6G air-space-ground integrated network resource allocation device based on evolutionary game

Abstract

The invention discloses a 6G air-space-ground integrated network resource allocation device based on an evolutionary game, and belongs to the technical field of calculation, reckoning or counting. According to the device, the problems of unbalanced resource allocation, task processing delay increase and the like caused by the fact that a plurality of tasks fight for limited computing resources at the same time in the 6G space-air-ground integrated network are considered, and the device for task unloading and resource allocation of the edge server is provided so as to meet the requirements of ground users for the tasks. According to the device, firstly, a 6G air-space-ground integrated calculation unloading network model is constructed; secondly, defining an optimization target and a constraint condition for minimizing task execution time delay; and finally, a task unloading and resource allocation problem is solved by adopting a combined task unloading and resource allocation algorithm. A simulation result shows that the algorithm has good network performance.

Description

Technical Field

[0001] This invention relates to communication technology, specifically disclosing a 6G integrated air-space-ground network resource allocation device based on evolutionary game theory, belonging to the technical field of calculation, estimation, or counting. Background Technology

[0002] Future 6G mobile communication will build a Space-Air-Ground integrated network (SAGIN) communication system. This system combines the advantages of space, air, and ground infrastructure to further meet the communication needs of various users anytime, anywhere, achieving the goal of ubiquitous mobile communication network coverage. Furthermore, the SAGIN network will promote the development of emerging applications such as the Internet of Things (IoT), the Internet of Vehicles (IoV), and the Industrial Internet. Therefore, the SAGIN network has profoundly changed people's lives, and its research has significant practical implications.

[0003] However, the varying sizes and computational demands of tasks initiated by ground user equipment lead to differences in task latency. Due to the limited computing resources of edge servers, uneven resource allocation can occur. Some devices may be overloaded due to excessive tasks, while others may be idle due to insufficient tasks, resulting in decreased overall resource utilization and impacting system efficiency and performance. Furthermore, when multiple tasks are simultaneously offloaded to the same device, exceeding its processing capacity can increase latency. Increased resource contention can also cause conflicts between tasks, potentially preventing the timely processing of some computationally intensive tasks. Therefore, offloading different types of tasks generated by user equipment to satellite edge servers, UAV edge servers, or base station edge servers to minimize latency is a significant challenge. Summary of the Invention

[0004] This invention addresses the problems of unbalanced resource allocation and increased task processing latency caused by multiple tasks simultaneously competing for limited computing resources. It designs a 6G integrated air-space-ground network resource allocation device based on evolutionary game theory. First, a 6G integrated air-space-ground computing offloading network model is constructed. Second, an optimization objective and constraints for minimizing task execution latency are defined. Finally, a joint task offloading and resource allocation algorithm is used to solve the task offloading and resource allocation problem.

[0005] The 6G air-space-ground integrated network resource allocation device based on evolutionary game theory of the present invention includes the following three steps:

[0006] 1) Construct a 6G integrated space-air-ground computing offloading network model. This network model consists of three layers: a space-based network, a space-based network, and a ground-based network. The space-based network contains N LEO satellites configured with edge servers, denoted as STS. The set of STSs can be represented as STS = {STS1, STS2, ..., STS}. n STS N |n∈N},STS n Let $n$ represent the nth LEO satellite, whose communication range covers the entire ground-based network. It can communicate with user equipment and handle computational tasks offloaded by user equipment. The airborne network consists of M unmanned aerial vehicles (UAVs), each equipped with a communication module and an edge server, denoted as $UAS$. The set of $UAS$ can be represented as $UAS = {UAS1, UAS2, ..., UAS}$. m , ..., UAS M |m∈M},UAS m Let r represent the effective communication radius of the m-th UAV. UAS It can provide communication / computing services for user equipment; the ground network consists of I user equipment (USE) and J base stations (BS), each BS deploying an edge server; the set of USE can be denoted as USE = {USE1, USE2, ..., USE...} i , ..., USE I |i∈I}, where USE i Let r represent the effective communication radius of the i-th user equipment. USE (r UAS >r BS >r USE It can send task offloading requests to BS, UAS, and STS; the set of BS can be denoted as BS = {BS1, BS2, ..., BS...} i , ..., BS J |j∈J}, where BS j Let r represent the effective communication radius of the j-th base station. BS It can provide computing resources for the tasks of USEs;

[0007] In this network model, only USE spontaneously generates computational tasks; STS, BS, and UAS can only execute tasks but do not spontaneously generate computational tasks. Within each time slot, USE... i Generate a computational task that needs to be processed, which can be represented by a triplet. <D i C i T i max > indicates that D i Indicates the size of the task data, C iT represents the computational load of a task, i.e., the computing resources required to complete the task. i max Indicates the maximum acceptable time to complete the task; USE i Tasks can be offloaded to edge servers in various network zones for processing; in the space-based zone, USE i Tasks can be offloaded to STS for execution via the USE-STS communication link; in the airborne region, USE i Tasks can be offloaded to the UAS for execution via the USE-UAS communication link; in the ground-based area, USE i Tasks can be offloaded to the BS for execution via the USE-BS communication link; the execution latency of a task consists of transmission latency and processing latency.

[0008] 2) Define the optimization objective and constraints to minimize task execution latency. Centering on the USE, the computational offloading process in the 6GSAG-CO network model is studied, and interference in the USE communication process is further discussed: First, interference between the USE-BS and USE-UAS communication links is considered; second, interference also exists between the STS-STS, UAS-UAS, and BS-BS communication links, but these are allocated to different frequency bands than the USE transmission links.

[0009] Tasks generated by USE are offloaded to the STS server via a wireless channel for execution. i With STS n The channel gain for information transmission between them can be expressed as:

[0010]

[0011] Among them, υ i，n ε represents the complex Gaussian variable representing Rayleigh fading. i,n The shadow fading indicates a log-normal distribution, d i,n Indicates USE i With STS n The distance between them Indicates the path loss index;

[0012] Communication between USE and STS uses the Ka band, which does not cause interference with other frequency bands; USE i To STS n The achievable task transmission rate is expressed as:

[0013]

[0014] Among them, B i Indicates USE i The channel bandwidth, pi Indicates USE i The transmit power, σ 2 Indicates Gaussian white noise;

[0015] Communication links between UAS and between BS share the same frequency band as USE-UAS. i To UAS m Signal-to-interference-plus-noise ratio γ i,m Represented as:

[0016]

[0017] in, Indicates USE i Offload the task to UAS within the communication range m Number of uses; Indicates interference caused by other USE-UAS communication links; G i，m Indicates USE i With UAS m The channel between; assuming the position of the UAS is quasi-fixed, i.e., the UAS is in a hovering state, USE i With UAS m The channel between them can be constructed as a Rician fading channel model, represented as:

[0018]

[0019] in, Denotes the line-of-sight (LoS) channel component, d i,m Indicates USE i To UAS m The distance between them, where λ represents the Rician factor. The path loss exponent, |g, represents the Loss of Path (LoS) in Rician fading. i，m |=1 indicates the LoS channel coefficient; G represents the non-line-of-sight (NLoS) channel component. i,m The NLoS channel coefficients represent the zero-mean, unit-variance Gaussian fading channel. This represents the path loss exponent of NLoS in Rician fading; therefore, USE i To UAS m The achievable task transmission rate is expressed as:

[0020] R i,m =B i log2(1+γ i,m (5)

[0021] USE can offload tasks to the BS. iReceived from base station BS j The signal-to-noise ratio is:

[0022]

[0023] in, Indicates interference caused by other USE-BS links; G i，j Indicates USE i With BS i The channel between them is represented as:

[0024]

[0025] Where g0 represents the channel power gain at a reference distance d0 = 1, d i，j Indicates USE i To BS j The distance between them; therefore, USE i To BS j The achievable task transmission rate is expressed as:

[0026] R i,j =B i log2(1+λ i，j (8)

[0027] When USE i Offload computing tasks to STS n At that time, the transmission latency of the computation task transmitted to the STS server via the wireless channel It can be represented as:

[0028]

[0029] Processing latency required to process tasks on the STS edge server Represented as:

[0030]

[0031] Among them, f i,n This indicates the computing resources allocated by STS to the task;

[0032] Therefore, when the task is unloaded to STS n When the task is executed on a server, the execution latency can be expressed by the formula:

[0033]

[0034] in, Represents binary decision; when The time indicates in STS n Tasks are processed on the server; This indicates that the task was not uninstalled to STS.n On the server;

[0035] When USE i Offload computing tasks to UAS m At that time, the computational task is transmitted via the wireless channel with a delay It can be represented as:

[0036]

[0037] Processing latency required to process tasks on the UAS edge server Represented as:

[0038]

[0039] Among them, f i,u This indicates the computing resources allocated by the UAS to the task;

[0040] Therefore, when the task is unloaded to the UAS m When the task is executed on a server, the execution latency can be expressed by the formula:

[0041]

[0042] in, Represents binary decision; when The time indicates in UAS m The server processes tasks; This indicates that the task was not uninstalled to the UAS. m Processing is performed on the server;

[0043] When USE i Offload computing tasks to BS j At that time, the computational task is transmitted via the wireless channel with a delay It can be represented as:

[0044]

[0045] BS j Computational latency required to execute tasks on the server Represented as:

[0046]

[0047] Among them, f i，j BS j Computational resources allocated to the task;

[0048] Therefore, when the task is unloaded to BS j When the task is executed on a server, the execution latency can be expressed by the formula:

[0049]

[0050] in, Represents binary decision; when The time indicates in BS j The server processes tasks; when This indicates that it is not in BS. j Tasks are processed on the server;

[0051] In summary, USE i The total execution delay of the task can be expressed as:

[0052]

[0053] The optimization objective is to minimize the execution latency of the task under the following constraints;

[0054] Unloading decision constraint: All user devices can only choose one unloading method for their tasks;

[0055]

[0056] Unloading decision value constraints: The unloading decision of the edge server of each network area for user tasks is a Boolean variable, which can only take the value 0 or 1;

[0057]

[0058] Execution latency constraint: The task execution latency must not exceed the maximum acceptable completion latency of the user equipment for the task;

[0059]

[0060] Computational resource allocation constraints: The computational resources allocated to user tasks by STS, UAS, and BS edge servers cannot exceed the maximum allocatable computational resources of the edge server, i.e.:

[0061]

[0062]

[0063] in, This indicates the maximum computing resource allocation capability of the STS; This indicates the maximum computing resource allocation capacity of the UAS; This indicates the maximum computing resource allocation capability of the BS.

[0064] The objective function of the optimization problem can be specifically expressed as:

[0065]

[0066] 3) A joint task unloading and resource allocation algorithm is used to solve the task unloading and resource allocation problem. First, the initial resource allocation result is substituted into formula (25) to obtain a subproblem related only to the task unloading decision, which is then solved using the particle swarm optimization algorithm. Then, the obtained task unloading decision is substituted into formula (25) to obtain a subproblem related only to resource allocation, which is then solved using the evolutionary dynamic game resource allocation algorithm. Finally, the optimal task unloading decision and resource allocation scheme are obtained.

[0067] The optimization problem (25) is transformed into a sub-optimization problem P1, denoted as:

[0068]

[0069] The PSO algorithm is applied to solve the complex task offloading strategy optimization subproblem P1, resulting in a Particle Swarm Task Offloading Optimization Algorithm (PSTOA). In each iteration of this algorithm, each particle continuously adjusts its position based on its velocity and location. By introducing a nonlinear inertia weighting factor, the particle's velocity is continuously adjusted according to the current iteration number, thereby constantly changing the particle's current search area.

[0070] In the PSTOA algorithm, multiple particles are randomly generated as a swarm, and the size of the swarm is represented by Q. Each particle represents a possible unloading decision scheme. Binary encoding is used to describe the task unloading decision, using... This represents all possible task unloading decisions.

[0071]

[0072] These represent the offloading decision sets for STS, UAS, and BS, respectively.

[0073] The fitness function measures the quality of a particle. Since the PSO algorithm is only applicable to unconstrained optimization problems, a penalty function needs to be added to the fitness function. The penalty function can be expressed as:

[0074]

[0075] The fitness function with added penalty function constraints is:

[0076]

[0077] Where τ is the penalty factor.

[0078] Let V be the velocity and position of the q-th particle. q and xq The update equations for its velocity and position are expressed as follows:

[0079]

[0080] in, This represents the velocity of particle q in the k-th iteration. It is the position of particle q in the kth iteration, w k It is the inertia weight factor at the k-th iteration. and Represents individual and group learning factors. and It is a random number within the range [0, 1]. and It represents the individual optimal position and the group optimal position of particle q in the kth iteration.

[0081] A nonlinear inertia weight decreasing method is introduced to improve the traditional fixed inertia weight. The nonlinear inertia weight factor w is... k The formula is expressed as follows:

[0082]

[0083] Among them, w k，max w k,min k represents the maximum and minimum values ​​of the inertia weighting factor. max φ represents the maximum number of iterations, and φ is a nonlinear constant that controls the shape of the decreasing curve.

[0084] Once the optimal task unloading decision κ is found, it is substituted into the optimization problem (25) and transformed into a sub-optimization problem P2, denoted as:

[0085]

[0086] This invention designs an Evolutionary Dynamic Game Computational Resource Disturbed Algorithm (EGCRD). First, using complex network theory, each task issued by a user device is considered as an independent node, and the set of task nodes can be represented as A = {A1, A2, ..., A...}. h A H H represents the total number of computing tasks offloaded to their respective servers. Nodes select appropriate strategies for resource allocation based on their current resource needs and network conditions, ultimately arriving at the optimal resource allocation mechanism based on the evolution results. The strategy evolution of different task nodes will collectively affect the overall resource allocation and the total task latency.

[0087] Evolutionary game theory generally consists of four parts: participants, groups, strategy sets, and game payoffs. For the problem of 6G SAGIN computational resource allocation, the components of the constructed evolutionary dynamic game computational resource allocation model are described below.

[0088] Participants: All task nodes.

[0089] Group: A group is a collection of tasks that are unloaded onto the same server.

[0090] Strategy: Each node has two game strategies to choose from: cooperation (C) and betrayal (D). Nodes that adopt the cooperation strategy will contribute resources to the common pool, while nodes that adopt the betrayal strategy will not contribute resources to the common pool.

[0091] Game payoff: Define node A h (h∈H) The supply and demand difference of computational resources in round t ψ h (t) represents the resource α allocated to the node in round t-1. h (t-1) and idle resources β in round t-1 h The sum of (t-1) and resource demand ξ h The difference between (t) is used to calculate the resource supply and demand gap ψ. h The formula for calculating (t) is shown below:

[0092] ψ h (t)=α h (t-1)+β h (t-1)-ξ h (t) (39)

[0093] Here, idle resources refer to the number of resources remaining in the common pool after the game ends. If t = 0, then the resources allocated in the previous round and the resource demand are 0, and the idle resources are the initial resources f. initial If resources are invested in this round of the game, then the idle resources in this round are the resource requirements for this round. If no resources are invested in this round, then the idle resources are the total resources from the previous round. Therefore, the idle resources in round t can be represented as:

[0094]

[0095] When the supply and demand difference of computing resources is ψ h When ψ(t) is greater than 0 and the current node adopts a cooperative game strategy, its allocated computing resources exceed its demand, and the node tends to allocate the excess computing resources to the pool for sharing with others; when the supply and demand difference of computing resources is ψ... h When (t) is less than 0 or the current node adopts a betrayal strategy, its allocated computational resources are less than its resource requirements, and the node will choose not to invest any resources. Therefore, the formula for calculating the investment cost is as follows:

[0096]

[0097] The profit calculation for the EGCRD algorithm is as follows: Node A h After engaging in a game with all other nodes, the total resources in the common pool are calculated, and then the total resources are evenly distributed among all participating nodes. When node A... h If a cooperative strategy is adopted, then node A h The number of computational resources that can be allocated, α h，C(t) and the resulting efficiency gains ∏ h,C (t) is shown below:

[0098]

[0099] ∏ h，C (t)=|ψ h (t)|-|ξ h (t)-α h，C (t)| (43)

[0100] Node A h If the betrayal strategy is adopted, then node A h The number of computational resources that can be allocated, α h,D (t) and the resulting efficiency gains ∏ h,D (t) is shown below:

[0101]

[0102] ∏ h,D (t)=|ψ h (t)|-|ξ h (t)-α h,D (t)| (45)

[0103] Where Z is node A h All participating game groups, z being one of the game groups, |z| being the number of nodes participating in the z-group game, x∈Ω z This indicates that x is a member participating in the z-group game. Represents node A h The amount of computational resources invested in game group z.

[0104] All nodes use the Fermi mechanism for synchronization, randomly selecting a neighbor and simultaneously replicating its strategy and node type. At time t, node A... h Imitating its neighbor A v The probability of (v≠h) is:

[0105]

[0106] Among them, s h(t+1)→s v (t) represents game node A h To his neighbor A v The strategy used for replication is s v (t) represents node A v The strategy at time t, s h (t+1) represents node A h The policy at time t+1, where ω represents the noise factor, is used to characterize the uncertainty of the node when making a choice. By introducing this uncertainty, we can better simulate the state of a node making a policy choice in reality when information is incomplete. When ∏ h (t)<∏ v At time (t), the neighbor node selection strategy brought more benefits to node A. h Will tend to learn node A v The greater the difference in payoff between the chosen game strategies, the higher the probability of imitation and learning; when π h (t)>∏ v At time (t), node A h The chosen strategy has already brought more benefits, so it will tend not to change its game strategy. Although it may still imitate and learn, the probability is relatively small.

[0107] The present invention adopts the above technical solution and has the following beneficial effects: The present invention comprehensively considers the task offloading and resource allocation problems in the 6G air-space-ground integrated network, and optimizes the problem with the goal of minimizing task execution latency, thereby improving network performance. Attached Figure Description

[0108] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0109] Figure 1 A network model for 6G integrated air-space-ground computing offloading:

[0110] Figure 2 A schematic diagram illustrating the joint optimization algorithm for task unloading and resource allocation;

[0111] Figure 3 This is a diagram comparing the number of tasks with the task execution latency.

[0112] Figure 4 This is a diagram comparing transmission power and task execution latency. Detailed Implementation

[0113] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0114] 1) Figure 1This is a 6G integrated space-air-ground computing offloading network model. The network model comprises three layers: a space-based network, a space-based network, and a ground-based network. The space-based network contains N LEO satellites configured with edge servers, denoted as STS. The set of STSs can be represented as STS = {STS1, STS2, ..., STS}. n ..., STS N |n∈N},STS n Let $n$ represent the nth LEO satellite, whose communication range covers the entire ground-based network. It can communicate with user equipment and handle computational tasks offloaded by user equipment. The airborne network consists of M unmanned aerial vehicles (UAVs), each equipped with a communication module and an edge server, denoted as $UAS$. The set of $UAS$ can be represented as $UAS = {UAS1, UAS2, ..., UAS}$. m , ..., UAS M |m∈M},UAS m Let r represent the effective communication radius of the m-th UAV. UAS It can provide communication / computing services for user equipment; the ground network consists of I user equipment (USE) and J base stations (BS), each BS deploying an edge server; the set of USE can be denoted as USE = {USE1, USE2, ..., USE...} i , ..., USE I |i∈I}, where USE i Let r represent the effective communication radius of the i-th user equipment. USE (r UAS >r BS >r USE It can send task offloading requests to BS, UAS, and STS; the set of BS can be denoted as BS = {BS1, BS2, ..., BS...} j , ..., BS J |j∈J}, where BS j Let r represent the effective communication radius of the j-th base station. BS It can provide computing resources for the tasks of USEs;

[0115] In this network model, only USE spontaneously generates computational tasks; STS, BS, and UAS can only execute tasks but do not spontaneously generate computational tasks. Within each time slot, USE... i Generate a computational task that needs to be processed, which can be generated by a triple <D i C i T i max > indicates that D i Indicates the size of the task data, C iT represents the computational load of a task, i.e., the computing resources required to complete the task. i max Indicates the maximum acceptable time to complete the task; USE i Tasks can be offloaded to edge servers in various network zones for processing; in the space-based zone, USE i Tasks can be offloaded to STS for execution via the USE-STS communication link; in the airborne region, USE i Tasks can be offloaded to the UAS for execution via the USE-UAS communication link; in the ground-based area, USE i Tasks can be offloaded to the BS for execution via the USE-BS communication link; the execution latency of a task consists of transmission latency and processing latency.

[0116] Figure 2 This is a flowchart illustrating the joint optimization algorithm for task unloading and resource allocation. First, the initial resource allocation result is substituted into formula (25) to obtain subproblem P1, which is only related to the task unloading decision, and solved using the PSTOA algorithm. Then, the obtained task unloading decision is substituted into formula (25) to obtain subproblem P2, which is only related to resource allocation, and solved using the EGCRD algorithm. Finally, the optimal task unloading decision and resource allocation scheme are obtained.

[0117] Figure 3 This diagram illustrates the comparison between the number of tasks and task execution latency. In this diagram, the horizontal axis represents the number of tasks, and the vertical axis represents task latency. The curves marked with blue circles, red crosses, yellow rectangles, and purple diamonds represent the Random Offloading Algorithm (RO), Genetic Resource Allocation Algorithm (GRA), Simulated Annealing Resource Allocation Algorithm (SARA), and the proposed algorithm (EGRDSS), respectively. This is because the EGRDSS algorithm can dynamically adjust its resource game strategy based on the current payoff and Fermi rule, optimizing resource allocation according to real-time task processing and resource supply-demand differences. Simultaneously, the improved PSO algorithm can effectively find the globally optimal task offloading decision across the entire search space, reducing task processing latency. Therefore, its task execution latency is lower than the other three algorithms. When the number of tasks is 50, compared to RO, GRA, and SARA, EGRDSS reduces task latency by 37.34%, 13.37%, and 21.9%, respectively.

[0118] Figure 4This diagram illustrates the comparison between transmission power and task execution latency. In this diagram, the horizontal and vertical axes represent transmission power and task execution latency, respectively. Furthermore, from left to right, the diagram shows the task execution latency under EGRDSS, GRA, SARA, and RATO. The diagram shows that as transmission power increases, the task latency of all four algorithms gradually decreases. This is because increasing transmission power significantly improves signal strength and quality, thereby increasing the task transmission rate and reducing transmission latency. In situations with significant interference, reducing transmission power avoids affecting other devices. Comparing the bar charts of different strategies, compared to the GRA, SARA, and RATO strategies, EGRDSS reduces task execution latency by 9.01%, 11.83%, and 28.94%, respectively.

[0119] The above specific embodiments further illustrate the inventive purpose, technical solution and beneficial effects of the present invention. It should be understood that the above specific embodiments are only illustrative examples and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions or alterations made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A 6G integrated air-space-ground network resource allocation device based on evolutionary game theory, characterized in that, Includes the following steps: 1) Construct a 6G integrated air-space-ground computing offloading network model; 2) Define the optimization objective and constraints to minimize task execution latency; 3) The task unloading and resource allocation problem is solved by using a joint task unloading and resource allocation algorithm.

2. The 6G air-space-ground integrated network resource allocation device based on evolutionary game theory as described in claim 1, characterized in that, Step 1) constructs a 6G integrated space-air-ground computing offloading network model. This network model consists of three layers: a space-based network, a space-based network, and a ground-based network. The space-based network contains N LEO satellites configured with edge servers, denoted as STS. The set of STSs can be represented as STS = {STS1, STS2, ..., STS}. n STS N |n∈N},STS n Let $n$ represent the nth LEO satellite, whose communication range covers the entire ground-based network. It can communicate with user equipment and handle computational tasks offloaded by user equipment. The airborne network consists of M unmanned aerial vehicles (UAVs), each equipped with a communication module and an edge server, denoted as $UAS$. The set of $UAS$ can be represented as $UAS = {UAS1, UAS2, ..., UAS}$. m , ..., UAS M |m∈M},UAS m Let r represent the effective communication radius of the m-th UAV. UAS It can provide communication / computing services for user equipment; the ground network consists of I user equipment (USE) and J base stations (BS), each BS deploying an edge server; the set of USE can be denoted as USE = {USE1, USE2, ..., USE...} i , ..., USE1|i∈I}, where USE i Let r represent the effective communication radius of the i-th user equipment. USE (r UAS >r BS >r USE It can send task offloading requests to BS, UAS, and STS; the set of BS can be denoted as BS = {BS1, BS2, ..., BS...} j , ..., BS J |j∈J}, where BS i Let r represent the effective communication radius of the j-th base station. BS It can provide computing resources for the tasks of USEs; In this network model, only USE spontaneously generates computational tasks; STS, BS, and UAS can only execute tasks but do not spontaneously generate computational tasks. Within each time slot, USE... i Generate a computational task that needs to be processed, which can be generated by a triple <D i C i T i max > indicates that D i Indicates the size of the task data, C i T represents the computational load of a task, i.e., the computing resources required to complete the task. i max Indicates the maximum acceptable time to complete the task; USE i Tasks can be offloaded to edge servers in various network zones for processing; in the space-based zone, USE i Tasks can be offloaded to STS for execution via the USE-STS communication link; in the airborne region, USE i Tasks can be offloaded to the UAS for execution via the USE-UAS communication link; in the ground-based area, USE i Tasks can be offloaded to the BS for execution via the USE-BS communication link; the execution latency of a task consists of transmission latency and processing latency.

3. The 6G air-space-ground integrated network resource allocation device based on evolutionary game theory as described in claim 1, characterized in that, Step 2) defines the optimization objective and constraints to minimize task execution latency. Centered on USE, it studies the computation offloading process in the 6G SAG-CO network model and further discusses the interference in the USE communication process: First, it considers the interference between USE-BS and USE-UAS communication links; second, there is also interference between STS-STS, UAS-UAS, and BS-BS communication links, but they are divided into different frequency bands from the USE transmission links. Tasks generated by USE are offloaded to the STS server via a wireless channel for execution. i With STS n The channel gain for information transmission between them can be expressed as: Among them, υ i，n ε represents the complex Gaussian variable representing Rayleigh fading. i，n The shadow fading indicates a log-normal distribution, d i,n Indicates USE i With STS n The distance between them Indicates the path loss index; Communication between USE and STS uses the Ka band, which does not cause interference with other frequency bands; USE i To STS n The achievable task transmission rate is expressed as: Among them, B i Indicates USE i The channel bandwidth, p i Indicates USE i The transmit power, σ 2 Indicates Gaussian white noise; Communication links between UAS and between BS share the same frequency band as USE-UAS. i To UAS m Signal-to-interference-plus-noise ratio γ i,m Represented as: in, Indicates USE i Offload the task to UAS within the communication range m Number of uses; Indicates interference caused by other USE-UAS communication links; G i.m Indicates USE i With UAS m The channel between; assuming the UAS position is quasi-fixed, i.e., the UAS is in a hovering state, USE i With UAS m The channel between them can be constructed as a Rician fading channel model, represented as: in, Denotes the line-of-sight (LoS) channel component, d i,m Indicates USE i To UAS m The distance between them, where λ represents the Rician factor. The path loss exponent, |g, represents the Loss of Path (LoS) in Rician fading. i，m |=1 indicates the LoS channel coefficient; G represents the non-line-of-sight (NLoS) channel component. i,m The NLoS channel coefficients represent the zero-mean, unit-variance Gaussian fading channel. This represents the path loss exponent of NLoS in Rician fading; therefore, USE i To UAS m The achievable task transmission rate is expressed as: R i,m =B i log2(1+γ i，m ) (5) USE can offload tasks to the BS. i Received from base station BS j The signal-to-noise ratio is: in, Indicates interference caused by other USE-BS links; G i,j Indicates USE i With BS j The channel between them is represented as: Where g0 represents the channel power gain at a reference distance d0 = 1, d i，j Indicates USE i To BS j The distance between them; therefore, USE i To BS j The achievable task transmission rate is expressed as: R i，j =B i log2(1+λ i，j ) (8) When USE i Offload computing tasks to STS n At that time, the transmission latency of the computation task transmitted to the STS server via the wireless channel It can be represented as: Processing latency required to process tasks on the STS edge server Represented as: Among them, f i，n This indicates the computing resources allocated by STS to the task; Therefore, when the task is unloaded to STS n When the task is executed on a server, the execution latency can be expressed by the formula: in, Represents binary decision; when The time indicates in STS n Tasks are processed on the server; This indicates that the task was not uninstalled to STS. n On the server; When USE i Offload computing tasks to UAS m At that time, the computational task is transmitted via the wireless channel with a delay It can be represented as: Processing latency required to process tasks on the UAS edge server Represented as: Among them, f i，u This indicates the computing resources allocated by the UAS to the task; Therefore, when the task is unloaded to the UAS m When the task is executed on a server, the execution latency can be expressed by the formula: in, Represents binary decision; when The time indicates in UAS m The server processes tasks; This indicates that the task was not uninstalled to the UAS. m Processing is performed on the server; When USE i Offload computing tasks to BS j At that time, the computational task is transmitted via the wireless channel with a delay It can be represented as: BS j Computational latency required to execute tasks on the server Represented as: Among them, f i，j BS j Computational resources allocated to the task; Therefore, when the task is unloaded to BS j When the task is executed on a server, the execution latency can be expressed by the formula: in, Represents binary decision; when The time indicates in BS j The server processes tasks; when This indicates that it is not in BS. j Tasks are processed on the server; In summary, USE i The total execution delay of the task can be expressed as: The optimization objective is to minimize the execution latency of the task under the following constraints; Unloading decision constraint: All user devices can only choose one unloading method for their tasks; Unloading decision value constraints: The unloading decision of the edge server of each network area for user tasks is a Boolean variable, which can only take the value 0 or 1; Execution latency constraint: The task execution latency must not exceed the maximum acceptable completion latency of the user equipment for the task; Computational resource allocation constraints: The computational resources allocated to user tasks by STS, UAS, and BS edge servers cannot exceed the maximum allocatable computational resources of the edge server, i.e.: in, This indicates the maximum computing resource allocation capability of the STS; This indicates the maximum computing resource allocation capacity of the UAS; This indicates the maximum computing resource allocation capability of the BS (Base Station). The objective function of the optimization problem can be specifically expressed as:

4. The 6G air-space-ground integrated network resource allocation device based on evolutionary game theory as described in claim 1, characterized in that, Step 3) Solve the task unloading and resource allocation problem using a joint task unloading and resource allocation algorithm. The algorithm steps are as follows: (1) Initialize the number of satellites, drones, base stations and user equipment, and the resource allocation results; (2) Calculate the task transmission rate of different channels based on parameters such as channel bandwidth, gain, and path loss index; (3) Solve the optimal unloading decision using the particle swarm task unloading optimization algorithm and the current optimal computing resource allocation scheme; (4) The optimal computing resource allocation scheme is solved by using the evolutionary dynamic game computing resource allocation algorithm and the current optimal unloading decision.

Images