Air-space-ground integrated mobile edge computing offloading and resource allocation optimization method

Through the integrated mobile edge computing method of space and earth, the synergy between satellites and drones is used to optimize system delay and energy consumption, solving the problems of shortage of computing resources and degradation of delay performance in IoT devices, and providing efficient computing Services and resource allocation improves user experience.

CN118524446BActive Publication Date: 2024-11-01JILIN UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410968674.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-01
Estimated Expiration
2044-07-19

AI Technical Summary

Technical Problem

The prior art is difficult to effectively offload computing tasks in IoT devices, especially in areas with sparse deployment of base stations or no cellular network coverage, resulting in shortage of computing resources and degradation of latency performance, which cannot meet the task requirements of delay-sensitive and complex computing.

Method used

Using the integrated mobile edge computing method of space and earth, the system model of ground user terminals including satellites, drones and cellular networks is constructed, and the system delay is optimized using Lagrangian multiplication method, sequential convex optimization and genetic algorithm. and energy consumption, allocate computing resources and determine the drone trajectory and offload ratio, and realize the coordination and integration of multi-platform resources.

Benefits of technology

In areas without cellular network coverage or affected by natural disasters, efficient computing services are provided to reduce system delay and energy consumption, improve user terminal experience quality, and realize effective unloading of computing tasks and reasonable allocation of resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118524446B_ABST
    Figure CN118524446B_ABST
Patent Text Reader

Abstract

The present invention is applicable to the field of communication technology, and provides an air-ground integrated mobile edge computing unloading and resource allocation optimization method, comprising the following steps: S101: constructing a system model; S102: establishing an optimization problem according to the constructed system model; S103: calculating the unloading ratio of each user terminal according to the Lagrange multiplier method; S104: converting the constraints into convex constraints according to the sequential convex optimization principle, and obtaining the optimal drone trajectory and computing resource allocation ratio; S105: based on the scheme of associating user terminals with drones in each given user terminal group, repeatedly iteratively executing S103 and S104, obtaining the corresponding system delay and using the corresponding system delay as a fitness evaluation standard, and finally obtaining the optimal association scheme according to the genetic algorithm. The present invention can meet the computing requirements of user terminals in the system, effectively improve the system energy efficiency, expand the throughput, and improve the experience quality of user terminals in the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular to an air-ground-space integrated mobile edge computing offloading and resource allocation optimization method. Background Art

[0002] In recent years, the Internet of Things (IoT) and the fifth generation of mobile communication (5G) technologies have developed rapidly. Although new mobile devices are becoming more and more powerful in terms of central processing units (CPUs), they may not be able to complete applications with large amounts of data and low latency requirements in a short time. To address this problem, computing tasks are offloaded to cloud servers with sufficient computing power for computing, which has led to the development of mobile cloud computing (MCC). However, MCC servers are usually far away from users, which may increase network burden and introduce high latency. To address this problem, the concept of mobile edge computing (MEC) was born. In a MEC system, MEC servers have more powerful computing power than local devices and are closer to mobile devices. The distributed structure of MEC servers helps to alleviate traffic congestion in the core network. MEC is to harvest a large amount of idle computing power and storage space distributed on the edge of the network and use these resources on mobile devices to handle latency-sensitive or computationally complex tasks generated by mobile devices. However, it is difficult to guarantee the performance of IoT edge computing by simply relying on offloading from ground base stations. On the one hand, IoT devices are usually power-constrained and cannot support remote transmission of task offloading, especially when base stations are sparsely deployed or unavailable nearby (such as automatic mining applications). On the other hand, the physical computing resources on the base station are scarce and slightly insufficient, but IoT computing tasks may encounter sudden conditions (such as smart transportation applications), which will lead to a shortage of computing resources and degrade latency performance.

[0003] To address these issues, satellites and unmanned aerial vehicles (UAVs) are considered to be a powerful complement to enhance terrestrial networks. Satellites can provide ubiquitous services with acceptable propagation delays (e.g., about 6.44 milliseconds). For UAVs with flexible deployment and agile management, they have been widely used in military and civilian applications to provide on-demand communication and computing resources. In addition, the Third Generation Partnership Project (3GPP) is also studying non-terrestrial networks and specifying new architectures to complement terrestrial cellular networks. The Space-Ground Integrated Network (SAGIN), an integrated system of satellite systems, aeronautical networks, and terrestrial communications, is considered to be an emerging architecture that can provide ubiquitous computing services to users. In heterogeneous network models, how to coordinate multiple platforms and integrate computing and communication resources to provide services to users is a hot research direction now. In view of the above status quo, it is urgent to develop an air-ground-ground integrated mobile edge computing offloading and resource allocation optimization method to overcome the shortcomings in current practical applications. Summary of the invention

[0004] The purpose of the present invention is to provide an air-space-ground integrated mobile edge computing offloading and resource allocation optimization method, aiming to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The method for optimizing the offloading and resource allocation of air-space-ground integrated mobile edge computing includes the following steps:

[0007] S101: Build a system model; build a mobile edge computing system model including satellites, drones, and ground user terminals without cellular network coverage, and initialize the number of user terminals, the number of drones, the number of computing tasks of the user terminals, and the channel model parameters of the communication links between the user terminals and satellites and between the user terminals and drones;

[0008] S102: Based on the system model constructed in S101, the computing task processing delay of the user terminal in local computing, UAV mobile edge computing and satellite cloud computing and the corresponding system energy consumption are obtained, and finally the system delay is obtained. With the goal of minimizing the weighted sum of system delay and energy consumption, a specific optimization problem is established;

[0009] S103: Under the optimization problem, based on the principle of minimizing the weighted sum of system delay and energy consumption, the offloading ratio of each user terminal is calculated according to the Lagrange multiplier method;

[0010] S104: transforming the constraints into convex constraints according to the sequential convex optimization principle to obtain the optimal UAV trajectory and computing resource allocation ratio;

[0011] S105: Based on the scheme of associating user terminals with drones in each given user terminal group, S103 and S104 are repeatedly iterated to calculate the corresponding system delay and use the corresponding system delay as the fitness evaluation criterion. Finally, the optimal association scheme and the corresponding optimal unloading ratio, drone trajectory and computing resource allocation ratio are obtained according to the genetic algorithm.

[0012] Furthermore, in S101, a satellite and drone-assisted mobile edge computing system model is constructed as follows:

[0013] S1011: Set user terminal set , the drone collection is , the total CPU resources of the MEC server on the drone are f m , each terminal device has a latency-sensitive task to perform ,in D k is the task data size,C k is the computing resource required for the computing task, For mobile devices k The computing power of the local task , the number of tasks offloaded to the MEC server of drone m , the number of tasks offloaded to the ground cloud server via satellite transfer ; Divide the UAV flight cycle into N time slots, where the size of each time slot is δ t , the UAV is considered to remain stationary within a time slot;

[0014] S1012: Define the matching factor between user terminal k and drone m as a k,m , when the user terminal k matches the drone, a k,m =1, otherwise a k,m =0; p k,m is the transmission power from user terminal k to drone m, p k,s is the transmission power from user terminal k to the satellite, assuming that the fixed altitude of the drone is H, and the channel gain between user terminal k and drone m in the nth time slot is:

[0015] ;

[0016] in, g k,m is the small-scale fading coefficient, q m ( n ) is the position coordinate of the UAV at the nth time slot, ω k is the location coordinate of user terminal k, β 0 is the channel gain at a reference distance of 1m; according to the Shannon formula, the unloading rate between the user terminal k and the drone m at the nth time slot is:

[0017] ;

[0018] in, B is the bandwidth of the communication link between the user terminal and the UAV, is the noise power;

[0019] The channel gain of user terminal k unloaded to the satellite is:

[0020]

[0021] in, and are the gains of the user terminal transmitting antenna and the satellite receiving antenna, ψ is the rain attenuation coefficient, is the Boltzmann constant, c is the speed of light, f c is the carrier frequency, H s is the satellite altitude;

[0022] Then the unloading rate of user terminal k to the satellite is:

[0023] ;

[0024] in, B s is the bandwidth of the user terminal and satellite communication link, is the noise power.

[0025] Furthermore, in S102, according to the system model constructed in S101, the computing task processing delay of the user terminal in local computing, UAV mobile edge computing and satellite cloud computing and the corresponding system energy consumption are obtained, and finally the system delay is obtained, which is as follows:

[0026] The local calculation delay of user terminal k is:

[0027] ;

[0028] in, C It is the computing resources required to perform a single computing task on the user terminal;

[0029] The total delay of processing part of the computing task of user terminal k on the drone includes the offloading delay, computing delay and return delay, which is expressed as , the unloading delay is approximately expressed as an integer multiple of the number of time slots, the return delay is ignored, and the calculated delay is expressed as:

[0030] ;

[0031] The UAV mobile edge computing latency is expressed as:

[0032] ;

[0033] in, is the average unloading rate of tasks from user terminal k to drone m, is the computing resource allocated by UAV m to user terminal k;

[0034] The total delay of processing part of the computing task of user terminal k on the satellite includes the transmission delay from the user terminal to the satellite, the processing delay between the satellite and the ground station and the gateway, the round-trip propagation delay between the user terminal and the satellite, and the round-trip propagation delay between the satellite and the ground station. The satellite cloud computing delay is expressed as:

[0035] ;

[0036] in, is the processing delay between the satellite and the ground station and gateway, s is the distance between the satellite and the ground station;

[0037] For user terminal k, its delay is the maximum value among the local computing delay, the UAV mobile edge computing delay and the satellite cloud computing delay, expressed as ;

[0038] The final system delay is:

[0039] ;

[0040] The energy consumption of user terminals includes the energy consumption unloaded to satellites and drones and the transmission energy consumption, which can be expressed as:

[0041] ;

[0042] in, It is the energy consumption factor related to the user terminal’s own equipment;

[0043] The energy consumption of drones includes flight energy consumption and computing energy consumption. The computing energy consumption is expressed as:

[0044] ;

[0045] in, is the energy consumption factor related to the chip structure of the MEC server on the drone, is the time it takes for the UAV to calculate the computational task of user terminal k;

[0046] The flight energy consumption of the UAV in the nth time slot is expressed as:

[0047] ;

[0048] in, d 0 is the drag ratio of the drone body, ρ is the air density, R S It is the solidity of the drone's wind wheel. A is the rotor disk area of ​​the drone, v m ( n) is the flight speed of UAV m in the nth time slot, P 0 is the UAV blade profile power, is the angular velocity of the drone rotor, R e is the rotor radius of the drone, P i is the induced power of the drone blades, v 0 is the average rotor induced speed when the UAV is hovering;

[0049] Finally, the system energy consumption is expressed as:

[0050] ;

[0051] in, is the weighting coefficient, Less than 1;

[0052] Furthermore, according to the system delay and energy consumption expressions, the specific steps of establishing the optimization problem are as follows:

[0053] S1021: According to the expressions of system delay and energy consumption, an optimization target is established as the weighted sum of system delay and energy consumption, and the expression is: ;

[0054] S1022: A single user terminal can only offload tasks to one drone, with the following constraints:

[0055] ;

[0056] S1023: The computing tasks of the user terminal are processed by local computing, UAV mobile edge computing, and satellite cloud computing, and finally the equality constraint is obtained. ; The computing resource constraints of the drone are:

[0057] ;

[0058] S1024: Divide the flight cycle of the UAV in motion into a number of time slots, assume that the UAV is stationary in the time slots, and have different channel parameters between the user terminal and the UAV in different time slots;

[0059] S1025: Considering the user terminal association constraints, the UAV flight and computing resource constraints, and the user terminal delay constraints, the user terminal computing task allocation coefficient, the UAV trajectory and the association matching factor, and the computing resource allocation are optimized; finally, the optimization problem is obtained:

[0060] ;

[0061] st ;

[0062] ;

[0063] ;

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] in, is the normalization coefficient, V max is the maximum flight speed of the drone. and are the position coordinates of drone i and j at the nth time slot, d min is the minimum distance between drones, is the upper limit of the computing resources of UAV m, is the upper limit of the latency of the computing task of user terminal k.

[0069] Furthermore, in S103, the specific steps of obtaining the offloading ratio of each user terminal by using the Lagrange multiplier method are as follows:

[0070] S1031: Decouple the problem into k independent sub-problems for solution;

[0071] S1032: Assume that the local calculation delay is the largest and the number of time slots is N k , the user terminal delay is represented by the maximum local computing delay, and the range constraint of the number of tasks unloaded from the user terminal to the UAV is added. At the same time, the average unloading rate in different time slots is represented as the rate of unloading from the user terminal to the UAV; the optimization objective and corresponding constraints corresponding to the user terminal k are as follows:

[0072] ;

[0073] st ;

[0074] ;

[0075] ;

[0076] ;

[0077] S1033: The solution with the minimum objective function value is taken as the optimal unloading ratio solution for user terminal k, and the unloading ratio solutions for all user terminals are obtained in sequence.

[0078] Furthermore, in S104, the specific steps of obtaining the UAV trajectory and computing resource allocation ratio by sequential convex optimization are as follows:

[0079] S1041: First, relax the UAV transmission rate in the optimization target, so that the UAV mobile edge computing delay and UAV computing energy consumption in the optimization target are transformed into a convex form, and increase the relaxation variable ,get:

[0080] ;

[0081] Where:

[0082] ;

[0083] ;

[0084] in, is the unloading rate from user terminal k to drone m, , slack variables satisfy , is the variable in the first loop The value taken at the first-order Taylor expansion, In the first cycle The approximate value of

[0085] For the flight energy consumption part of the optimization objective, add relaxation variables:

[0086] ;

[0087] Right now ;

[0088] After the transformation, the flight energy consumption becomes convex:

[0089] ;

[0090] S1042: Then relax the distance constraint between the UAVs to make the non-convex constraint a linear constraint, and obtain:

[0091] ;

[0092] in, qi ( n )and q j ( n ) are the position coordinates of drone i and j at the nth time slot, and They are q i ( n )and q j ( n ) is the value of the first-order Taylor expansion at the lth cycle, is the relaxation of the distance between UAVs i and j at the nth time slot;

[0093] S1043: The converted optimization objectives and constraints are as follows:

[0094] ;

[0095] st. ;

[0096] ;

[0097] ;

[0098] ;

[0099] , ;

[0100] in, E m is the energy consumption of UAV m;

[0101] The interior point method of the convex optimization algorithm is used to solve the problem, and the UAV trajectory and computing resource allocation ratio in the current cycle are obtained. The solved result is used as the initial value and the convex optimization solution is continued until convergence, and the optimal UAV trajectory and computing resource allocation plan in the current cycle is obtained.

[0102] Furthermore, in S105, the specific steps of using the genetic algorithm to determine the optimal association scheme between the user terminal and the drone are as follows:

[0103] S1051: Set the maximum evolution number T, encode various association situations in binary form as the population, and randomly generate P individuals as the initial population;

[0104] S1052: Calculate the fitness of individuals in the population, specifically: after obtaining the association scheme corresponding to the individual, calculate the corresponding system delay according to S104, invert it and add a fixed constant as the size of the fitness;

[0105] S1053: After selection, inheritance, and mutation operations, the worst individual is replaced to update the population;

[0106] S1054: When the average fitness of the population has not converged, repeat S1052~1053; when the average fitness of the population converges or reaches the maximum evolution number, the optimal association scheme corresponding to the individual with the highest fitness in the population is output.

[0107] Compared with the prior art, the present invention has the following beneficial effects:

[0108] 1. The present invention constructs an air-ground-integrated mobile edge computing model including multiple drones and satellites. The user terminal adopts a partial offloading mode. The computing tasks of the user terminal will be processed in parallel by three methods: local computing, offloading to the drone's MEC server, and offloading to the ground cloud server through satellite relay. Considering the characteristics and computing power of each method, the computing delay of each method is analyzed, and then the communication model is constructed separately to analyze the delay consumed by task transmission under different modes; at the same time, the flight condition restrictions of multiple drones and the computing resource capacity of drones are considered, and the long propagation delay of satellite communication is considered, so as to reasonably allocate computing tasks; mathematical methods are used to solve the computing offloading and resource allocation problems of the air-ground-integrated multi-layer offloading scenario with limited number of drones and computing resources, and an approximate optimal solution is obtained.

[0109] 2. The present invention uses satellites and drones to provide efficient communication computing services for user terminals in remote areas without cellular network coverage or in areas where communication networks are destroyed by natural disasters. It can process computing tasks for user terminals and reduce system latency and energy consumption, further improving the quality of experience of user terminals in the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 Schematic diagram of the mobile edge computing system model established for the present invention.

[0111] Figure 2 It is a flow chart of the present invention.

[0112] Figure 3 This is a graph showing the solution of the association scheme based on the genetic algorithm in the present invention. DETAILED DESCRIPTION

[0113] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0114] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0115] like Figure 1 and Figure 2 As shown, an air-ground integrated mobile edge computing offloading and resource allocation optimization method provided by an embodiment of the present invention includes the following steps:

[0116] S101: Build a system model; build a mobile edge computing system model including satellites, drones, and ground user terminals without cellular network coverage, and initialize the number of user terminals, the number of drones, the number of computing tasks of the user terminals, and the channel model parameters of the communication links between the user terminals and satellites and between the user terminals and drones;

[0117] Specifically, in S101, a satellite and drone-assisted mobile edge computing system model is constructed as follows:

[0118] S1011: Set user terminal set , the drone collection is , the total CPU resources of the MEC server on the drone are f m , each terminal device has a latency-sensitive task to perform ,in D k is the task data size, C k is the computing resource required for the computing task, For mobile devices k Because the computing power and battery capacity of the user terminal itself are limited, some tasks need to be offloaded to the MEC server on the drone, or offloaded to the ground cloud server via satellite for processing. The computing tasks are divided into local tasks and , the number of tasks offloaded to the MEC server of drone m , the number of tasks offloaded to the ground cloud server via satellite transfer ; Divide the UAV flight cycle into N time slots, where the size of each time slot is δ t , the UAV is considered to remain stationary within a time slot;

[0119] S1012: In order to avoid interference between user terminals, frequency division multiple access technology is used for communication. The matching factor between user terminal k and drone m is defined asa k,m , when the user terminal k matches the drone, a k,m =1, otherwise a k,m =0; p k,m is the transmission power from user terminal k to drone m, p k,s is the transmission power from user terminal k to the satellite, assuming that the fixed altitude of the drone is H, and the channel gain between user terminal k and drone m in the nth time slot is:

[0120] ;

[0121] in, g k,m is the small-scale fading coefficient, q m ( n ) is the position coordinate of the UAV at the nth time slot, ω k is the location coordinate of user terminal k, β 0 is the channel gain at a reference distance of 1m; according to the Shannon formula, the unloading rate between the user terminal k and the drone m at the nth time slot is:

[0122] ;

[0123] in, B is the bandwidth of the communication link between the user terminal and the UAV, is the noise power;

[0124] The channel gain of user terminal k unloaded to the satellite is:

[0125]

[0126] in, and are the gains of the user terminal transmitting antenna and the satellite receiving antenna, ψ is the rain attenuation coefficient, is the Boltzmann constant, c is the speed of light, f c is the carrier frequency, H s is the satellite altitude;

[0127] Then the unloading rate of user terminal k to the satellite is:

[0128] ;

[0129] in,B s is the bandwidth of the communication link between the user terminal and the satellite, is the noise power;

[0130] S102: Based on the system model constructed in S101, the computing task processing delay of the user terminal in local computing, UAV mobile edge computing and satellite cloud computing and the corresponding system energy consumption are obtained, and finally the system delay is obtained. With the goal of minimizing the weighted sum of system delay and energy consumption, a specific optimization problem is established;

[0131] Specifically, in S102, according to the system model constructed in S101, the computing task processing delay of the user terminal in local computing, UAV mobile edge computing and satellite cloud computing and the corresponding system energy consumption are obtained, and finally the system delay is obtained, which is as follows:

[0132] The local calculation delay of user terminal k is:

[0133] ;

[0134] in, C It is the computing resources required to perform a single computing task on the user terminal;

[0135] The total delay of processing part of the computing task of user terminal k on the drone includes the offloading delay , calculate delay and return delay , expressed as , the unloading delay is approximately expressed by an integer multiple of the number of time slots. Since the amount of backhaul data is small, the backhaul delay is ignored and the calculated delay is expressed as:

[0136] ;

[0137] The UAV mobile edge computing latency is expressed as:

[0138] ;

[0139] in, is the average unloading rate of tasks from user terminal k to drone m, is the computing resource allocated by UAV m to user terminal k;

[0140] The total delay of processing part of the computing task of user terminal k on the satellite includes the transmission delay from the user terminal to the satellite, the processing delay between the satellite and the ground station and the gateway, the round-trip propagation delay between the user terminal and the satellite, and the round-trip propagation delay between the satellite and the ground station. The satellite cloud computing delay is expressed as:

[0141] ;

[0142] in, is the processing delay between the satellite and the ground station and gateway, s is the distance between the satellite and the ground station, c is the speed of light, is the unloading rate between user terminal k and the satellite;

[0143] Since the computing tasks of the user terminal need to be processed in parallel, the final delay of the user terminal k is the maximum value of the local computing delay, the UAV mobile edge computing delay, and the satellite cloud computing delay, which can be expressed as The final system delay is .

[0144] The energy consumption of user terminals includes the energy consumption unloaded to satellites and drones and the transmission energy consumption, which can be expressed as:

[0145] ;

[0146] in, It is the energy consumption factor related to the user terminal’s own equipment;

[0147] The energy consumption of drones includes flight energy consumption and computing energy consumption. The computing energy consumption is expressed as:

[0148] ;

[0149] in, is the energy consumption factor related to the chip structure of the MEC server on the drone, is the time it takes for the UAV to calculate the computational task of user terminal k;

[0150] The flight energy consumption of the UAV in the nth time slot is expressed as:

[0151] ;

[0152] in, d 0 is the drag ratio of the drone body, ρ is the air density, R S It is the solidity of the drone's wind wheel. A is the rotor disk area of ​​the drone, v m ( n ) is the flight speed of UAV m in the nth time slot, P 0 is the UAV blade profile power, is the angular velocity of the drone rotor, R e is the rotor radius of the drone, P i is the induced power of the drone blades,v 0 is the average rotor induced speed when the UAV is hovering;

[0153] Finally, the system energy consumption is expressed as:

[0154] ;

[0155] in, is the weighting coefficient, set Less than 1, indicating that in this method, reducing the energy consumption of user terminals is more important;

[0156] According to the system delay and energy consumption expressions, the specific steps of establishing the optimization problem are as follows:

[0157] S1021: According to the expressions of system delay and energy consumption, an optimization target is established as the weighted sum of system delay and energy consumption, and the expression is: ;

[0158] S1022: One of the key issues in resource allocation is how to solve the integer constraint problem of matching user terminals and drones. A single user terminal can only offload tasks with one drone, and the constraint is:

[0159] ;

[0160] S1023: The computing tasks of the user terminal are processed by local computing, UAV mobile edge computing, and satellite cloud computing, and finally the equality constraint is obtained. ; The computing resources that can be allocated to the UAV are limited, and the computing resource constraints are:

[0161] ;

[0162] S1024: Divide the flight cycle of the UAV in motion into several time slots. It is assumed that the UAV is stationary in the time slots. The channel parameters between the user terminal and the UAV in different time slots are different. The key to solving the trajectory problem is to take into account the flight energy consumption of the UAV and make the position of the UAV more suitable for unloading by the user terminal.

[0163] S1025: Considering the user terminal association constraints, the UAV flight and computing resource constraints, and the user terminal delay constraints, in order to reduce the system delay and the weighted sum of energy consumption, we optimize the user terminal computing task allocation coefficient, the UAV trajectory and association matching factor, and the computing resource allocation; finally, the optimization problem is obtained:

[0164] ;

[0165] st ;

[0166] ;

[0167] ;

[0168] ;

[0169] ;

[0170] ;

[0171] ;

[0172] in, is the normalization coefficient, V max is the maximum flight speed of the drone. and are the position coordinates of drone i and j at the nth time slot, d min is the minimum distance between drones, is the upper limit of the computing resources of UAV m, is the upper limit of the latency of the computing task of user terminal k.

[0173] S103: Under the optimization problem, based on the principle of minimizing the weighted sum of system delay and energy consumption, the offloading ratio of each user terminal is calculated according to the Lagrange multiplier method;

[0174] Specifically, in S103, the specific steps of obtaining the uninstallation ratio of each user terminal by using the Lagrange multiplier method are as follows:

[0175] S1031: First, given the association between the user terminal and the drone, the drone trajectory and the computing resource allocation, the drone flight energy consumption has been determined, and the optimization target in this part can be ignored. At this time, the computing task allocation of the user terminal does not affect each other, so the problem can be decoupled into k independent sub-problems for solution;

[0176] S1032: First, we conduct a classified discussion, assuming that the local computing, edge computing, and cloud computing delays are the maximum delays, and the user terminal delay can be expressed by one of them. We discuss by time slots, add a range constraint on the number of tasks that the user terminal unloads to the drone, and use the average unloading rate in different time slots as the rate at which the user terminal unloads to the drone. We get the local computing task delay, the drone task delay, and the satellite cloud computing processing delay. Assuming that the local computing delay is the largest and the number of time slots is N k , the optimization objectives and corresponding constraints corresponding to user terminal k are as follows:

[0177] ;

[0178] st ;

[0179] ;

[0180] ;

[0181] ;

[0182] in, is the upper limit of the latency of the computing task of user terminal k, is the energy consumption factor related to the user terminal’s own equipment, is the energy consumption factor related to the chip structure of the MEC server on the drone, is the time it takes for the drone to calculate the computational task of user terminal k, is the computing resource allocated by UAV m to user terminal k;

[0183] S1033: After the classification discussion is completed, the solution with the minimum objective function value is used as the optimal uninstallation ratio solution for user terminal k, and the uninstallation ratio solutions for all user terminals are obtained in turn.

[0184] S104: transforming the constraints into convex constraints according to the sequential convex optimization principle to obtain the optimal UAV trajectory and computing resource allocation ratio;

[0185] Specifically, in S104, the specific steps of obtaining the trajectory of the UAV and calculating the resource allocation ratio by sequential convex optimization are as follows:

[0186] S1041: At this time, it is known that the optimization target is the UAV computing processing delay and the UAV energy consumption. First, the UAV transmission rate in the optimization target is relaxed so that the UAV mobile edge computing delay and UAV computing energy consumption in the optimization target are transformed into a convex form, and the relaxation variable is added. ,get:

[0187] ;

[0188] Where:

[0189] ;

[0190] ;

[0191] in, is the unloading rate from user terminal k to drone m, , slack variables satisfy , is the variable in the first loop The value taken at the first-order Taylor expansion, In the first cycle The approximate value of

[0192] For the flight energy consumption part of the optimization objective, add relaxation variables:

[0193] ;

[0194] Right now ;

[0195] After the transformation, the flight energy consumption becomes convex:

[0196] ;

[0197] in, d 0 is the drag ratio of the drone body, ρ is the air density, R S It is the solidity of the drone's wind wheel. A is the rotor disk area of ​​the drone, v m ( n ) is the flight speed of UAV m in the nth time slot, P 0 is the UAV blade profile power, Ω e is the angular velocity of the drone rotor, R e is the rotor radius of the drone, P i is the induced power of the drone blades, v 0 is the average rotor induced speed when the UAV is hovering;

[0198] S1042: Then relax the distance constraint between the UAVs to make the non-convex constraint a linear constraint, and obtain:

[0199] ;

[0200] in, q i ( n )and q j ( n ) are the position coordinates of drone i and j at the nth time slot, and They are q i ( n )and q j ( n ) is the value of the first-order Taylor expansion at the lth cycle, is the relaxation of the distance between UAVs i and j at the nth time slot;

[0201] S1043: The converted optimization objectives and constraints are as follows:

[0202] ;

[0203] st. ;

[0204] ;

[0205] ;

[0206] ;

[0207] , ;

[0208] in, E m is the energy consumption of UAV m;

[0209] The interior point method of the convex optimization algorithm is used to solve the problem, and the UAV trajectory and computing resource allocation ratio in the current cycle are obtained. The solved result is used as the initial value and the convex optimization solution is continued until convergence, and the optimal UAV trajectory and computing resource allocation plan in the current cycle is obtained.

[0210] S105: Based on the scheme of associating user terminals with drones in each user terminal group, S103 and S104 are repeatedly executed iteratively to obtain the corresponding system delay and use the corresponding system delay as the fitness evaluation criterion, and finally obtain the optimal association scheme and the corresponding optimal unloading ratio, drone trajectory and computing resource allocation ratio according to the genetic algorithm;

[0211] Specifically, in S105, the specific steps of using the genetic algorithm to determine the optimal association scheme between the user terminal and the drone are as follows (eg Figure 3 shown):

[0212] S1051: Set the maximum evolution number T, encode various association situations in binary form as the population, and randomly generate P individuals as the initial population;

[0213] S1052: Calculate the fitness of individuals in the population, specifically: after obtaining the association scheme corresponding to the individual (establishing the association scheme and the chromosome mapping relationship in the genetic algorithm in each user terminal group, determining the population size and the number of iterations), calculate the corresponding system delay according to S104, invert it and add a fixed constant as the size of the fitness;

[0214] S1053: After selection, inheritance, and mutation operations, the worst individual is replaced to update the population;

[0215] S1054: When the average fitness of the population has not converged, repeat S1052~1053; when the average fitness of the population converges or reaches the maximum evolution number, the optimal association scheme corresponding to the individual with the highest fitness in the population is output.

[0216] The above are only preferred embodiments of the present invention. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These should also be regarded as the protection scope of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. An air-ground-space integrated mobile edge computing offloading and resource allocation optimization method, characterized in that: The following steps are involved: S101: Build a system model; build a mobile edge computing system model including satellites, drones, and ground user terminals without cellular network coverage, and initialize the number of user terminals, the number of drones, the number of computing tasks of the user terminals, and the channel model parameters of the communication links between the user terminals and satellites and between the user terminals and drones; S102: Based on the system model constructed in S101, the computing task processing delay of the user terminal in local computing, UAV mobile edge computing and satellite cloud computing and the corresponding system energy consumption are obtained, and finally the system delay is obtained. With the goal of minimizing the weighted sum of system delay and energy consumption, a specific optimization problem is established; S103: Under the optimization problem, based on the principle of minimizing the weighted sum of system delay and energy consumption, the offloading ratio of each user terminal is calculated according to the Lagrange multiplier method; S104: transforming the constraints into convex constraints according to the sequential convex optimization principle to obtain the optimal UAV trajectory and computing resource allocation ratio; S105: Based on the scheme of associating user terminals with drones in each user terminal group, S103 and S104 are repeatedly executed iteratively to obtain the corresponding system delay and use the corresponding system delay as the fitness evaluation criterion, and finally obtain the optimal association scheme and the corresponding optimal unloading ratio, drone trajectory and computing resource allocation ratio according to the genetic algorithm; In the S101, a satellite and drone-assisted mobile edge computing system model is constructed as follows: S1011: Assume that the user terminal set K = {1, 2, ..., K}, the drone set M = {1, 2, ..., M}, and the total CPU resources of the MEC server on the drone are f m , each terminal device has a latency-sensitive task to perform Where D k is the task data size, C k is the computing resource required for the computing task, is the computing power of mobile device k; the computing tasks are divided into local tasks Number of tasks offloaded to the MEC server of drone m Number of tasks offloaded to ground cloud servers via satellite transfer The UAV flight cycle is divided into N time slots, where the size of each time slot is δ t , the UAV is considered to remain stationary within a time slot; S1012: Define the matching factor between user terminal k and drone m as a k,m , when the user terminal k matches the drone, a k,m =1, otherwise a k,m =0; let p k,m is the transmission power from user terminal k to drone m, p k,s is the transmission power from user terminal k to the satellite, assuming that the fixed altitude of the drone is H, and the channel gain between user terminal k and drone m in the nth time slot is: Among them, g k,m is the small-scale fading coefficient, q m (n) is the position coordinate of the UAV at the nth time slot, ω k is the position coordinate of user terminal k, β0 is the channel gain with a reference distance of 1m; according to Shannon's formula, the unloading rate between user terminal k and drone m at the nth time slot is: Where B is the bandwidth of the communication link between the user terminal and the UAV, σ 2 is the noise power; The channel gain of user terminal k unloaded to the satellite is: in, and are the gains of the user terminal transmitting antenna and the satellite receiving antenna, ψ is the rain attenuation coefficient, η is the Boltzmann constant, c is the speed of light, and f c is the carrier frequency, H s is the satellite altitude; Then the unloading rate of user terminal k to the satellite is: Among them, B s is the bandwidth of the communication link between the user terminal and the satellite, σ 2 is the noise power; In S102, according to the system model constructed in S101, the computing task processing delay of the user terminal in local computing, UAV mobile edge computing and satellite cloud computing and the corresponding system energy consumption are obtained, and finally the system delay is obtained, which is as follows: The local calculation delay of user terminal k is: Where C is the computing resources required to execute a single computing task of the user terminal; The total delay of processing part of the computing task of user terminal k on the drone includes the offloading delay, computing delay and return delay, which is expressed as The unloading delay is approximately expressed as an integer multiple of the number of time slots, and the return delay is ignored. The calculated delay is expressed as: The UAV mobile edge computing latency is expressed as: in, is the average unloading rate of tasks from user terminal k to drone m, is the computing resource allocated by UAV m to user terminal k; The total delay of processing part of the computing task of user terminal k on the satellite includes the transmission delay from the user terminal to the satellite, the processing delay between the satellite and the ground station and the gateway, the round-trip propagation delay between the user terminal and the satellite, and the round-trip propagation delay between the satellite and the ground station. The satellite cloud computing delay is expressed as: Among them, t trip is the processing delay between the satellite and the ground station and gateway, s is the distance between the satellite and the ground station; The delay of user terminal k is the maximum value among the local computing delay, the UAV mobile edge computing delay and the satellite cloud computing delay, expressed as The final system delay is: The energy consumption of user terminals includes the energy consumption unloaded to satellites and drones and the transmission energy consumption, which can be expressed as: Among them, κ is the energy consumption factor related to the user terminal’s own equipment; The energy consumption of drones includes flight energy consumption and computing energy consumption. The computing energy consumption is expressed as: Among them, κ e is the energy consumption factor related to the chip structure of the MEC server on the drone, is the time it takes for the UAV to calculate the computational task of user terminal k; The flight energy consumption of the UAV in the nth time slot is expressed as: Where d0 is the drag ratio of the drone body, ρ is the air density, and R S is the solidity of the UAV wind wheel, A is the area of ​​the UAV rotor disk, v m (n) is the flight speed of UAV m in the nth time slot, P0 is the UAV blade profile power, Ω e is the angular velocity of the UAV rotor, R e is the rotor radius of the UAV, P i is the induced power of the UAV blade, and v0 is the average rotor induced speed of the UAV when it is hovering; Finally, the system energy consumption is expressed as: Among them, ε is the weighting coefficient, ε is less than 1; According to the system delay and energy consumption expressions, the specific steps of establishing the optimization problem are as follows: S1021: According to the expressions of system delay and energy consumption, establish an optimization target as the weighted sum of system delay and energy consumption, expressed as min T sum +φE sum ; S1022: A single user terminal can only offload tasks to one drone, with the following constraints: S1023: The computing tasks of the user terminal are processed by local computing, UAV mobile edge computing, and satellite cloud computing, and finally the equality constraint is obtained. The computing resource constraints of the drone are: S1024: Divide the flight cycle of the UAV in motion into a number of time slots, assume that the UAV is stationary in the time slots, and have different channel parameters between the user terminal and the UAV in different time slots; S1025: Considering the user terminal association constraints, the UAV flight and computing resource constraints, and the user terminal delay constraints, the user terminal computing task allocation coefficient, the UAV trajectory and the association matching factor, and the computing resource allocation are optimized; finally, the optimization problem is obtained: as sum +φE sum ; Among them, φ is the normalization coefficient, V max is the maximum flight speed of the drone, q i [n] and q j [n] are the position coordinates of drones i and j at the nth time slot, d min is the minimum distance between drones, is the upper limit of the computing resources of UAV m, is the upper limit of the latency of the computing task of user terminal k.

2. The air-ground-space integrated mobile edge computing offloading and resource allocation optimization method according to claim 1 is characterized in that: In S103, the specific steps of obtaining the offloading ratio of each user terminal by using the Lagrange multiplier method are as follows: S1031: Decouple the problem into k independent sub-problems for solution; S1032: Assume that the local calculation delay is the largest and the number of time slots is N k , the user terminal delay is represented by the maximum local computing delay, and the task number range constraint of the user terminal unloading to the UAV is added. At the same time, the average unloading rate in different time slots is represented as the rate of unloading from the user terminal to the UAV; the optimization objective and corresponding constraints corresponding to the user terminal k are as follows: S1033: The solution with the minimum objective function value is taken as the optimal unloading ratio scheme for user terminal k, and the unloading ratio schemes for all user terminals are obtained in sequence.

3. The method for offloading and optimizing resource allocation of integrated air-ground-space mobile edge computing according to claim 1, characterized in that: In S104, the specific steps of obtaining the UAV trajectory and calculating the resource allocation ratio by sequential convex optimization are as follows: S1041: First, relax the UAV transmission rate in the optimization target, so that the UAV mobile edge computing delay and UAV computing energy consumption in the optimization target are transformed into a convex form, and increase the relaxation variable y k,m (n), we get: Where: Among them, R k,m (n) is the unloading rate from user terminal k to drone m, φ k,m (n) = p k,m g k,m 2 β0 / σ 2 , the slack variable y k,m (n) satisfies y k,m (n)≥||q m (n)-ω k || 2 , is the variable y in the first loop k,m (n) The value taken at the first-order Taylor expansion, is R in the first cycle k,m (n) approximate value; For the flight energy consumption part of the optimization objective, add relaxation variables: Right now After the transformation, the flight energy consumption becomes convex: S1042: Then relax the distance constraint between the UAVs to make the non-convex constraint a linear constraint, and obtain: Among them, q i (n) and q j (n) are the position coordinates of drone i and j at the nth time slot, and They are q i (n) and q j (n) is the value of the first-order Taylor expansion at the lth cycle, is the relaxation of the distance between UAVs i and j at the nth time slot; S1043: The converted optimization objectives and constraints are as follows: Among them, E m is the energy consumption of UAV m; The interior point method of the convex optimization algorithm is used to solve the problem, and the UAV trajectory and computing resource allocation ratio in the current cycle are obtained. The solved result is used as the initial value and the convex optimization solution is continued until convergence, and the optimal UAV trajectory and computing resource allocation plan in the current cycle is obtained.

4. The method for offloading and optimizing resource allocation of integrated air-ground-space mobile edge computing according to claim 1, characterized in that: In S105, the specific steps of using the genetic algorithm to determine the optimal association scheme between the user terminal and the drone are as follows: S1051: Set the maximum evolution number T, encode various association situations in binary form as the population, and randomly generate P individuals as the initial population; S1052: Calculate the fitness of individuals in the population, specifically: after obtaining the association scheme corresponding to the individual, calculate the corresponding system delay according to S104, invert it and add a fixed constant as the size of the fitness; S1053: After selection, inheritance, and mutation operations, the worst individual is replaced to update the population; S1054: When the average fitness of the population has not converged, repeat S1052 to S1053; when the average fitness of the population converges or reaches the maximum evolution number, the optimal association scheme corresponding to the individual with the highest fitness in the population is output.