Policy for solving cloud edge collaborative task unloading

Abstract

The invention relates to the technical field of traffic safety, in particular to a cloud edge collaborative task unloading solution strategy, which comprises the following steps of: determining the specific definition of a problem, collecting all related data required by task unloading, calculating the fitness value of each candidate solution by using an improved genetic algorithm, and adopting a roulette selection strategy to obtain a cloud edge collaborative task unloading solution. And selecting individuals from the current population according to the fitness value to enter the next generation of population, dynamically adjusting the crossover rate and the mutation rate, combining new individuals generated after crossover and mutation operations with parent individuals to form a new generation of population, updating the number of iterations, and checking whether a termination condition is met or not. Tasks of users are processed through cooperation of edge nodes and cloud server nodes, a two-layer computing framework is provided to solve the cloud edge cooperation task unloading problem, the total delay of a network and the total energy consumption of an edge server are minimized, an improved self-adaptive genetic algorithm is used for solving, and the task unloading efficiency is improved. And finally, proving the performance of the proposed two-layer computing architecture unloading strategy through an experimental result.

Description

Technical Field

[0001] The present invention relates to the technical field of collaborative task offloading, and specifically to a method for solving the cloud-edge collaborative task offloading strategy. Background Art

[0002] With the development of the network and the popularization of user equipment (UE), people can easily access the network through UE and obtain the desired services. The terrestrial network only covers about 20% of the Earth's land area, less than 6% of the Earth's surface. Moreover, the terrestrial network is usually deployed in densely populated areas such as cities. In sparsely populated areas such as rural areas, deserts, and forests, the terrestrial network is not deployed due to high construction difficulty or cost. UEs in these areas cannot rely on the terrestrial network to obtain services. In addition, the terrestrial network is vulnerable to natural disasters. Compared with the terrestrial network, the satellite network has a wider coverage range and is not affected by climate and geographical conditions. The satellite network can supplement and expand the terrestrial network of the Fifth Generation Mobile Communication (5G). In the future Sixth Generation Mobile Communication (6G) network, the satellite network can be used as a supplement to the terrestrial cellular network to achieve seamless global coverage. To achieve goals such as expanding the service range, providing reliable and on-demand services, etc., the satellite-terrestrial integrated network has become a current research hotspot and has received extensive attention from the academic and industrial communities.

[0003] Compared with the traditional terrestrial network, the satellite-terrestrial integrated network usually has a higher communication delay due to the long propagation distance of the satellite-terrestrial link. In particular, cloud services are widely used in the current communication network due to the advantages of high computing power in the cloud. However, in the satellite-terrestrial integrated network, the need for computing services from the cloud may lead to a higher communication delay because users in areas without traditional terrestrial network coverage can only access the cloud through satellites. To avoid frequent communication with the cloud, the Mobile Edge Computing (MEC) architecture can be applied to place the "cloud" closer to the user at the network edge. Different from the traditional cloud computing system, the MEC server is deployed at the wireless access point (AP) and is geographically closer to the mobile device (MD), which can effectively overcome the disadvantages of long transmission delay and high communication overhead between the MD and the cloud server. However, the computing power of the MEC server is generally limited. Therefore, it is necessary to carefully design the offloading strategy of computing tasks to optimize the service capacity and delay performance of the network. As an optimization operation, the goal of computing offloading is diverse: reducing delay, controlling costs, saving energy, improving the load balancing rate, and so on.

[0004] As described above, existing work mainly focuses on binary offloading using MEC servers on ground base stations or satellites, and fails to fully utilize the collaborative computing architecture to improve latency performance. Especially in the satellite-ground backhaul network architecture, how to optimize latency performance through collaborative multi-layer computing remains an open issue.

[0005] Based on this, the present invention provides a method for solving the cloud-edge collaborative task offloading strategy to address the above-mentioned technical problems. Summary of the Invention

[0006] The object of the present invention is to provide a method for solving the cloud-edge collaborative task offloading strategy to address the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for solving the cloud-edge collaborative task offloading strategy is provided, including the following steps:

[0009] S1. Clearly define the specific problem, including the network model, computing model, latency, and energy consumption model, define the cloud-edge collaborative task offloading problem in the satellite-ground integrated network, and construct a two-layer computing architecture including base stations, user equipment, edge servers, and cloud data centers;

[0010] S2. Collect all relevant data required for task offloading, including the data volume, computing requirements of user tasks, computing capabilities of base stations and the cloud, channel bandwidth, and transmission power, to provide necessary data support for the task offloading strategy;

[0011] S3. Use an improved genetic algorithm to generate an initial population containing multiple candidate solutions, providing an initial solution set for the genetic algorithm as the starting point for the search;

[0012] S4. Calculate the fitness value of each candidate solution to evaluate its performance in minimizing latency and energy consumption, providing a basis for selection, crossover, and mutation operations to ensure that the algorithm evolves towards the optimization goal;

[0013] S5. Adopt the roulette wheel selection strategy to select individuals from the current population into the next generation population according to the fitness value, retaining excellent individuals and eliminating inferior individuals to ensure the evolution direction of the population;

[0014] S6. Dynamically adjust the crossover rate and perform crossover operations on the selected individuals to generate new candidate solutions, producing new solutions through gene recombination to increase the diversity of the population;

[0015] S7. Dynamically adjust the mutation rate and perform mutation operations on some individuals to introduce new genes, producing new solutions through gene mutation to further increase the diversity of the population and prevent the algorithm from falling into local optimality;

[0016] S8. Combine the new individuals generated after the crossover and mutation operations with the parental individuals to form a new generation of population, update the iteration count, continuously evolve the population, and gradually approach the optimal solution;

[0017] S9. Check whether the termination condition is satisfied. If it is satisfied, terminate the algorithm; otherwise, return to step S4 to continue the iteration, ensure that the algorithm ends within a reasonable time, and output the final task offloading strategy.

[0018] Preferably, the specific implementation steps of step S1 are as follows:

[0019] S1.1. Define a network including several base stations through a network model. Each base station serves several users, and each base station is equipped with an MEC server to form a two-layer computing architecture of base station edge computing and cloud computing;

[0020] S1.2. Through a computing model, each user has a computing task, calculate the CPU cycles required to calculate one-bit input data, and divide the task into independent parts to be processed simultaneously at the BS and the cloud;

[0021] S1.3. Calculate the latency and energy consumption of edge computing and cloud computing through a latency and energy consumption model.

[0022] Preferably, the specific implementation steps of step S2 are as follows:

[0023] S2.1. Collect the data volume and computing requirements of user tasks;

[0024] S2.2. Collect the computing capabilities of the base station and the cloud;

[0025] S2.3. Collect channel parameters, including channel bandwidth, background noise power, transmission power, and channel gain.

[0026] Preferably, each candidate solution in step S3 represents the allocation of tasks to the computing nodes of the BS or the cloud.

[0027] Preferably, the specific implementation steps of step S4 are as follows:

[0028] S4.1. Calculate the fitness value of each candidate solution through the fitness function fitness.

[0029] Preferably, the specific implementation steps of step S6 are as follows:

[0030] S6.1. Improve the genetic algorithm and dynamically adjust the crossover rate:

[0031] S6.2. Judge whether the maximum fitness value of the currently two crossed individuals is greater than the average fitness value of the population, and calculate the crossover rate respectively;

[0032] S6.3. Perform crossover operation on the selected individuals to generate new candidate solutions.

[0033] Preferably, the specific implementation steps of step S7 are as follows:

[0034] S7.1. Improve the genetic algorithm and dynamically adjust the mutation rate:

[0035] S7.2. Determine whether the fitness value of the current individual is greater than the average fitness value of the population, and calculate the mutation rate respectively;

[0036] S7.3. Perform mutation operation on some individuals, randomly change the task allocation situation, and introduce new genes.

[0037] Preferably, the specific implementation steps of step S8 are as follows:

[0038] S8.1. Combine the new individuals generated after crossover and mutation operations with the parent individuals to form a new generation of population;

[0039] S8.2. Update the iteration times and prepare for the next round of iteration.

[0040] Preferably, the specific implementation steps of step S9 are as follows:

[0041] S9.1. Check whether the termination conditions are met, including reaching the maximum number of iterations or the fitness value converging;

[0042] S9.2. If the termination conditions are met, terminate the algorithm and output the optimal candidate solution in the current population as the final task offloading strategy.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] Based on a two-layer computing framework, the present invention solves the cloud-edge collaborative task offloading problem to minimize the total network delay and the total energy consumption of the edge server. Finally, an improved adaptive genetic algorithm is used for solving. Experiments show that the present invention has obtained relatively good results in various experiments, solving the problems raised in the background technology. In summary, the present invention processes the user's tasks through the cooperation of edge nodes and cloud server nodes, proposes a two-layer computing framework to solve the cloud-edge collaborative task offloading problem, minimizes the total network delay and the total energy consumption of the edge server, uses an improved adaptive genetic algorithm for solving, and finally proves the performance of the proposed two-layer computing architecture offloading strategy through experimental results. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is the network model topology diagram of the present invention;

[0046] Figure 2 It is the main simulation parameter value table of communication and computing of the present invention;

[0047] Figure 3 This is the value table of algorithm parameters in the experiment of the present invention;

[0048] Figure 4 This is the bar chart showing the impact of the number of user tasks on the total cost in the present invention;

[0049] Figure 5 This is the bar chart showing the impact of the number of user tasks on the total cost in the present invention;

[0050] Figure 6 This is the bar chart showing the impact of the computing power of the BS server on the total cost in the present invention;

[0051] Figure 7 This is the line chart showing the impact of the number of iterations on the total cost in the present invention. Detailed implementation manners

[0052] Next, in combination with the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment:

[0054] Embodiment 1

[0055] Please refer to Figures 1 to 3 , the present invention proposes a method for solving the cloud-edge collaborative task offloading strategy, including the following steps:

[0056] S1. Clearly define the specific problem, including the network model, computing model, delay and energy consumption model, define the cloud-edge collaborative task offloading problem in the satellite-terrestrial integrated network, and construct a two-layer computing architecture including base stations, user equipment, edge servers and cloud data centers;

[0057] In this embodiment, it should also be noted that the specific implementation steps of step S1 are as follows:

[0058] S1.1. Define a network including several base stations through the network model. Each base station serves several users, and each base station is equipped with an MEC server to form a two-layer computing architecture of base station edge computing and cloud computing;

[0059] S1.2. Through the computing model, each user has a computing task, calculate the CPU cycles required to calculate one-bit input data, and divide the task into independent parts to be processed simultaneously at the BS and the cloud;

[0060] S1.3. Calculate the latency and energy consumption of edge computing and cloud computing through the latency and energy consumption models.

[0061] S2. Collect all relevant data required for task offloading, including the data volume, computing requirements of user tasks, computing capabilities of the base station and the cloud, channel bandwidth, transmission power, to provide necessary data support for the task offloading strategy;

[0062] In this embodiment, it should also be noted that the specific implementation steps of step S2 are as follows:

[0063] S2.1. Collect the data volume and computing requirements of user tasks;

[0064] S2.2. Collect the computing capabilities of the base station and the cloud;

[0065] S2.3. Collect channel parameters, including channel bandwidth, background noise power, transmission power, and channel gain.

[0066] S3. Use the improved genetic algorithm to generate an initial population containing multiple candidate solutions, providing an initial solution set for the genetic algorithm as the starting point for the search;

[0067] In this embodiment, it should also be noted that each candidate solution in step S3 represents the allocation of tasks to the BS or cloud computing nodes.

[0068] S4. Calculate the fitness value of each candidate solution, evaluate its performance in minimizing latency and energy consumption, providing a basis for selection, crossover, and mutation operations to ensure that the algorithm evolves towards the optimization goal;

[0069] In this embodiment, it should also be noted that the specific implementation steps of step S4 are as follows:

[0070] S4.1. Calculate the fitness value of each candidate solution through the fitness function fitness.

[0071] S5. Adopt the roulette wheel selection strategy to select individuals from the current population into the next generation population according to the fitness value, retain excellent individuals, and eliminate inferior individuals to ensure the evolution direction of the population;

[0072] S6. Dynamically adjust the crossover rate, perform crossover operations on the selected individuals to generate new candidate solutions, produce new solutions through gene recombination, and increase the diversity of the population;

[0073] In this embodiment, it should also be noted that the specific implementation steps of step S6 are as follows:

[0074] S6.1. Improve the genetic algorithm and dynamically adjust the crossover rate:

[0075] S6.2. Determine whether the maximum fitness value of the current two crossed individuals is greater than the population average fitness value, and calculate the crossover rate respectively;

[0076] S6.3. Perform a crossover operation on the selected individuals to generate new candidate solutions.

[0077] S7. Dynamically adjust the mutation rate, perform a mutation operation on some individuals, introduce new genes, generate new solutions through gene mutation, further increase the diversity of the population, and prevent the algorithm from falling into a local optimum;

[0078] In this embodiment, it should also be noted that the specific implementation steps of step S7 are as follows:

[0079] S7.1. Improve the genetic algorithm and dynamically adjust the mutation rate:

[0080] S7.2. Determine whether the fitness value of the current individual is greater than the population average fitness value, and calculate the mutation rate respectively;

[0081] S7.3. Perform a mutation operation on some individuals, randomly change the task allocation situation, and introduce new genes.

[0082] S8. Combine the new individuals generated after the crossover and mutation operations with the parent individuals to form a new generation of population, update the iteration times, continuously evolve the population, and gradually approach the optimal solution;

[0083] In this embodiment, it should also be noted that the specific implementation steps of step S8 are as follows:

[0084] S8.1. Combine the new individuals generated after the crossover and mutation operations with the parent individuals to form a new generation of population;

[0085] S8.2. Update the iteration times and prepare for the next iteration.

[0086] S9. Check whether the termination condition is satisfied. If it is satisfied, terminate the algorithm; otherwise, return to step S4 to continue the iteration, ensure that the algorithm ends within a reasonable time, and output the final task offloading strategy.

[0087] In this embodiment, it should also be noted that the specific implementation steps of step S9 are as follows:

[0088] S9.1. Check whether the termination condition is satisfied, including reaching the maximum number of iterations or the fitness value converging;

[0089] S9.2. If the termination condition is satisfied, terminate the algorithm and output the optimal candidate solution in the current population as the final task offloading strategy.

[0090] Embodiment 2

[0091] Please refer toFigures 1 to 7 , A practical application for solving the cloud-edge collaborative task offloading strategy. Specifically, it includes the following steps:

[0092] (1) Architecture of the network model:

[0093] (1.1) Network model

[0094] Please refer to Figure 1 , There are N base stations in the network, and each base station provides services for K users within its coverage area;

[0095] Then, the total number of users in the network is

[0096] Each base station is equipped with an MEC server, and a two-layer computing architecture for the cloud-edge network is formulated, namely base station edge computing and cloud computing;

[0097] Due to the limited computing power of user devices, part of the computing tasks of users are offloaded to relevant base stations to reduce computing latency;

[0098] Then, each base station also offloads part of the computing tasks to the cloud data center through satellite backhaul transmission;

[0099] (2.2) Computing model

[0100] Task model: In the network, each user has a computing task that needs to be offloaded and processed;

[0101] For user k associated with BS n, the computing task is represented by Q n,k =(D n,k , C n,k ), where D n,k is the task input data size (in bits), and C n,k is the number of CPU cycles required to compute one bit of input data;

[0102] Then, the total number of CPU cycles required for task Q n,k is

[0103] For BS n, the task set of all users within its coverage area is composed of

[0104] In, it is considered that the computing task is divisible, and based on this, a partial offloading model is applied;

[0105] Each task Q n,k is arbitrarily divided into independent parts and processed simultaneously at the BS and in the cloud;

[0106] Offloading model: As shown in Figure 1 the two-layer computing architecture;

[0107] Due to the limited computing power of user equipment, part of the user's computing tasks are offloaded to the relevant BS to reduce computational latency.

[0108] Then, each base station also offloads the computing tasks to the cloud through satellite backhaul transmission.

[0109] (1.3) Latency Analysis

[0110] Edge computing: When the user task L is offloaded to the BS, similar to most existing work, the size of the computation result is considered to be much smaller than the size of the task input data. The data transmission latency between the user and the base station can be expressed as Equation (1). When the user task L is offloaded to different edge servers, the amount of data to be transmitted is d v’v , then the transmission latency can be expressed as Equation (2);

[0111]

[0112] where D n,k C n,k represents the amount of data of the task, represents the computing power of the edge server, and T tran is the data transmission latency;

[0113]

[0114] where R m'm is the transmission rate from edge server m' to edge server m, and T m'm represents the transmission latency between m' and m, which can be calculated by the Shannon formula:

[0115]

[0116] where B is the channel bandwidth, P m' is the transmission power of edge server m', h m'm is the channel gain between m' and m, and N0 represents the noise power;

[0117] Therefore, the total time T for task L to complete can be expressed as Equation (4):

[0118] T total = T m'm + T tran (4);

[0119] Cloud data center computing: When task L is offloaded to the cloud server, the total latency of cloud computing data processing includes the communication latency from the user to the base station and the communication latency from the base station to the cloud. Similarly, the communication latency from the user to the base station can be expressed by Equation (1), and the communication latency from the base station to the cloud can be expressed by Equation (5):

[0120]

[0121] where o b,c is the transmission distance between the base station and the cloud, c is the speed of light, and t b,c is the communication delay from the base station to the cloud

[0122] Therefore, the total time T for task L to complete can be expressed as Equation (6);

[0123] T = T tran + T b,c (6);

[0124] (1.4) Energy consumption analysis

[0125] Edge computing: When user task L is offloaded to the BS, similar to the existing work, the execution energy consumption of task L on the edge server can be expressed as Equation (7). When user task L is offloaded to different BSs, its transmission energy consumption can be expressed as (8):

[0126] e vm = κT tran (7);

[0127] where κ is the energy coefficient of the edge server, and its specific value is related to the specific parameters of the edge server;

[0128]

[0129] where P v' represents the transmission power between BSs;

[0130] Therefore, the total energy consumption of the edge server is expressed as (9):

[0131] E = e vm + e v'v (9);

[0132] 2.5 Research problem description

[0133] Research the cloud-edge collaborative task offloading problem considering joint delay and energy consumption. According to Equations (6) and (9), the research problem is optimized as follows:

[0134]

[0135] (2) Algorithm improvement and construction:

[0136] (2.1) Improved genetic algorithm

[0137] The genetic algorithm (GA) is a heuristic search algorithm that mimics the biological evolution mechanism in the process of natural selection;

[0138] When solving optimization problems, the genetic algorithm iteratively optimizes candidate solutions by simulating operations such as crossover (or pairing), mutation, and selection in the biological genetic process;

[0139] In the genetic algorithm, the crossover process refers to the simulation of the biological sexual reproduction process and is the main way of gene recombination;

[0140] Crossover modes include: single-point crossover, double-point crossover, and multi-point crossover;

[0141] The mutation operation describes the gene mutation process at local positions of chromosomes or individual positions during biological evolution or reproduction;

[0142] The crossover rate and mutation rate are dynamically adjusted in the manner shown in Equations (11), (12), (13), and (14):

[0143]

[0144] where P c and P m represent the crossover rate and mutation rate of the current population, P cmax and P cmin represent the maximum crossover rate and minimum crossover rate of the population, P mmax and P mmin represent the maximum mutation rate and minimum mutation rate of the population, f represents the fitness value of the current individual, favg represents the average fitness value of the current population, and f max represents the maximum fitness value of the current population;

[0145] (2.2) Fitness Function Construction

[0146] For the genetic algorithm, the selection of the fitness function is crucial, which directly affects the result of the algorithm operation;

[0147] Normally, most of them select the objective function as the fitness function. However, the established cloud-edge collaboration model is a multi-objective problem, and the two objective functions have a large difference in order of magnitude. If they are directly added by taking weights, it doesn't make much sense. Therefore, the following processing is carried out:

[0148] Assume that all tasks are offloaded to the edge server. At this time, the execution time of the task is T B , and the energy consumption is E B , and assume that all tasks are offloaded to the cloud data center. At this time, the execution time and energy consumption of the task are T C and E C , and the designed fitness function fitness is Equation (15):

[0149]

[0150] (2.3) Algorithm Description

[0151] The improved genetic algorithm, GA-plus algorithm process is as follows;

[0152] Step 1. Let g = 0, generate the initial population POP(g);

[0153] Step 2. Calculate the fitness values of the individuals in POP(g) according to Equation (15);

[0154] Step 3. When the maximum fitness value f of the current two crossed individuals is greater than or equal to the average fitness value favg of the current population, the crossover rate Pc is calculated according to Equation (11), and the mutation rate Pm is calculated according to Equation (13). Otherwise, the crossover rate is calculated according to Equation (12), and the mutation rate is calculated according to Equation (14);

[0155] Step 4. Cross and mutate the population POP(g) according to the crossover rate and mutation rate in Step 3;

[0156] Step 5. Select POPSIZE individuals from Step 4 by roulette wheel and put them into the next generation population POP(g+1)

[0157] Step 6. Let g = g + 1;

[0158] Step 7. Judge whether the termination condition is satisfied. If so, end; otherwise, go to Step 2;

[0159] (3) In the simulation, the default height from the base station to the cloud is set to 9000 kilometers, and the time delay from the base station to the cloud is set to 100 milliseconds;

[0160] The total number of base stations N = 15, and each base station serves 8 users within a coverage radius of 300m;

[0161] The ground channel is assumed to be a Rayleigh channel, while the satellite channel is modeled as a Rice channel;

[0162] Please refer to Figure 2 and Figure 3 , for each computing task, the input data size follows a random distribution of D n,k ∈ [100, 1000];

[0163] In the comparative experiment, when the number of tasks is set to 100, 200, 500, 1000, 2000, 5000, and 10000 respectively, the minimum values of the tasks are solved by four algorithms: the Greedy algorithm, the Simulated Annealing algorithm (SA), the Genetic algorithm (GA), and the improved Genetic algorithm (GA-plus);

[0164] Please refer to Figure 4 and Figure 5 For each algorithm, run it randomly 50 times to solve the total cost of the tasks. The final experimental results are shown in the figure;

[0165] It can be seen from the figure that when the number of tasks changes from 100 to 10,000, the solutions obtained by the traditional genetic algorithm GA and simulated annealing algorithm SA are generally larger, indicating that their overall optimization capabilities are relatively poor;

[0166] The optimization ability of the greedy algorithm Greedy is relatively good, but the best-performing algorithm is the proposed GA-plus algorithm, and the total cost it solves is the smallest;

[0167] Please refer to Figure 6 to show the influence of the computing power of the BS server on the total cost when the number of tasks is 1,000;

[0168] Please refer to Figure 7 to show the convergence process of the GA-plus algorithm, traditional genetic algorithm GA, and simulated annealing algorithm SA when the number of tasks is 1,000. It is found that the evolution curve of the GA-plus algorithm is relatively steep from the 1st generation to about the 27th generation, indicating that its optimization ability is better than the other two algorithms;

[0169] Therefore, study the multi-layer collaborative computing in the space-ground integrated network, and based on the proposed two-layer computing framework, solve the cloud-edge collaborative task offloading problem to minimize the total network delay and the total energy consumption of the edge server;

[0170] Finally, use the improved adaptive genetic algorithm to solve it;

[0171] In summary, the experimental results fully show that the method obtains relatively good results in various set experiments.

[0172] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0173] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A method for solving the cloud-edge collaborative task offloading strategy, characterized in that It includes the following steps: S1. Clearly define the specific problem, including the network model, computing model, latency, and energy consumption model, define the cloud-edge collaborative task offloading problem in the space-air-ground integrated network, and construct a two-layer computing architecture including base stations, user equipment, edge servers, and cloud data centers; S2. Collect all relevant data required for task offloading, including the data volume, computing requirements of user tasks, computing capabilities of base stations and the cloud, channel bandwidth, and transmission power, to provide necessary data support for the task offloading strategy; S3. Use an improved genetic algorithm to generate an initial population containing multiple candidate solutions, providing an initial solution set for the genetic algorithm as the starting point for searching; S4. Calculate the fitness value of each candidate solution, evaluate its performance in minimizing latency and energy consumption, providing a basis for selection, crossover, and mutation operations to ensure that the algorithm evolves towards the optimization goal; S5. Adopt the roulette wheel selection strategy to select individuals from the current population into the next generation population according to the fitness value, retain excellent individuals, and eliminate inferior individuals to ensure the evolution direction of the population; S6. Dynamically adjust the crossover rate, perform crossover operations on the selected individuals to generate new candidate solutions, generate new solutions through gene recombination, and increase the diversity of the population; S7. Dynamically adjust the mutation rate, perform mutation operations on some individuals, introduce new genes, generate new solutions through gene mutation, further increase the diversity of the population, and prevent the algorithm from falling into local optimality; S8. Merge the new individuals generated after crossover and mutation operations with the parent individuals to form a new generation population, update the iteration count, continuously evolve the population, and gradually approach the optimal solution; S9. Check whether the termination condition is met. If it is met, terminate the algorithm; otherwise, return to step S4 to continue the iteration, ensure that the algorithm ends within a reasonable time, and output the final task offloading strategy.

2. The method for solving the cloud-edge collaborative task offloading strategy according to claim 1, wherein The specific implementation steps of step S1 are as follows: S1.

1. Define a network containing several base stations through the network model. Each base station serves several users, and each base station is equipped with an MEC server to form a two-layer computing architecture of base station edge computing and cloud computing; S1.

2. Through the computing model, each user has a computing task, calculate the CPU cycles required to calculate one bit of input data, and divide the task into independent parts to be processed simultaneously at the BS and in the cloud; S1.

3. Calculate the latency and energy consumption of edge computing and cloud computing through the latency and energy consumption models.

3. The method for solving the cloud-edge collaborative task offloading strategy according to claim 2, wherein The specific implementation steps of step S2 are as follows: S2.

1. Collect the data volume and computing requirements of user tasks; S2.

2. Collect the computing capabilities of base stations and the cloud; S2.

3. Collect channel parameters, including channel bandwidth, background noise power, transmission power, and channel gain.

4. The method for solving the cloud-edge collaborative task offloading strategy according to claim 3, wherein Each candidate solution in step S3 represents the allocation of tasks to computing nodes at the BS or in the cloud.

5. A method for solving the cloud-edge collaborative task offloading strategy according to claim 4, characterized in that, The specific implementation steps of step S4 are as follows: S4.

1. Calculate the fitness value of each candidate solution through the fitness function fitness.

6. The method for solving the cloud-edge collaborative task offloading strategy according to claim 5, wherein The specific implementation steps of step S6 are as follows: S6.

1. Improve the genetic algorithm and dynamically adjust the crossover rate: S6.

2. Determine whether the maximum fitness value of the current two intersecting individuals is greater than the population average fitness value, and calculate the crossover rate respectively; S6.

3. Perform crossover operations on the selected individuals to generate new candidate solutions.

7. A method for solving the cloud-edge collaborative task offloading strategy according to claim 6, characterized in that The specific implementation steps of step S7 are as follows: S7.

1. Improve the genetic algorithm and dynamically adjust the mutation rate: S7.

2. Determine whether the fitness value of the current individual is greater than the population average fitness value, and calculate the mutation rate respectively; S7.

3. Perform mutation operations on some individuals, randomly change the task allocation situation, and introduce new genes.

8. A method for solving the cloud-edge collaborative task offloading strategy according to claim 7, characterized in that The specific implementation steps of step S8 are as follows: S8.

1. Merge the new individuals generated after crossover and mutation operations with the parent individuals to form a new generation of population; S8.

2. Update the number of iterations and prepare for the next round of iteration.

9. The method for solving the cloud-edge collaborative task offloading strategy according to claim 8, wherein The specific implementation steps of step S9 are as follows: S9.

1. Check whether the termination conditions are met, including reaching the maximum number of iterations or the fitness value converges; S9.

2. If the termination conditions are met, terminate the algorithm and output the optimal candidate solution in the current population as the final task offloading strategy.

Images