Satellite-ground network end-side cloud cooperative computing unloading method based on satellite coverage time

Abstract

The invention provides a satellite-ground network end-side cloud cooperative computing unloading method based on satellite coverage time, and belongs to the technical field of wireless communication. The invention provides an end-side cloud cooperative computing unloading method based on satellite coverage time to solve the problems that task interruption is caused by limited satellite coverage time and on-orbit life is affected by limited energy. The method comprises the following steps: firstly, predicting the remaining service time of satellites, and designing an inter-satellite collaborative unloading scheme according to whether the satellites are in a visible time window before task unloading and in an unloading process; secondly, calculating time delay and energy consumption of tasks processed at each node, and constructing an end-side cloud cooperative computing unloading model with minimum energy consumption; then, decomposing the model into an unloading decision model, a computing resource allocation model and a communication resource allocation model by using a block coordinate descent method; and finally, providing a whale algorithm based on simulated annealing and adaptive neighborhood search to solve an optimal unloading decision, and further forming an end-side cloud cooperative computing unloading scheme with minimum energy consumption.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology. Specifically, it relates to a method for offloading edge-cloud collaborative computing in satellite-ground networks based on satellite coverage time. Background Technology

[0002] The 6G-oriented space-ground converged network leverages the advantages of integrating wide-area satellite coverage, stable and efficient terrestrial networks, and flexible and intelligent edge-cloud collaboration to construct a highly efficient, flexible, and robust communication system. In this system, the core value of edge-cloud collaboration lies in the deep integration of communication, computing, and storage resources among terminals, edge nodes, and the cloud. Overall performance optimization is achieved through the rational allocation of tasks among terminals, the edge, and the cloud. Computation offloading is a crucial step in this allocation process: by migrating some or all of the terminal's computing tasks to edge nodes or the cloud, computation offloading alleviates the pressure on terminal computing power and energy consumption, while also improving task processing efficiency through network collaboration. This is one of the core mechanisms supporting the efficient operation of the space-ground converged network.

[0003] However, in the scenario of space-ground integrated network based on edge-cloud collaboration, the efficient implementation of computation offloading is still limited by the characteristics of the satellites themselves: On the one hand, satellites in space-ground integrated networks rely on solar power and have limited battery capacity. In addition, the orbital operation involves changes in the light cycle. If energy consumption is not optimized, it will accelerate energy consumption and shorten on-orbit life, affecting the continuity of network coverage. On the other hand, the dynamic changes in the coverage area caused by the orbital movement of satellite nodes lead to forced rescheduling during the mission offloading process when satellites leave the coverage area. This not only increases communication overhead but also causes a surge in energy consumption due to mission interruption and retransmission mechanisms. Therefore, edge-cloud collaboration in space-ground integrated network scenarios still needs further exploration.

[0004] Another application by the inventors' team, CN119382762A, discloses a computational offloading method for satellite-ground fusion and edge-cloud collaboration, belonging to the field of wireless communication technology. Addressing the problem of high task completion latency caused by large ground processing latency and limited satellite processing capabilities, this application proposes an edge-cloud collaborative computational offloading method by uploading edge computing to satellite. In the task modeling stage, the dependencies between the decomposed computational subtasks are modeled using linked lists and directed acyclic graphs, generating a task model that effectively reduces task processing latency. In the computational offloading stage, the task completion latency is defined based on node offloading strategies and link transmission rates. A complete offloading strategy is used to generate a latency list based on the nodes that can be offloaded from the task. Under the constraints of remaining satellite coverage time, node computing power, and link offloading bandwidth, a real-number encoded genetic algorithm based on task completion latency is proposed to find the optimal offloading point for the computational task. The resulting optimal computational offloading strategy effectively reduces latency and improves task completion rate.

[0005] However, the aforementioned method, which classifies unfinished tasks within the satellite coverage period as failures and reschedules them, intensifies competition for satellite resources within the visible time window. Meanwhile, satellite resources about to be moved out of the visible time window and those outside the visible time window remain idle. This approach not only wastes system resources but also increases system energy consumption. This method utilizes coordinated offloading of satellites both within and outside the visible time window. An inter-satellite coordinated offloading scheme is designed based on whether the satellite is within the visible time window before and during task offloading. This scheme effectively solves the problem of rescheduling unfinished tasks within the visible time window, while simultaneously integrating and utilizing satellite resources outside and about to be moved out of the visible time window, thus alleviating the pressure of competition for satellite resources within the visible time window.

[0006] To address these issues, this invention seeks protection for an edge-cloud collaborative computing offloading method for satellite-ground networks based on satellite coverage time. Addressing the problems of mission interruption due to limited satellite coverage time and on-orbit lifespan impacted by limited energy, this invention proposes an edge-cloud collaborative computing offloading method based on satellite coverage time. First, the remaining service time of satellites is predicted using satellite position information, and an inter-satellite collaborative offloading scheme is designed based on whether the satellites are within the visible time window before and during mission offloading. Second, the processing latency and energy consumption of the mission at each node are calculated to obtain the total latency and total system energy consumption during edge-cloud collaborative offloading of a single mission. An edge-cloud collaborative computing offloading model minimizing energy consumption is constructed under constraints of maximum tolerable latency, communication resources, and computing resources. Then, the model is decomposed into an offloading decision model, a computing resource allocation model, and a communication resource allocation model using a block coordinate descent method. Finally, a whale algorithm based on simulated annealing and adaptive neighborhood search is proposed to solve for the optimal offloading decision, thereby forming an edge-cloud collaborative computing offloading scheme minimizing energy consumption. Summary of the Invention

[0007] This invention aims to solve the problems of the prior art. It proposes a satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time. The technical solution of this invention is as follows:

[0008] A method for offloading edge-cloud collaborative computing in a satellite-ground network based on satellite coverage time includes the following steps:

[0009] S1: Construct a satellite-ground integrated network model based on edge-cloud collaboration, and calculate the remaining satellite coverage time and the transmission rate of the communication links between each node;

[0010] S2: Design an inter-satellite collaborative unloading scheme based on whether the satellite is within the visible time window before and during the unloading process; calculate the processing latency and energy consumption of the task at each node to obtain the total latency and total system energy consumption of a single task during edge-cloud collaborative unloading.

[0011] S3: Construct an edge-cloud collaborative computing offloading model that minimizes energy consumption, and decompose the model into an offloading decision model, a computing resource allocation model, and a communication resource allocation model using the block coordinate descent method;

[0012] S4: A whale algorithm based on simulated annealing and adaptive neighborhood search is proposed to solve the optimal offloading decision, thereby forming an edge-cloud collaborative computing offloading scheme that minimizes energy consumption.

[0013] The advantages and beneficial effects of this invention are as follows:

[0014] This invention analyzes the problems of mission interruption caused by limited satellite coverage time and the impact of limited energy on on-orbit lifespan, and proposes an edge-cloud collaborative computational offloading method based on satellite coverage time. The main innovations of this invention are: 1) designing an inter-satellite collaborative offloading scheme based on whether the satellite is within the visible time window before and during mission offloading; 2) proposing a whale algorithm based on simulated annealing and adaptive neighborhood search to solve the optimal offloading decision for edge-cloud collaborative offloading. In most studies on computational offloading and resource allocation based on satellite-ground fusion networks, the key characteristic of limited coverage time caused by high-speed satellite movement has not been considered. In the few studies involving satellite coverage time constraints, tasks not completed within the satellite coverage time are usually judged as failures and rescheduled. This approach not only wastes system resources but also increases additional energy consumption. This paper proposes an edge-cloud collaborative computational offloading method based on satellite coverage time, in which an inter-satellite collaborative offloading scheme is designed based on whether the satellite is within the visible time window before and during mission offloading. Edge-cloud collaborative offloading is then performed through inter-satellite collaboration, combined with the base station (BS) and cloud computing center, rationally utilizing the resources of the satellite-ground fusion network to minimize energy consumption. While the traditional whale algorithm has certain advantages in global optimization problems, it is prone to getting trapped in local optima when facing complex optimization scenarios with multiple constraints and high dynamics, such as task offloading in satellite networks. This invention introduces a simulated annealing mechanism, leveraging its probabilistic acceptance of inferior solutions to effectively enhance the algorithm's ability to escape local optima and avoid missing the globally optimal offloading decision due to premature convergence. Simultaneously, by combining an adaptive neighborhood search strategy, the search range can be dynamically adjusted within a set threshold based on changes in fitness. This improved hybrid algorithm exhibits higher accuracy and robustness in solving optimal offloading decisions, providing strong algorithmic support for efficient computational offloading within the edge-cloud collaborative framework. Therefore, this invention demonstrates innovation and feasibility in its solution approach. Attached Figure Description

[0015] Figure 1 This is a network model diagram constructed according to a preferred embodiment of the present invention;

[0016] Figure 2 This is a preferred embodiment of the present invention that constructs a LEO satellite and user location relationship map;

[0017] Figure 3 This is a flowchart of the satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time, as described in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0019] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0020] A method for offloading computation through edge-cloud collaborative computing in satellite-ground networks based on satellite coverage time. This method addresses the issues of mission interruption due to limited satellite coverage time and on-orbit lifespan impacted by limited energy resources, minimizing energy consumption under constraints of maximum mission latency, computational resources, and communication resources. The specific steps are as follows:

[0021] Step 1: Set network scenario parameters, including the number of nodes and resource information; import task information, including data size and maximum tolerable latency.

[0022] Step 2: Construct a satellite-ground integrated network model based on edge-cloud collaboration, and calculate the remaining satellite coverage time and the transmission rate of the communication links between each node according to the network scenario parameters;

[0023] Step 3: Design an inter-satellite collaborative unloading scheme based on whether the satellite is within the visible time window before and during the unloading process; calculate the processing latency and energy consumption of the task at each node to obtain the total latency and total system energy consumption of a single task during edge-cloud collaborative unloading.

[0024] Step 4: Under the constraints of maximum tolerable latency, communication resources, and computing resources, construct an edge-cloud collaborative computing offloading model that minimizes energy consumption. Use the block coordinate descent method to decompose the model into an offloading decision model, a computing resource allocation model, and a communication resource allocation model.

[0025] Step 5: A whale algorithm based on simulated annealing and adaptive neighborhood search is proposed to solve the optimal offloading decision, thereby forming an edge-cloud collaborative computing offloading scheme with minimal energy consumption. First, the fitness of each population is solved to update the global optimum. Second, the whale algorithm is used to explore new populations, and the population is updated using the probabilistic reception mechanism of the simulated annealing algorithm. Then, an adaptive neighborhood search strategy is used to explore better populations. Finally, iteration stops when the convergence condition is met or the maximum number of iterations is reached.

[0026] The main symbols and parameters involved in this invention and their meanings are listed in Table 1.

[0027] Table 1. Main Symbols, Parameters, and Their Meanings

[0028]

[0029]

[0030] Preferably, in the fourth step, a minimum energy consumption edge-cloud collaborative computing offloading model is constructed under the constraints of maximum tolerable latency, communication resources, and computing resources. The model is decomposed into an offloading decision model, a computing resource allocation model, and a communication resource allocation model using the block coordinate descent method.

[0031] (1) Construct an edge-cloud collaborative computing offloading model that minimizes energy consumption

[0032] Allow users to uninstall the decision matrix Communication resource allocation matrix B = [B BS B LEO B C ] T Calculate the resource allocation matrix F = [F L ,F BS ,F LEO ,F C ] T The calculation unloading model can then be expressed as:

[0033]

[0034] The constraints in the problem are as follows: C1 represents the total task latency of user n being subject to the task's maximum tolerable latency. The constraints are as follows: C2 and C3 indicate that the user adopts a complete offload, and user n can only offload the task to one of the following nodes: local, BS edge server, cloud computing center, and LEO satellite edge server. C4-C8 indicate that the sum of the computing resources allocated to the user by the user equipment, BS edge server, cloud computing center, and LEO satellite edge server cannot exceed the maximum value of their respective computing resources, and the computing resources allocated to the user by each of them are greater than or equal to 0. C9-C12 indicate that the sum of the communication resources allocated to the user by the BS edge server and LEO satellite edge server cannot exceed the maximum value of their respective communication resources, the sum of the communication resources allocated to the LEO satellite by the cloud computing center will not exceed the maximum value of its own communication resources, and the communication resources allocated by each of them are greater than or equal to 0.

[0035] (2) Model decomposition using block coordinate descent method

[0036] The edge-cloud collaborative computing offloading model is decomposed into an offloading decision model, a computing resource allocation model, and a communication resource allocation model using the block coordinate descent method:

[0037]

[0038] The offloading decision model is shown in equation (2). It is assumed that the allocation of communication resources and computing resources is fixed, and the optimal offloading decision is solved under the constraints of offloading decision and maximum tolerable delay of the task. Since the constraints C4-C12 are only related to the communication resource allocation matrix B and the computing resource allocation matrix F, they do not constitute a constraint on the computing offloading decision matrix U, so these constraints are not considered.

[0039]

[0040] The computational resource allocation model is shown in equation (3). It is assumed that the offloading decision and communication resource allocation are fixed, and the optimal computational resource allocation scheme is solved under the constraints of computational resources and the maximum tolerable delay of the task. Since constraints C2 and C3 are only related to the offloading decision matrix U, and constraints C9-C12 are only related to the communication resource allocation matrix B, these constraints are not considered.

[0041]

[0042] The communication resource allocation model is shown in equation (4). It is assumed that the offloading decision and computing resources are fixed, and the optimal communication resource allocation scheme is solved under the constraints of communication resources and the maximum tolerable delay of the task. Since constraints C2 and C3 are only related to the offloading decision matrix U, and constraints C4-C8 are only related to the computing resource allocation matrix F, these constraints are not considered.

[0043] Preferably, in the fifth step, a whale algorithm based on simulated annealing and adaptive neighborhood search is proposed to solve the optimal offloading decision, thereby forming an edge-cloud collaborative computing offloading scheme that minimizes energy consumption. First, the fitness of each population is solved, and the global optimal solution is updated. Second, the whale algorithm is used to explore new populations, and the population is updated using the probabilistic reception mechanism of the simulated annealing algorithm. Then, an adaptive neighborhood search strategy is used to explore better populations. Finally, iteration stops when the convergence condition is met or the maximum number of iterations is reached. In this subsection, fitness is defined as the total energy consumption of the system.

[0044] The algorithm first uses a Logistic chaotic mapping to generate δ whale populations, each represented as U. i(i = 1, 2, ..., δ). First, solve the resource allocation scheme for each population, calculate the fitness of each population, and update the current global optimum. Next, based on the whale algorithm, obtain new populations in stages according to random numbers and shrinkage factors, and check and repair the new populations to ensure their legitimacy. Introduce a probabilistic acceptance mechanism of simulated annealing to update the population, and accept solutions according to the temperature decay rule to escape local optima. Then, adopt an adaptive neighborhood search strategy, and perform large-scale neighborhood search or local neighborhood search according to whether the fitness change is within a threshold. In addition, a local neighborhood search is performed on the global optimum in each iteration. Finally, the iteration terminates when the maximum number of iterations is reached or the fitness change is less than the threshold for 3 consecutive times.

[0045] The Logistic mapping is a typical chaotic system that can ensure the diversity and balance of the initial population. Its iterative equation is:

[0046] x n+1 =4x n (1-x n (5)

[0047] First, a chaotic sequence is generated using the Logistic chaotic mapping. The initial value x0 is randomly selected, but fixed points such as 0, 0.5, and 1 are avoided. Then, the chaotic sequence is converted into a matrix corresponding to the number of rows and columns of the unloading decision matrix. The maximum value in each row is set to 1, and the other values ​​are set to 0. Finally, an initial population is generated.

[0048] (1) Fitness calculation

[0049] For each population, the communication resource allocation scheme and the computation resource allocation scheme must be solved before the fitness can be calculated. Before solving these two resource allocation schemes, it must be proven that the problems corresponding to the communication resource allocation model and the computation resource allocation model are convex problems, so that they can be solved directly using the CVX toolbox. Theorem 1 is used to prove that the objective function of the model is a convex function; Theorem 2 is used to prove that the constraints of the model are a convex set.

[0050] Theorem 1: For the function satisfy

[0051] f(αx+(1-α)y)≤αf(x)+(1-α)f(y) (6)

[0052] Or satisfy

[0053] Then the function f(x) is called a convex function.

[0054] Theorem 2: If set For any two points x and y in λ, the following condition holds for λ∈[0,1]:

[0055] If λx+(1-λ)y∈C(8), then set C is a convex set.

[0056] To obtain the communication and computing resource allocation scheme, the following steps are required: First, generate the initial communication resource allocation matrix and computing resource allocation matrix based on the offloading decision and using the average allocation strategy; Second, optimize the computing resource allocation matrix using CVX while fixing the communication resource allocation matrix; Third, optimize the communication resource allocation matrix using CVX while fixing the computing resource allocation matrix; Fourth, alternately optimize the communication resource allocation matrix and computing resource allocation matrix using the block coordinate descent algorithm according to the first two steps until the fitness change is within the set threshold, or the iteration stops when the maximum number of iterations is reached, thus obtaining the communication and computing resource allocation scheme.

[0057] The fitness change is defined as the difference between the fitness of the current global optimum and the fitness of the global optimum in the previous iteration, divided by the fitness of the global optimum in the previous iteration, as shown below:

[0058]

[0059] Where f pre (U best f represents the fitness of the global optimum in the previous iteration. current (U best ) represents the fitness of the current global optimal solution.

[0060] (2) New population generation based on whale algorithm

[0061] The whale optimization algorithm dominates position updates by simulating the typical behavior of humpback whales: it is divided into the following two stages depending on the generation of random numbers.

[0062] When the random number rand < 0.5, the predation phase begins:

[0063] If the contraction factor |A| < 1, the whale moves closer to the global optimum to focus on local exploitation. The position update formula is as follows:

[0064] U new =U best -A·|C·U best -U current | (10)

[0065] Among them U best U represents the globally optimal individual. current U represents the current individual. new This represents a newly generated individual. The contraction factor A = 2a·r1 - a, where a represents the decay factor from 2 to 0, and r1 is a random number in the range [0,1]. C = 2·r2, where r2 is a random weight in the range [0,1].

[0066] When the contraction factor |A|≥1, other whales are randomly selected to guide the search to enhance global exploration. The position update formula is as follows:

[0067] U new =U random -A·|C·U random -U current | (11)

[0068] Among them U random Other individual locations were randomly selected.

[0069] When the random number rand ≥ 0.5, the spiral update phase begins.

[0070] The spiral update mechanism simulates the bubble-net hunting behavior of humpback whales, generating spiral trajectories near the global optimum to achieve fine-grained neighborhood search, balancing the needs of exploration and development. The position update formula is as follows:

[0071] U new =U best +|U best -U current |·e b·l ·cos(2πl) (12)

[0072] Where b represents the spiral shape parameter, and l is a random number in the range [-1, 1], used to control the randomness of the spiral trajectory.

[0073] (3) Probabilistic receiving mechanism of simulated annealing

[0074] The probabilistic acceptance mechanism of simulated annealing injects the algorithm with the ability to escape local optima. When updating individual optimal solutions, the algorithm not only prioritizes accepting better solutions but also sets a probabilistic acceptance rule based on temperature decay for worse solutions. The probability formula is as follows:

[0075]

[0076] Where f(U) new ) represents the fitness of the newly formed population, f(U) current ) represents the fitness of the current population, T represents the initial temperature, α represents the temperature decay factor, and r represents the number of iterations.

[0077] As the temperature decreases with the iteration exponent, the early high-temperature phase allows for a higher probability of accepting suboptimal solutions to expand the exploration scope, while the later low-temperature phase reduces the acceptance probability to focus on development. This mechanism enables the algorithm to broadly explore the solution space in the early stages of iteration and gradually converge to high-quality solutions in the later stages.

[0078] (4) Adaptive Neighborhood Search Strategy

[0079] The adaptive neighborhood search strategy intelligently adjusts the search intensity based on whether the fitness change is within a set threshold, further optimizing the search process. A parameter ω represents whether the fitness change is within the threshold. When ω = 0, it indicates the fitness change is not within the threshold, and a large-scale neighborhood search is performed to quickly explore a broad solution space. When ω = 1, it indicates the fitness change is within the threshold, and only a local neighborhood search is performed to refine the optimal solution. The large-scale neighborhood search generates candidate solutions by randomly swapping any two rows in the unloading decision matrix, executing μ times. If the optimal solution in the candidate solution is better than the original solution, the initial solution is replaced. The local neighborhood search randomly selects ν rows and swaps the 0-1 positions in these rows to generate candidate solutions, i.e., changing the unloading position of a single user. If the optimal solution in the candidate solution is better than the original solution, the original solution is replaced. After each iteration, a local neighborhood search is performed on the global optimum to generate a new solution and update the global optimum, ensuring its neighborhood is fully explored and avoiding getting trapped in local optima.

[0080] (5) Iteration termination condition

[0081] The algorithm employs a dual strategy for iterative termination, combining a maximum iteration limit with a convergence detection mechanism. Firstly, when the number of iterations reaches a preset maximum, the algorithm terminates forcibly, ensuring controllable computation. Secondly, a convergence detection mechanism is introduced; if the fitness change is less than a set threshold for three consecutive iterations, the algorithm is considered converged and terminates early. This design avoids infinite iteration due to getting trapped in local optima while stopping promptly when the solution tends to stabilize, balancing solution quality and computational efficiency.

[0082] The model involved in this invention is as follows:

[0083] 1. Network Model

[0084] The primary application of this invention is in space-ground integrated networks, such as... Figure 1 As shown. This network model consists of n users, m BSs, i LEO satellites, and one cloud computing center. The set of users is represented as... Each BS is equipped with one edge server, and the BS set is represented as follows: Each LEO satellite is also equipped with an edge server, and the set of LEO satellites is represented as follows: Its communication range can cover areas without terrestrial networks, but its coverage time is limited due to orbital motion. As a computing power hub, the cloud computing center possesses massive computing resources and can handle highly complex, non-latency-sensitive computing tasks.

[0085] The system operates based on a discrete time-slot model, where each time slot τ maintains a constant length, and the set of time slots is represented as follows: At the start of each time slot, each user will generate a task, which can be represented by a triple. Let D represent this, where D n (t) represents the data size of the task (in bits), C n (t) represents the number of CPU cycles required to compute 1 bit of data (in CPU cycles / bit). This indicates the maximum tolerable delay for the task (in seconds). Additionally, using... The four parameters represent the user's uninstallation decision, indicating whether to uninstall to the local machine, the BS edge server, the LEO satellite edge server, or the cloud computing center. A parameter equal to 1 indicates uninstallation to that node, while a parameter equal to 0 indicates uninstallation not to that node.

[0086] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions.

[0087] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0088] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0089] The above embodiments should be understood as illustrative only and not as limiting the scope of protection of the present invention. After reading the description of the present invention, those skilled in the art can make various alterations or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A method for offloading edge-cloud collaborative computing in a satellite-ground network based on satellite coverage time, characterized in that, Includes the following steps: S1: Construct a satellite-ground integrated network model based on edge-cloud collaboration, and calculate the remaining satellite coverage time and the transmission rate of the communication links between each node; S2: Design an inter-satellite collaborative unloading scheme based on whether the satellite is within the visible time window before and during the unloading process; calculate the processing latency and energy consumption of the task at each node to obtain the total latency and total system energy consumption of a single task during edge-cloud collaborative unloading. S3: Construct an edge-cloud collaborative computing offloading model that minimizes energy consumption, and decompose the model into an offloading decision model, a computing resource allocation model, and a communication resource allocation model using the block coordinate descent method; S4: A whale algorithm based on simulated annealing and adaptive neighborhood search is proposed to solve the optimal offloading decision and form an edge-cloud collaborative computing offloading scheme with minimal energy consumption.

2. The satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time according to claim 1, characterized in that, Step S1 involves constructing a satellite-ground fusion network model based on edge-cloud collaboration, calculating the remaining satellite coverage time, and the transmission rate of the communication links between each network node. Specifically, this includes: The network model consists of three parts: user terminals, terrestrial network, and satellite network. The user terminals include various sensors and smartphones, which are mainly responsible for data acquisition and task generation. The terrestrial network consists of cloud computing centers and base stations (BS) equipped with edge servers, which work together to process massive amounts of terrestrial tasks. The satellite network consists of low-Earth orbit (LEO) satellites equipped with edge servers. By deploying edge computing on satellites to expand on-board computing resources, the satellite network provides computing services to users. Calculate the remaining satellite service time: The remaining coverage time between LEO satellite i and ground users is: Among them, v i The velocity of LEO satellite i is represented by l. i This indicates the coverage arc length of LEO satellite i to ground users; Construct a channel model and calculate the transmission rate: Three channel models are considered: terrestrial channel model, satellite-to-terrestrial channel model, and inter-satellite channel model. The terrestrial channel model is used for communication between users and the BS, the satellite-to-terrestrial channel model is used for communication between users and LEO satellites and between LEO satellites and cloud computing centers, and the inter-satellite channel model is used for communication between LEO satellites. Inter-satellite interference and terrestrial user interference are ignored.

3. The satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time according to claim 2, characterized in that, The ground channel model, satellite-to-ground channel model, and inter-satellite channel model are as follows: 1) Ground channel model Non-line-of-sight transmission dominates terrestrial communication, and the corresponding channel is modeled as a Rayleigh channel; the transmission rate between user n and BSm is given by the following equation: Among them B n,m (t) represents the bandwidth allocated by BSm to user n. It is the transmit power of user n, |h n,m | 2 σ is the channel gain between user n and BSm. 2 This represents the Gaussian white noise power of the link; 2) Satellite-to-Ground Channel Model Users communicate with LEO satellites via Ka-band satellite-to-ground links; because line-of-sight transmission dominates these links, the corresponding channels are modeled as Ricean channels; the transmission rate between user n and LEO i is expressed as: Among them B n,i (t) represents the bandwidth allocated to user n by LEO satellite i. This represents the transmit power of user n. This represents the antenna gain of user n. The receiving antenna gain of the LEO satellite is represented by |h n,i | 2 Indicates the channel gain of the link; When a user offloads a task to the cloud computing center, the task needs to be offloaded to a LEO satellite first, and then forwarded by the LEO satellite to the ground-based cloud computing center. Therefore, the uplink transmission rate from user n to LEO satellite i is shown in equation (3), and the downlink transmission rate from LEO satellite i to the cloud computing center can be expressed as: Among them B i,c (t) represents the bandwidth allocated to LEO satellite i by the cloud computing center. This indicates the transmission power of LEO satellite i. This indicates the antenna gain of LEO satellite i. The receiving antenna gain of the cloud computing center, |h i,c | 2 Indicates the channel gain of the link; 3) Inter-satellite channel model Inter-satellite links are primarily line-of-sight transmissions, and therefore are modeled as Ricean channels. The transmission rate of inter-satellite links is similar to that in equation (3), and can be expressed as: Among them B k,i (t) represents the bandwidth allocated to LEO satellite k by LEO satellite i. This indicates the transmission power of LEO satellite k. This represents the antenna gain of LEO satellite k. The receiving antenna gain of LEO satellite i is represented by |h k,i | 2 This indicates the channel gain of the link.

4. The satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time according to claim 3, characterized in that, Step S2 involves designing an inter-satellite collaborative unloading scheme based on whether the satellite is within the visible time window before and during task unloading; calculating the processing latency and energy consumption of the task at each node to obtain the total latency and total system energy consumption during edge-cloud collaborative unloading for a single task. The specific details are as follows: (1) Inter-satellite cooperative offloading scheme 1) If the target satellite, i.e. the satellite that will ultimately process the task, is within the visible time window before the task is unloaded, and is also within the visible time window when the target satellite finishes processing the task, then the target satellite will directly send the processing result back to the user. 2) If the target satellite is within the visible time window before the task is unloaded, but has moved out of the visible time window after the task is completed, the processing result needs to be forwarded to the user via the visible satellite; 3) If the target satellite is outside the visible time window before the task is unloaded, the task needs to be forwarded via the visible satellite, and then the processing result needs to be forwarded to the user via the visible satellite. According to the above scheme, when a user offloads a task to a satellite, the calculation of latency and energy consumption should take the following into account: For the first and second offloading schemes, it is necessary to calculate the latency from task transmission to completion, and then compare it with the remaining service time of the satellite to determine which offloading scheme to use to calculate latency and energy consumption; For the third offloading scheme, latency and energy consumption can be calculated directly according to the offloading process. The time delay from task transmission to processing completion represents the period from the start of task transmission to the completion of task processing on the target satellite, as shown below: in, This indicates that the mission is offloaded to the LEO satellite, f n,i (t) represents the computing resources allocated to user n by LEO satellite i, D n (t) represents the task size, C n (t) represents the number of CPU cycles required to compute 1 bit of data, s n,i (t) represents the distance between user n and LEO satellite i, and c represents the speed of light; The delay for user n to offload the task to LEO satellite i is: in This represents the time delay from task transmission to computation completion. Indicates the remaining satellite service time, s n,g (t), s n,k (t) represents the distances between user n and LEO satellites g and k, respectively, and s k,i (t), s g,i (t) represents the distances between LEO satellites k and g and LEO satellite i, respectively; the first row of formulas indicates that before task unloading, target satellite i is outside the visible time window, and in this case, it needs to be forwarded to the target satellite via visible satellite k and then forwarded back to the target satellite via visible satellite g; the second row of formulas indicates that target satellite i is within the visible time window both before and after task unloading, and in this case, target satellite i can communicate directly with the user; the third row of formulas indicates that target satellite i is within the visible time window before task unloading, but outside the visible time window after task completion, and in this case, it needs to be forwarded back to the target satellite via visible satellite k. To facilitate the derivation of the formula, equation (7) is modified as follows: in This indicates that the target satellite was within the visible time window before the mission was unloaded. This indicates that the target satellite was outside the visible time window before the mission unloading; they can only take the values ​​0 or 1, and This indicates that target satellite i will complete its task within the visible time window. This indicates that target satellite i is outside the visible time window when it finishes its task; they can only take the values ​​0 or 1, and According to the inter-satellite coordination offloading strategy, the energy consumption for user n to offload the task to LEO satellite i is: in and Let represent the transmit power of user n and LEO satellite k, respectively; where κ represents the energy consumption factor of the device's computing task, which is related to the physical structure of the device. (2) Local computing When performing local computation, only the latency and energy consumption of task processing need to be considered. Therefore, the latency for user n to process the task locally is: in Indicates local computation, f n (t) represents the computing resources of user n; The energy consumption for local processing is: (3) Unload to BS When a user offloads tasks to the BS, the latency in completing these tasks consists of two parts: the latency of the user transmitting the tasks and the latency of the BS processing the tasks; the latency for user n to offload tasks to the BS edge server m is: in This indicates that the task is unloaded to the BS, f n,m (t) represents the computing resources allocated by the BS edge server m to user n; The energy consumption for user n to offload the task to the BS edge server m is: (4) Unload to the cloud computing center When a task is offloaded to the cloud computing center for processing, it is relayed to the center via LEO satellites. The latency of the entire process includes transmission latency, propagation latency, and processing latency, while the energy consumption includes transmission energy consumption and processing energy consumption. The latency for user n to offload the task to the cloud computing center is: in, This indicates that the task is offloaded to the cloud computing center, s i,c (t) represents the distance between LEO satellite i and the cloud computing center, f n,c (t) represents the computing resources allocated to user n by the cloud computing center; The energy consumption for user n to offload tasks to the cloud computing center is: (5) Total latency and total system energy consumption of single task edge-cloud collaborative offloading The total latency of a single task's edge-cloud collaborative unloading can be expressed as: The total energy consumption of the system can be expressed as:

5. The satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time according to claim 4, characterized in that, Step S3 constructs a minimum energy consumption edge-cloud collaborative computing offloading model. The model is decomposed into an offloading decision model, a computing resource allocation model, and a communication resource allocation model using the block coordinate descent method. The specific details are as follows: (1) Construct an edge-cloud collaborative computing offloading model that minimizes energy consumption Allow users to uninstall the decision matrix in, Indicates the number of users. Indicates the number of BS. U represents the number of LEO satellites. n Represents the unloading decision of user n; the communication resource allocation matrix B = [B BS B LEO B C ] T Among them, B BS This represents the communication resource allocation matrix of B and B. LEO B represents the communication resource allocation matrix for LEO satellites. C The communication resource allocation matrix of the cloud computing center; the computing resource allocation matrix F = [F L F BS ,F LEO ,F C ] T , of which F L F represents the computing resource allocation matrix for user equipment. BS F represents the computational resource allocation matrix of BS. LEO F represents the computational resource allocation matrix for LEO satellites. C The computational resource allocation matrix of the cloud computing center is represented; therefore, the computational offloading model in step S5 can be described as follows: in, This represents the maximum tolerable latency for tasks generated by user n. These indicate whether the task is offloaded to the local machine, BSM, LEO satellite i, and cloud computing center, respectively. These represent the maximum computing resources of user equipment n, BSm, LEO satellite i, and cloud computing center, respectively. B c,max , C1 represents the maximum communication resources of BSm, LEO satellite i, and cloud computing center, respectively; C2 represents the total task latency of user n subject to the maximum tolerable latency of the task. The constraints are as follows: C2 and C3 indicate that the user adopts a complete offload, and user n can only offload the task to one of the following nodes: local, BS edge server, cloud computing center, and LEO satellite edge server; C4-C8 indicate that the sum of the computing resources allocated to the user by the user equipment, BS edge server, cloud computing center, and LEO satellite edge server cannot exceed the maximum value of their own computing resources, and the computing resources allocated to the user by them are all greater than or equal to 0; C9-C12 indicate that the sum of the communication resources allocated to the user by the BS edge server and LEO satellite edge server cannot exceed the maximum value of their own communication resources, the sum of the communication resources allocated to the LEO satellite by the cloud computing center will not exceed the maximum value of its own communication resources, and the communication resources allocated by them are all greater than or equal to 0. (2) Model decomposition using block coordinate descent method Using the block coordinate descent method, the edge-cloud collaborative computing offloading model is decomposed into an offloading decision model, a communication resource allocation model, and a computing resource allocation model. The offloading decision model optimizes offloading decisions through fixed communication and computing resource allocation schemes, thereby determining the optimal execution location of tasks and solving the discrete optimization problem of where tasks should be processed, thus balancing the load and avoiding single-point bottlenecks. The communication resource allocation model optimizes communication resource allocation through fixed offloading decisions and computing resource allocation schemes, thereby optimizing data transmission efficiency and providing communication links for task offloading. The computing resource allocation model optimizes computing resource allocation through fixed offloading decisions and communication resource allocation schemes, thereby minimizing processing energy consumption under existing offloading decisions and communication resource allocations.

6. The satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time according to claim 1, characterized in that, In step S4, a whale algorithm based on simulated annealing and adaptive neighborhood search is proposed to solve the optimal offloading decision, thereby forming an edge-cloud collaborative computing offloading scheme that minimizes energy consumption. The fitness is defined as the total energy consumption of the system, specifically including: The algorithm first generates a whale population using a Logistic chaotic mapping, where the population represents the unloading decision matrix. It proves that the communication and computing resource allocation model is a convex problem, obtains a resource allocation scheme and calculates fitness through alternating optimization, and updates the global optimum. Then, based on the whale algorithm, it obtains a new population in stages according to random numbers and shrinkage factors, and checks and repairs the new population to ensure its legitimacy. It introduces a probabilistic acceptance mechanism of simulated annealing to update the population, and accepts solutions according to the temperature decay rule to escape local optima. It adopts an adaptive neighborhood search strategy, performing a large-scale neighborhood search or a local neighborhood search depending on whether the fitness change is within a threshold. In addition, a local neighborhood search is performed on the global optimum in each iteration. The iteration terminates when the maximum number of iterations is reached or the fitness change is less than the threshold for three consecutive times.

7. A storage medium that internally stores a computer program, characterized in that, When the computer program is read by the processor, it executes the satellite-ground network edge-cloud collaborative computing offloading method based on satellite coverage time as described in any one of claims 1 to 6.

Images