6, C7, C8, C9, C10 and C11 are the offloading bandwidth constraints, and C12 is the offloading decision variable constraint.

[0080] The advantages and beneficial effects of the present invention are as follows:

[0081] The solution of the present invention aims at the problem of high task completion delay caused by limited satellite processing capacity and large ground processing delay in satellite-ground fusion communication, and proposes a computational unloading method for satellite-ground fusion end-edge-cloud collaboration. The main innovations of the present invention are: 1) The functional module subdivision strategy is adopted to split the task, define the sequence, selection and parallel dependency relationships, and combine the linked list with the directed acyclic graph to model the task dependency to determine the subtask priority; 2) According to the node unloading strategy and link transmission rate, the serial, parallel and hybrid task completion delays are defined, and the task completion delays of edge computing, cloud computing and satellite nodes are calculated using a complete unloading strategy, and a task completion delay list based on task unloading nodes is generated. 3) Under the multiple constraints of the remaining coverage time of the satellite, the computing power of the node and the link unloading bandwidth, a real number coded genetic algorithm based on the task completion delay is proposed to find the optimal unloading point of the computing task, thereby forming a computational unloading strategy that minimizes the task completion delay. In existing studies, most of the tasks are divided arbitrarily, resulting in an overly idealized task model. Therefore, the present invention has the characteristics of creativity and easy implementation in terms of solution means. Furthermore, the present invention analyzes whether the signal bypasses obstacles through reflection, diffraction or scattering, models the channel as Rayleigh and Rice channels, and comprehensively considers the influencing factors of path loss and rain attenuation to calculate the transmission rate in different links. In existing studies, most of them consider the impact of a single factor on the transmission rate. Therefore, the present invention is unique and practical. Finally, the present invention also adopts a complete offloading method to offload tasks to edge computing, cloud computing and satellite nodes for processing. In existing studies, most of them consider a layered partial offloading method. Although this method does not require the computing power of edge computing, cloud computing and satellite nodes, the resulting data migration will increase the propagation and transmission delays. Therefore, the present invention is reasonable and creative. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a network model of satellite-ground fusion end-edge-cloud collaboration constructed by the preferred embodiment provided by the present invention;

[0083] Figure 2It is a task model diagram;

[0084] Figure 3 It is a diagram of the relative positions of the satellite and the user;

[0085] Figure 4 This is a schematic diagram of the computing offloading process of the satellite-ground integrated terminal-edge-cloud collaboration described in the present invention. DETAILED DESCRIPTION

[0086] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0087] The technical solution of the present invention to solve the above technical problems is:

[0088] A satellite-ground fusion end-edge-cloud collaborative computing offloading method is proposed to reduce the task completion delay. Firstly, the functional module subdivision strategy is adopted to split the task, define the sequential, parallel and selective dependencies, and combine the linked list with the directed acyclic graph to model the task dependency to determine the subtask priority; secondly, according to the node offloading strategy and link transmission rate, the serial, parallel and hybrid task completion delays are defined, and the complete offloading strategy is adopted to calculate the task completion delay of edge computing, cloud computing and satellite nodes, and generate a task completion delay list based on the task offloading nodes; finally, under the multiple constraints of the remaining satellite coverage time, node computing power and link offloading bandwidth, a real number coded genetic algorithm based on task completion delay is proposed to find the optimal offloading point of the computing task, thereby forming a computing offloading strategy that minimizes the task completion delay.

[0089] The specific process is as follows:

[0090] Step 1: Build a satellite-ground integrated end-edge-cloud collaborative network model to calculate the transmission rates of the ground and satellite-ground links.

[0091] Step 2: Use the functional module segmentation strategy to split the task into several subtasks.

[0092] Step 3: Define sequential, parallel, and selective dependencies, combine linked lists with directed acyclic graphs, and build serial, parallel, and hybrid task dependency models.

[0093] Step 4: Define the completion delay of serial, parallel and hybrid tasks according to the node offloading strategy and link transmission rate.

[0094] Step 5: Use the full offloading strategy to calculate the task completion delay of edge computing, cloud computing and satellite nodes, and generate a task completion delay list based on task offloadable nodes.

[0095] Step 6: Based on the distance and elevation angle between the user and the satellite, calculate the arc length of the satellite's coverage of the user and the remaining coverage time.

[0096] Step 7: Under the multiple constraints of satellite remaining coverage time, node computing power and link offloading bandwidth, a real-coded genetic algorithm based on task completion delay is proposed to find the optimal offloading point of the computing task, and then form a computing offloading strategy that minimizes the task completion delay.

[0097] Preferably, the third step is to define sequential, parallel and selection dependencies, as follows:

[0098] (1) Sequential dependency: Only when the current subtask is completed and data is transferred to the next subtask, can the next subtask continue to execute.

[0099] (2) Parallel dependency relationship: A task has multiple parallel subtasks, and the execution of subtasks does not interfere with each other.

[0100] (3) Select dependency row relationship: During the execution of a task, after the previous subtask is completed, the next subtask to be executed may have a choice between two or more subtasks.

[0101] Preferably, in the fourth step, according to the node offloading strategy and the link transmission rate, the serial, parallel and hybrid task completion delays are defined, and the formula is as follows:

[0102] When a user terminal generates a computing task, the user can choose to offload the task to different servers, which will then process the task. In this process, the propagation delay, transmission delay, and processing delay are:

[0103]

[0104] Among them, f represents the computing power of edge computing, cloud computing, and satellite servers, that is, the number of CPU cycles that the server can provide, c represents the propagation speed of the signal in the optical fiber, R represents the transmission rate in different links, d represents the distance from the user terminal to the server, and q represents the distance from the user terminal to the server. k represents the amount of data of task k itself, w k Represents the amount of computation required to process task k.

[0105] Generally speaking, the calculation result is small for the original task data, so the time to transmit the calculation result to the user is often ignored. For the convenience of analysis, this article does not consider the queuing delay of the task in the processing process, so the completion delay of a single task is:

[0106] T=T tran +T prop +T proc (35)

[0107] According to the above task model, there may be sequential, parallel, and selection dependencies between subtasks. Different dependencies lead to different execution orders of subtasks, resulting in different task completion delays. At this time, the completion delay of the kth serial task should be:

[0108]

[0109] The completion delay of the kth parallel task should be

[0110]

[0111] in, represents the maximum amount of computation in task k, Indicates the maximum amount of data in task k.

[0112] The completion delay of the kth mixed task should be:

[0113]

[0114] in, It is represented by the amount of computation required for the kth subtask in the task, that is, the number of cycles required for the computation task. ∩ represents the intersection. Represents task y k The subtask with the largest computational complexity and dependency relationship, u s Represents task y k Number of neutron tasks.

[0115] The models involved in the present invention are as follows:

[0116] 1. Network Model

[0117] The network model of satellite-ground fusion and end-edge-cloud collaboration is as follows Figure 1 As shown in the figure, the model includes user terminals, ground and satellite networks. Users mainly include various sensors and smart phones, which are used for data collection, task generation, task division and task reconstruction. The ground network is mainly composed of edge computing and cloud computing, which are jointly responsible for processing a large number of ground tasks and providing users with low-latency and intelligent services. The satellite network is mainly composed of satellites and on-board edge servers. Putting edge computing on the satellite can expand on-board computing resources. When the computing power of the ground infrastructure is insufficient, the satellite network provides computing services to users with its advantages of wide coverage, large capacity and high flexibility, effectively supporting the computing needs of various user businesses.

[0118] 2. The technical solution of the present invention is as follows:

[0119] The present invention proposes a satellite-ground fusion end-edge-cloud collaborative computing offloading method to reduce the task completion delay. First, the functional module subdivision strategy is adopted to split the task, define the sequential, parallel and selective dependencies, and combine the linked list with the directed acyclic graph to model the task dependency to determine the subtask priority; secondly, according to the node offloading strategy and link transmission rate, the serial, parallel and hybrid task completion delays are defined, and the complete offloading strategy is adopted to calculate the task completion delay of edge computing, cloud computing and satellite nodes, and generate a task completion delay list based on task offloadable nodes; finally, under the multiple constraints of the remaining satellite coverage time, node computing power and link offloading bandwidth, a real number coded genetic algorithm based on task completion delay is proposed to find the optimal offloading point of the computing task, thereby forming a computing offloading strategy that minimizes the task completion delay.

[0120] 3. Objective function and constraints:

[0121]

[0122] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions.

[0123] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0124] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A satellite-ground fusion end-edge-cloud collaborative computing offloading method, characterized in that: The following steps are involved: S1: Build a satellite-ground integrated end-edge-cloud collaborative network model to calculate the transmission rates of ground and satellite-ground links; S2: Use functional module segmentation strategy to split the task into several subtasks; S3: Define sequential, parallel and selective dependencies, combine linked lists with directed acyclic graphs, and build serial, parallel and hybrid task dependency models; S4: Define the completion delay of serial, parallel and hybrid tasks according to the node offloading strategy and link transmission rate; S5: Calculate the task completion delay of edge computing, cloud computing and satellite nodes using a complete offloading strategy, and generate a task completion delay list based on task offloading nodes; S6: Calculate the satellite coverage arc length and remaining coverage time according to the distance and pitch angle between the user and the satellite; S7: Under the multiple constraints of satellite remaining coverage time, node computing power and link offloading bandwidth, a real-coded genetic algorithm based on task completion delay is proposed to find the optimal offloading point of the computing task, and then form a computing offloading strategy that minimizes the task completion delay.

2. According to claim 1, a satellite-ground fusion end-edge-cloud collaborative computing offloading method is characterized in that: In step S1, the ground and satellite-to-ground link transmission rates are calculated as follows: (1) Ground network link The link uses frequency division multiplexing, and the channel is modeled as a Rayleigh channel. User terminal i connects to ground edge computing j 1 and cloud computing servers 2 The link transmission rate between them is: in, are different link channel bandwidths, is the transmission power of the user equipment in different ground network links, N0 is the power of the additive white Gaussian noise of the link, is the channel gain of different links, and its calculation formula is: in, is the channel coefficient, and its calculation formula is: in, is the Rayleigh random variable, γ is the path loss exponent; (2) Satellite-to-ground uplink The satellite-to-ground uplink uses the Ka band for communication. Line-of-sight transmission dominates the uplink. The channel is modeled as a Ricean channel. The link transmission rate between user terminal i and satellite m is: Among them, B i,m is the channel bandwidth of the satellite-to-ground uplink, P i,m is the transmit power of user terminal i in the uplink, W i is the antenna gain of user terminal i, G m is the receiving antenna gain of satellite m, and its calculation formula is: G m ≈(4D RA f u / c) 2 (5) Among them, D RA is the diameter of the satellite receiving antenna, f u is the uplink carrier frequency, c is the speed of light; Considering the impact of free path loss and rain attenuation on satellite signal transmission quality, the uplink channel coefficient is: Among them, ρ i,m is a Ricean random variable, As the rain fades, is the free path loss, which is calculated as:

3. According to claim 1, a satellite-ground fusion end-edge-cloud collaborative computing offloading method is characterized in that: The sequential, parallel and selection dependency relationships in step S3 are: (1) Sequential dependency: Only when the current subtask is completed and data is transferred to the next subtask, can the next subtask continue to execute; (2) Parallel dependency relationship: A task has multiple parallel subtasks, and the execution of subtasks does not interfere with each other; (3) Select dependency row relationship: During the execution of a task, after the previous subtask is completed, the next subtask to be executed may have a choice between two or more subtasks.

4. According to claim 1, a satellite-ground fusion end-edge-cloud collaborative computing offloading method is characterized in that: The task dependency model in step S3 is: The tasks generated by the user terminal are represented by a set, namely Y = {y1, y2, y3, …, yu y }, uy is the number of tasks; the subtasks divided by the user terminal are represented by a set, that is, u s represents the number of task subtasks; the task model is modeled as a four-tuple to characterize the task requirements, that is, y k =(S k ,ξ k ,w k ,q k );q k represents the amount of data of the kth task itself, w k represents the amount of computation required to process the kth task; at the same time, Represents the dependency between subtasks. , indicating a subtask and There is no dependency between , indicating a subtask and There is a timing, data, or dual timing and data dependency between the subtasks; based on whether there is a timing, data, or dual timing and data dependency between the subtasks, a linked list and directed acyclic graph are combined to construct serial, parallel, and hybrid task dependency models.

5. According to claim 1, a satellite-ground fusion end-edge-cloud collaborative computing offloading method is characterized in that: The completion delay of the serial, parallel and mixed tasks in step S4 is: The propagation delay, transmission delay, and processing delay are: Where f represents the computing power of edge computing, cloud computing, and satellite servers, that is, the number of CPU cycles that the server can provide, R represents the transmission rate in different links, d represents the distance from the user terminal to the server, and q k Represents task y k The amount of data itself, w k Represents processing task y k The amount of calculation required; c represents the propagation speed of the signal in the optical fiber; R represents the transmission rate in different links; The completion delay of a single task is: T=T tran +T prop +T proc (11) The completion delay of the kth serial task is: The completion delay of the kth parallel task is in, Represents task y k The maximum amount of calculation in Represents task y k The maximum amount of data in the The completion delay of the kth mixed task should be: in, Represented as task y k The amount of computation required for the nth subtask is the number of cycles required for the computation task. ∩ represents the intersection. Represents task y k The subtask with the largest computational complexity and dependency relationship, u s Represents task y k Number of neutron tasks.

6. According to claim 5, a satellite-ground fusion end-edge-cloud collaborative computing offloading method is characterized in that: The completion delay of the serial, parallel and hybrid tasks offloaded to edge computing, cloud computing and satellite in step S5 is: Task y k Uninstall strategy vector To express, Represents task y k Binary decision variables for offloading to ground edge computing servers, cloud computing servers, satellites, and on-board edge computing servers. The user terminal offloads the task to the ground edge server; if Indicates that the user terminal offloads the task to the cloud computing server; if Indicates that the user terminal offloads the task to the satellite; if Indicates that the user terminal offloads the task to the on-board edge computing server; (1) The completion delay of serial, parallel and hybrid tasks offloaded to the ground edge computing server is: in, The computing power of the ground edge server, is the distance between the user terminal and the ground edge computing server, is the transmission rate between the user terminal and the ground edge computing server; (2) The completion delay of serial, parallel and hybrid tasks offloaded to the cloud computing server is: in, The computing power of cloud computing servers. is the distance between the user terminal and the cloud computing server, is the transmission rate between the user terminal and the cloud computing server; (3) The completion delay of serial, parallel and hybrid tasks unloaded to the satellite is: Among them, f m is the computing power of the satellite, d i,m is the distance between the user terminal and the satellite, R i,m is the transmission rate between the user terminal and the satellite; (4) When the satellite cannot process due to limited computing resources, part of the edge computing is carried out on the satellite. Therefore, the completion delay of serial, parallel and hybrid tasks offloaded to the on-board edge server is: in, The computing power of the on-board edge computing server.

7. The method for offloading computing in satellite-ground fusion end-edge-cloud collaboration according to claim 6 is characterized in that: The remaining satellite coverage time and coverage arc length in step S6 are: (1) Based on the relative position of the satellite and the user terminal, the elevation angle θ between the user terminal and the satellite i,m It is expressed as: Among them, h m Represents the height between the user equipment and the satellite orbit, R e represents the radius of the Earth, d i,m represents the distance between the user terminal and the satellite, It represents the angle between the satellite and the straight line passing through the center of the earth and perpendicular to the tangent line of the user terminal; (2) The geocentric angle ζ1 corresponding to the satellite coverage area is expressed as: (3) The longest communication time between user terminal i and the satellite is expressed as: Among them, v i,m is the speed of the LEO satellite, L is the arc length of the satellite's coverage of the user equipment, and the calculation formula is:

8. The method for offloading computing in satellite-ground fusion terminal-edge-cloud collaboration according to claim 7 is characterized in that: The constraint for minimizing the task completion delay in step S7 should be: in, They represent the computing power of ground edge computing, cloud computing and on-board edge computing servers, respectively. m represents the computing power of the satellite, represents the offload bandwidth of the links between the user terminal and the ground edge computing, cloud computing and on-board edge computing servers, respectively, and B max is the total offloading bandwidth, C1 is the satellite coverage time constraint, C2, C3, C4 and C5 are the node computing capacity constraints, C6, C7, C8, C9, C10 and C11 are the offloading bandwidth constraints, and C12 is the offloading decision variable constraint.

Images