CN116528298A - LEO satellite and HAP cooperation-based computing and unloading strategy in wide-range disaster area - Google Patents

Abstract

The invention discloses a calculation unloading strategy based on LEO satellite and HAP cooperation in a large-scale disaster area, wherein the strategy is to allocate subcarrier resources for tasks by adopting a cooperative game algorithm under the scheme of unloading decision; then, performing resource competition on the task unloaded to the edge server with the maximum system QoE value as a target, and realizing optimal allocation of computing resources; finally, the unloading decision problem is proved to be a potential game process, and an unloading decision scheme under Nash equilibrium is found by utilizing a Multi-user game calculation unloading (Multi-user Game Computing Offloading, MUGCO) algorithm. Simulation results show that compared with a comparison strategy, the provided strategy can effectively improve the total QoE value of the system.

Description

LEO satellite and HAP cooperation-based computing and unloading strategy in wide-range disaster area

Technical Field

The invention relates to a wireless communication technology in a large-scale disaster area, in particular to a calculation unloading strategy based on LEO satellite and HAP cooperation in the large-scale disaster area.

Background

In recent years, a large-scale disaster (such as earthquake, flood, hurricane, etc.) frequently occurs, and when such a large-scale disaster occurs, a communication network is paralyzed by destruction of a ground communication infrastructure (such as a base station), resulting in interruption of wireless service in a disaster area. In such emergency situations, it is imperative for disaster victims to broadcast their own location through a social platform on the user device and send out real-time dangerous situations. For rescue workers, the rescue workers are sent to disaster areas, and rescue tasks are performed by means of intelligent rescue user equipment (such as rescue robots). In order to handle these data tasks (e.g. video streaming, big data analysis, image processing) running on the ground user equipment (Ground User Equipment) in time so that post-disaster rescue can be performed smoothly, mobile edge computing (Mobile Edge Computing, MEC) is typically integrated into an emergency communication network to provide a computing offload service.

In a traditional disaster relief communication network, unmanned aerial vehicles (Unmanned Aerial Vehicle, UAVs) play a vital role in post-disaster relief due to the advantages of low cost, flexible deployment and high maneuverability. Currently, zhao, N et al propose that UAV cooperates with surviving base stations, and reduces the delay of communication between a disaster recovery person and an unmanned aerial vehicle by jointly optimizing the flight trajectory and scheduling of UAV. However, due to the limited load of the UAV itself and battery life, MEC servers deployed on the UAV are actually difficult to handle for a large number of intensive tasks. To overcome the difficulty of limited computational resources of UAV. Chen, W et al propose to offload the task of the upe to a surviving base station or drone deployed with an edge server using UAV relay. Xu, K et al propose a network architecture in which UVA cooperates with a remote wide area network, offloading tasks to the remote wide area network with rich computing resources by way of UAV flexible relay. Although they can temporarily solve the problem of limited computational resources on UAVs, the number of UAVs needed to achieve full coverage of wireless connections will double for a wide range of disaster areas, which is almost impossible to achieve, due to the small coverage limitations of UAVs themselves.

Compared to UAVs, the new HAP has the ability to carry heavy 5G infrastructure and provides longer length of service, greater area coverage and wider spectrum resources, making HAP integrated mobile edge computing (Mobile Edge Computing, MEC) well suited for emergency situations for large scale disaster relief. However, for a wide disaster area, overload is easily caused by unloading excessive GUE tasks onto HAP, so that the total QoE of GUE is reduced, and expansion of a cloud data center with rich computing power resources in a network becomes a good choice. LEO satellites, by virtue of their large coverage, have the ability to communicate directly with cloud data centers outside of disaster areas, so LEO satellites can act as an edge node for aggregation and relay forwarding in a wide range of disaster areas. Currently, star-air-ground integrated networks like this are emerging as research hotspots. For this reason, L.Zhang et al have proposed an SAIC architecture that explores joint user association and offloading decision-making problems in two-tier networks targeting maximization and rate. Ding, J. Et al propose a SAIECN architecture that minimizes the total energy consumption of the system by jointly optimizing GUE association, MU-MIMO transmission precoding, resource allocation and offloading decisions. The two-layer structure of the cooperation of the HAP and the LEO satellite is mainly considered, and the auxiliary effect of the cloud data center in the process of computing and unloading is ignored.

Disclosure of Invention

Aiming at the problems, the invention provides a calculation unloading strategy based on LEO satellite and HAP cooperation in a large-scale disaster area, aiming at improving GUE wireless service experience quality (Quality of Experience, qoE), and modeling the optimization problem as a weighted sum of maximizing all task unloading processing time delay and relatively reducing energy consumption.

The technical scheme for realizing the aim of the invention is as follows: firstly, under the scheme of unloading decision, adopting a cooperative game algorithm to allocate subcarrier resources for tasks; then, performing resource competition on the task unloaded to the edge server with the maximum system QoE value as a target, and realizing optimal allocation of computing resources; finally, the unloading decision problem is proved to be a potential game process, and an unloading decision scheme under Nash equilibrium is found by utilizing a Multi-user game calculation unloading (Multi-user Game Computing Offloading, MUGCO) algorithm. Simulation results show that compared with a comparison strategy, the provided strategy can effectively improve the total QoE value of the system.

The invention discloses a calculation unloading strategy based on LEO satellite and HAP cooperation in a large-scale disaster area, which comprises the following steps:

(1) Establishing a communication model: uplink transmission of tasks to be offloaded to HAP is carried out through Non-orthogonal multiple access (Non-orthogonal Multiple Access, NOMA) technology, K subcarriers are provided in the system, each HAP has full channel state information of the K subcarriers, tasks to be offloaded are divided into K groups, and GUEs in each group can share the same subcarrier; with set s= { S ₁ ,s ₂ ,...s _k And represents the K packets, where s _k A packet indicating the uploading of tasks on subcarrier k, the number of tasks in the group being Q, is represented by the set Γ= [1,. Q..q]A representation; in addition, useRepresenting the channel gain for task q upstream over subcarrier k, wherein +.>Then->Expressed by the formula (1):

wherein X is _i,j For GUE j, the horizontal distance with HAP i as reference, eta ₀ For channel gain at reference distance 1m, τ is the path loss index between GUE and HAP, the channel gain for the task of uplink transmission over subcarrier k satisfies

When task is defined according to NOMA technology _n,m The signal-to-interference-and-noise ratio of the uplink transmission to the HAP over subcarrier k can be expressed as:

in the middle ofRepresenting subcarrier selection factors, when task _n,m When subcarrier k is selected, < >>Otherwise, the reverse is performed; />Representing the transmission power, σ, of user q uploading data over subcarrier K ² Is background noise; the task can be obtained according to the formula _n,m The transmission rate from the terrestrial uplink to the HAP is:

in B of _k Representing the transmission bandwidth of subcarrier k;

(2) Establishing a calculation model: under the SAGE disaster relief network, the tasks generated by GUE have three unloading processing modes, namely o is defined _n,m E {0,1,2,3} is task _n,m Is a factor for offloading decisions:

when o _n,m When=0, the task is processed locally, and the time delay is representedWherein->The computing power of GUE; energy consumption->Where κ is a constant related to the CPU architecture;

when o _n,m When=1, the MEC server process for unloading the task to the HAP is represented, the time delay isWherein the method comprises the steps ofAssigning a task to an MEC server on a HAP _n,m Is a computing resource of (a); energy consumption->

When o _n,m When=2, it means that the task is offloaded to the satelliteProcessing by MEC server, delayWherein->Assigning a task to an MEC server on an LEO satellite _n,m L of the computing resource of (1) _n,s The distance from HAP n to LEO satellite, c is the speed of light; energy consumption->

When o _n,m When the number of the samples is =3,representing task offloading to cloud data center processing, latencyWherein->Representing a latency of a task from the LEO satellite to a cloud data center for processing; energy consumption->

Obtaining the task _n,m The time delay and energy consumption after the scheduling are as follows:

z in _{*} If the formula is true, Z _{*} =1, otherwise Z _{*} ＝0；

(3) Modeling of optimization problems: task _n,m The scheduled QoE function is defined as:

wherein the method comprises the steps ofWeights respectively representing time delay and energy consumption; the problem is modeled as maximizing the QoE function values for all the gus, namely:

wherein F represents a resource allocation scheme, S represents a subcarrier selection scheme, and W represents an unloading decision scheme; constraint condition C1 represents the maximum tolerance time delay of a task, C2 represents the task unloading decision selection space, C3 represents the subcarrier decision factor of the task, C4-C5 represent that the total calculation resources allocated by the unloading task cannot exceed the total calculation resources of the corresponding MEC server, and C6-C7 represent the minimum calculation resources required by the task to be unloaded to the corresponding edge server;

(4) Constructing priority according to two indexes of maximum time delay constraint of task and CPU period number required by calculating task, wherein task _n,m The mathematical model of the priority score can be expressed as:

wherein 0 is less than or equal to lambda _n,m Weights of two indexes which are less than or equal to 1 can be obtained by an entropy method; after the score of each task is obtained, the tasks are classified into four grades from high to low by adopting a sorting algorithm, and the four grades are stored in a matrix A= [ A ] ₁ ,A ₂ ,A ₃ ,A ₄ ]In (a) and (b); the task can be obtained _n,m The initialization pre-unloading scheme of (1) is as follows: when score _n,m ∈A ₁ ，o _n,m =1; when score _n,m ∈A ₂ ，o _n,m =2; when score _n,m ∈A ₃ ，o _n,m =3; when a task score _n,m ∈A ₄ ，o _n,m ＝0；

(5) A game model phi= { Y, S, G } is constructed for the allocation of subcarrier resources, wherein Y represents all task sets needing uplink transmission, and S= { S ₁ ,s ₂ ,...s _k The packet set of subcarriers, G the utility function of the task, proportional to the uplink transmission rate, subcarriers s _k The calculation formula of the utility function is as follows:q∈s _k in the formula->Representing the transmission rate of task q over subcarrier k:

definition 1: { s _ι ,s _k }→{s _ι ∪q,s _k { q } represents packet s _k Task q in (a) leaves and joins group s _ι Wherein s is _ι S is E S, and S _ι ≠s _k ；

Definition 2:representing task q versus group s _k Sum s _t For example, group s _k Task q preference group s in (a) _t Can be expressed as: />

Definition 3: if group s _k Task q and group s in (a) _ι If the task q' in (a) satisfies the definition 2, a switching behavior is generated, and the update packet is as follows: { s _ι ,s _k }→{s _ι ∪{q}\{q’},s _k ∪{q’}\{q}}；

Initializing subcarrier random distribution when GUE task uplink transmission, after iteration of cooperative game is carried out, randomly selecting subcarriers different from the last subcarrier to carry out grouping operation in the definition until each grouping is stable, thereby realizing highest efficiency of NOMA system;

for the problem of allocation of edge server computing resources, the mathematical model can be derived from equation (9) as:

s.tC4C5C6C7

(6) Modeling the offload decision problem as game ψ= { Z, W, H _n,m (where Z represents a set of gus, w= { (o) _n,m ):o _n,m E {0,1,2,3}, n E I, m E J } represent the selection policy space for each GUE, H _n,m The represented QoE function is called a potential game when a function χ exists in the game ψ that satisfies the following conditions:

and demonstrates that the potential function in game ψ is:

the invention has the advantages that:

(1) Problem modeling: and taking QoE values of all GUEs in the system as optimization targets, planning the problem as a weighted sum of task unloading processing time delay and relative reduction of energy consumption, and establishing an unloading decision and resource allocation problem model based on the SAGE disaster relief network.

(2) Mathematical analysis: the optimization problem is broken down into two sub-problems. The first sub-problem is the allocation problem of the sub-carrier resources and the computing resources under the unloading scheme, and the allocation problem is solved by adopting a cooperative game algorithm and a computing resource competition algorithm based on task priority respectively; the second sub-problem is an unloading decision problem with optimal resource allocation, which is modeled as a potential game process, and then an unloading decision scheme under Nash equilibrium is found by MUGCO algorithm.

(3) Simulation verification: the proposed offloading decision and resource allocation scheme is analyzed based on matlab software simulation and compared to a reference scheme. Simulation results show that compared with a reference strategy, the provided strategy can effectively improve the total QoE value of the system.

Drawings

FIG. 1 is a system model diagram of an embodiment of the present invention;

FIG. 2 is a flowchart of the MUGCO algorithm according to an embodiment of the present invention;

FIG. 3 is a graph showing the relationship between the number of GUEs and the average transmission rate of GUEs according to the embodiment of the present invention;

fig. 4 is a schematic diagram of QoE values of an edge server under different computing resource allocation algorithms according to an embodiment of the present invention;

fig. 5 is a graph of the amount of GUE versus the total QoE value of the system according to an embodiment of the present invention.

Detailed Description

The invention is further illustrated, but not limited, by the following figures and examples.

Examples:

the SAGIC relief network considered herein is shown in FIG. 1. In the air layer, there are I HAPs and one LEO satellite, using a setAnd->And (3) representing. They are all equipped with MEC servers to handle tasks running in the upe. Which is a kind ofIn which HAP is deployed above disaster area L _H And floats on the stratosphere in a quasi-static manner, and can process the GUE unloading task and also can be used as a relay node to communicate with LEO satellites. Furthermore, LEO satellites equipped with enhanced MEC servers are deployed at an orbital altitude L _s Where it may communicate with the cloud data center over a backhaul link. Suppose that each high-altitude platform serves J GUEs distributed randomly in the disaster area, with the set +.>Representing, and each GUE has a data task to be processed, using task _n,m Representing the task generated by the mth GUE under the nth HAP, wherein +.>Use array->Respectively represent task _n,m Wherein d is _n,m Representing the data size of a task, c _n,m Representing the number of CPU cycles required to process a task,representing the maximum tolerated latency of the task.

Establishing a communication model

The tasks to be offloaded are up-transferred to the HAP based on NOMA technology. It is assumed that there are K subcarriers in the system and that each HAP has full channel state information for the K subcarriers. We divide the tasks that need to be offloaded into K groups, the upes in each group can share the same subcarrier, with the set s= { S ₁ ,s ₂ ,...s _k And } represents. Wherein s is _k The ue packet uploaded on subcarrier k is represented, the number of tasks in the group is Q, and the set Γ= [1,. Q..q]And (3) representing. We useRepresenting the channel gain for task q upstream over subcarrier k, wherein +.>Then->The formula is expressed as:

wherein X is _i,j For GUE j, the horizontal distance with HAP i as reference, eta ₀ For channel gain at a reference distance of 1m, τ is the path loss index between GUE and HAP. Without loss of generality, it is assumed that the channel gain for the task of uplink transmission over subcarrier k is satisfied

When task is defined according to NOMA technology _n,m The signal-to-interference-and-noise ratio of the uplink transmission to the HAP over subcarrier k can be expressed as:

wherein the method comprises the steps ofRepresenting subcarrier selection factors, when task _n,m When subcarrier k is selected, < >>Otherwise, the reverse is performed; />Representing the transmission power, σ, of user q uploading data over subcarrier K ² Is background noise. According to the above, the task can be obtained from shannon's formula _n,m The transmission rate from the terrestrial uplink to the HAP is:

in B of _k Representing the transmission bandwidth of subcarrier k. For ease of analysis, a task is provided herein _n,m The transmission rate from the HAP to the LEO satellite is

Establishing a calculation model

Under the SAGE disaster relief network, the tasks generated by GUE have three unloading processing modes, namely o is defined _n,m E {0,1,2,3} is task _n,m Is a factor for offloading decisions. When o _n,m When=0, the task is processed locally, and the time delay is representedWherein->The computing power of GUE; energy consumption->Where κ is a constant related to the CPU architecture.

When o _n,m When=1, the MEC server process for unloading the task to the HAP is represented, the time delay isWherein the method comprises the steps ofAssigning a task to an MEC server on a HAP _n,m Is a computing resource of (a); energy consumption->

When o _n,m When=2, it means that the task is offloaded to the satelliteUpper part of the cylinderMEC server processing, latencyWherein->Assigning a task to an MEC server on an LEO satellite _n,m L of the computing resource of (1) _n,s The distance from HAP n to LEO satellite, c is the speed of light; energy consumption->

When o _n,m When=3, the task is unloaded to the cloud data center for processing, and the time delay is representedWherein->Representing a latency of a task from the LEO satellite to a cloud data center for processing; energy consumption->Furthermore, we use the matrix w= { o _1,1 ,o _1,2 ,...,o _n,m And all GUEs' offloading decision schemes.

In summary, a task can be obtained _n,m The time delay and energy consumption after the scheduling are as follows:

z in _{*} If the formula is true, Z _{*} =1, otherwise Z _{*} ＝0。

Optimization problem modeling

Based on the above, task _n,m The scheduled QoE function is defined as:

wherein the method comprises the steps ofRespectively representing the weights of the time delay and the energy consumption. The goal here is to maximize the QoE function values for all the gus, namely:

wherein F represents a resource allocation scheme and S represents a subcarrier selection scheme. Constraint condition C1 represents maximum tolerance time delay of a task, C2 represents unloading decision selection space of the task, C3 represents subcarrier decision factors of the task, C4-C5 represent that the total computing resources allocated by the unloading task cannot exceed the total computing resources of the corresponding MEC server, and C6-C7 represent minimum computing resources required by the task to be unloaded to the corresponding edge server.

Joint offloading decision and resource allocation policy

The optimization variables in problem (7) have both integer and continuous variables, so the problem is a mixed integer nonlinear programming problem. In addition, due to the strong coupling of the optimization variables, in order to facilitate mathematical analysis, the problem is decomposed into two sub-problems, wherein the first sub-problem is a resource allocation problem under a pre-unloading scheme, comprising the allocation of subcarrier resources and edge server computing resources, and the problem is solved by adopting a cooperative game and a computing resource competition algorithm based on task priority; the second sub-problem is an offloading decision problem with optimal resource allocation, which is first demonstrated as a potential function for offloading decisions, and then an offloading decision scheme under Nash equilibrium is found using MUGCO algorithm.

Initialization process

In disaster areas, different tasks have different service requirements, and therefore, priorities are built according to two indexes of maximum time delay constraint of the tasks and CPU period number required by calculation tasks. Wherein task _n,m The mathematical model of the priority score can be expressed as:

wherein 0 is less than or equal to lambda _n,m The weight of the two indexes is less than or equal to 1, and can be obtained by an entropy method. After the score of each task is obtained, the task is divided into four grades from high to low by adopting a sorting algorithm, and the four grades are stored in a matrix A= [ A ] ₁ ,A ₂ ,A ₃ ,A ₄ ]Is a kind of medium. The task can be obtained _n,m The initialization pre-unloading scheme of (1) is as follows: when score _n,m ∈A ₁ ，o _n,m =1; when score _n,m ∈A ₂ ，o _n,m =2; when score _n,m ∈A ₃ ，o _n,m =3; when a task score _n,m ∈A ₄ ，o _n,m ＝0。

Resource allocation

In the subcarrier resource allocation process, in order to efficiently group the tasks of a plurality of GUEs, so as to improve the uplink transmission rate of the tasks, a cooperative game algorithm is adopted, and a game model phi= { Y, S, G } is constructed. Wherein Y represents all task sets requiring uplink transmission, s= { S ₁ ,s ₂ ,...s _k The symbol G represents the utility function of the task, which is proportional to the uplink transmission rate. For example, subcarrier s _k The calculation formula of the utility function is as follows:q∈s _k in the followingRepresenting the transmission rate of task q over subcarrier k. The process of the cooperative game will be described in detail as follows:

definition 1: { s _ι ,s _k }→{s _ι ∪q,s _k { q } represents packet s _k Task q in (a) leaves and joins group s _ι Wherein s is _ι S is E S, and S _ι ≠s _k 。

Definition 2:representing task q versus group s _k Sum s _t For example, group s _k Task q preference group s in (a) _t Can be expressed as: />

Definition 3: if group s _k Task q and group s in (a) _ι If the task q' in (a) satisfies definition 2, they will generate switching behavior and update the packet as: { s _ι ,s _k }→{s _ι ∪{q}\{q’},s _k ∪{q’}\{q}}。

When the GUE task is transmitted in the uplink, the initialized subcarriers are allocated randomly. After iteration of the cooperative game is entered, randomly selecting subcarriers different from the last subcarrier to perform grouping operation in the definition until each grouping is stable, so that the highest efficiency of the NOMA system is realized. The specific cooperative game process is shown in the following algorithm 1:

lemma 1: algorithm 1 can converge to the final set of packets S in a finite step _end 。

And (3) proving: in order to optimize the utility function value G, grouping operations are continuously performed between users. In the kth to k+1 iterations, if the set of packets is from S _k Updated to S _k+1 The utility function G in the game must be strictly increased, namely:the utility function G is always incremented in game iterations, namely: s is S _start →S ₁ →...→S _end . Since the number of users for uplink transmission is limited, the packet set is limited and based on the bell number, the packet set will converge to S as the game iteration increases _end 。

For the problem of allocation of edge server computing resources, the mathematical model can be derived from equation (7) as:

s.tC4C5C6C7

aiming at the difference of task service demands of different GUEs in disaster areas, a computing resource competition algorithm based on task priority is designed, namely, on the premise of task priority, the minimum computing resources meeting the maximum delay requirement are sequentially distributed to users unloaded to an edge server from high to low according to the priority, then idle computing resources of the server are divided into a plurality of resource blocks, and the unloaded tasks realize the maximization of the total QoE value of the system through the competition resource blocks, wherein the specific steps are as shown in algorithm 2:

offloading decision-making

For GUEs in disaster areas, it is desirable to design an offloading decision scheme that is low in complexity and relatively satisfactory for all GUEs to achieve the goal of maximizing the total QoE value of the system. In game theory, potential games have the advantage of fast convergence in studying distributed computing offload strategy problems due to their limited improved nature. Thus, here based on potential gaming, the offload decision problem is first modeled as game ψ= { Z, W, H _n,m (where Z represents a set of gus, w= { (o) _n,m ):o _n,m E {0,1,2,3}, n E I, m E J } represent the selection policy space for each GUE, H _n,m Represented QoE function. We then demonstrate that game ψ is a potential game. The detailed description is as follows:

definition 4: a game ψ is said to be a potential game if it has a function χ that satisfies the following conditions:

and (4) lemma 2: the potential function in game ψ is:

case one: when task _n,m Is determined by o _n,m Update =0 to o' _n,m At > 0, the partially offloaded task may be due to task _n,m The computing resources of the transmitting sub-carriers and the edge servers are occupied, resulting in their transmission performance and computation performance being degraded. By task _x,y Representing the portion of the task that is affected. If H _n,m (o _n,m ,o _-(n,m) )-H _n,m (o’ _n,m ,o _-(n,m) ) < 0, then the following can be deduced:

wherein T is _x,y And E is _x,y Representing task _n,m Updating task before offloading decision _x,y Time delay and energy consumption of (a); t'. _x,y And E' _x,y Representing task _n,m Task after updating unloading decision _x,y Time delay and energy consumption of (a); t'. _n,m And E' _n,m Representing task _n,m And updating the time delay and the energy consumption after the unloading decision.

Without losing generality, letIt can be deduced from the above: x (W) -X (W') < 0.

And a second case: when task _n,m Is determined by o _n,m Update to o 'from > 0' _n,m When=0, the partially unloaded task is due to task _n,m The computing resources of the transmitting sub-carriers and the edge servers are released, so that their transmission performance and computation performance are improved. If using task _x,y Representing the affected part of the task, availableH _x,y (o _x,y ,o _-(x,y) )＜H _x,y (o’ _x,y ,o _-(x,y) ). If H _n,m (o _n,m ,o _-(n,m) )-H _n,m (o’ _n,m ,o _-(n,m) ) < 0, then the following can be deduced:

and a third case: when task _n,m Is determined by o _n,m Update to o 'from > 0' _n,m > 0, but o _n,m ≠o’ _n,m . At this time, three cases can be discussed as follows:

①task _n,m is determined by 0 < o _n,m Update < 3 to 0 < o' _n,m < 3. Without loss of generality, we assume a task _n,m Is determined by o _n,m Update to o 'for =1' _n,m =2, this update releases the task _n,m Offloading required computing resources on HAP competing with tasks offloaded to LEO satellites while also adding tasks _n,m Is used for the transmission delay of the (a). Here use task _x1,y1 Andrespectively because of task _n,m Freeing and occupying computing resources.

If H _n,m (o _n,m ,o _-(n,m) )-H _n,m (o’ _n,m ,o _-(n,m) ) < 0, then the following can be deduced:

/>

in the middle ofAnd->Representing +.>And->A divided computing resource; f's' _x1,y1 And f' _x2,y2 Indicating +.>And->And (5) the separated computing resources.

In the middle ofAnd->Respectively represent task _n,m Processing delay of the affected task before and after releasing the computing resources, +.>And->Respectively represent task _n,m Processing delays for affected tasks before and after the computing resources are occupied.

②task _n,m Is determined by o _n,m Update to 0 < o 'for =3' _n,m And < 3, the update can cause the calculation performance of part of tasks to be reduced because the update occupies the calculation resources of the edge server, and the proving method is similar to the first case and is not repeated here.

③task _n,m Is determined by 0 < o _n,m < 3 update to o' _n,m The update releases the computing resources of the server, so that the transmission performance and the computing performance of some users are improved, and the proving method is similar to the second case, and will not be repeated here.

In summary, the evidence function χ (S) is the potential function of game ψ, so game ψ is the potential game. From the limited improved nature of potential gaming, game ψ can find the Nash equilibrium of the system after a limited number of iterations. Based on this property, a musco algorithm is proposed herein to realize nash equalization, and the core idea of the MUGCO algorithm is: and updating the unloading decision of the GUE by not more than one GUE at a time, wherein each updating rule can find the optimal response of the GUE, and if the GUE finds the optimal response, an unloading decision scheme under Nash equilibrium can be obtained. The optimal response calculation method is shown as follows:

algorithm complexity analysis

In order to maximize the QoE value of the system, a combined unloading decision and resource allocation strategy is proposed, wherein the allocation of subcarrier resources adopts a cooperative game algorithm, K times of calculation are carried out in each iteration, and when the iteration reaches the Kth time, the complexity is O (K ² ) The method comprises the steps of carrying out a first treatment on the surface of the The computing resource allocation adopts a computing resource competition algorithm based on task priority, and the complexity is O (I multiplied by J); the offload decision algorithm complexity is O (K ² ×I×J)。

Simulation analysis

Simulation experiment analysis is performed on the proposed scheme in the SAGE disaster relief scene. In a wide disaster area with a diameter of 100KM, GUEs are randomly distributed, and 10 HAP and 1 LEO satellites are deployed at a height L above the disaster area _H =20 km and track height L _s At=500 km. The computing resource of the edge server on the HAP is 8GHz, and the computing resource of the edge server on the LEO satellite is 20GHz; the remaining simulation parameters are shown in table 1:

table 1 simulation parameters

On the premise of being based on the NOMA technology, the cooperative game (cooperative game-NOMA) algorithm is adopted as a reasonable subcarrier pairing task, so that the overall efficiency of the NOMA system is highest, and the average uplink transmission rate of GUE is improved. In contrast to the random allocation scheme (random-NOMA), the orthogonal multiple access scheme (OMA). As shown in fig. 3, as the number of the upe-stream upe, the average transmission rate of upe will decrease continuously. This is because the subcarrier multiplexing frequency increases, resulting in an increase in interference experienced by the ue during transmission. However, compared to the other two schemes, the average speed obtained by the GUE is highest when the cooperative game-NOMA algorithm is adopted.

In the scheme of computing resource allocation of the edge server, a computing resource competition algorithm based on task priority is provided, under the algorithm, tasks unloaded to the edge server are firstly ordered from high to low according to the corresponding priority, then minimum computing resources required by task processing are allocated for the tasks under the condition that the maximum delay requirement of GUE is met, and finally the residual computing resources are maximized according to a multi-user competition resource rule. Compared to the convex optimization scheme and the minimum resource allocation scheme. As shown in fig. 4, when j=17, the edge servers calculate QoE values obtained by resource allocation according to different schemes, where the edge servers No. 1-10 belong to HAP and No. 11 belong to LEO satellite, and as can be seen from the figure, the QoE values obtained by the algorithm herein are better than the other two schemes.

In the unloading decision scheme, in order to maximize the total QoE value of the system, on the premise of being based on potential games, the MUGCO algorithm is adopted to find the Nash balance In the system, and the Nash balance is compared with a Greedy algorithm (Greedy), a random unloading algorithm (Rand) and an All-to-HAP algorithm (All-In-HAP). As shown in fig. 5, in the case of a smaller amount of gus, the offload tasks can obtain a higher transmission rate and more computing resources, so that the total QoE value in the system increases rapidly; however, as the amount of GUEs increases, the transmission rate decreases and the edge server computing resources compete more, resulting in a decrease in QoE that is achieved when the GUE task is offloading processing, which results in a slow increase in the overall QoE value in the system. But the system QoE value obtained by the MUGCO algorithm is highest compared to the other three offloading decision algorithms.

Conclusion(s)

The problem of GUE resource allocation and unloading decision in a wide disaster area based on SAGE network is studied, and an optimization problem model is established by taking the total QoE value of the system as a target. Because the QoE requirements of different GUEs in disaster areas are different, priorities are set according to the attributes of GUE tasks, and an unloading decision scheme is initialized according to the priority level; then reasonably distributing subcarrier resources for the uplink GUE by adopting a cooperative game algorithm, and distributing computing resources for tasks of the GUE unloaded to the edge server by adopting a computing resource competition algorithm based on task priority; finally, an unloading decision scheme under Nash equilibrium is obtained through a MUGCO algorithm, so that the total QoE value of the system is effectively improved.

Claims (1)

1. The calculation unloading strategy based on the cooperation of LEO satellites and HAP in a wide-range disaster area is characterized in that: comprising the following steps:

(1) Establishing a communication model: the tasks to be offloaded are uplink-transmitted to HAP through non-orthogonal multiple access NOMA technology, K subcarriers are provided in the system, each HAP has the full channel state information of the K subcarriers, the tasks to be offloaded are divided into K groups, and GUEs in each group can share the same subcarrier; with set s= { S ₁ ,s ₂ ,...s _k And represents the K packets, where s _k A packet indicating the uploading of tasks on subcarrier k, the number of tasks in the group being Q, is represented by the set Γ= [1,. Q..q]A representation; in addition, useRepresenting the channel gain for task q upstream over subcarrier k, wherein +.>Then->Expressed by the formula (1):

wherein X is _i,j For GUE j, the horizontal distance with HAP i as reference, eta ₀ For channel gain at reference distance 1m, τ is the path loss index between GUE and HAP, the channel gain for the task of uplink transmission over subcarrier k satisfies

When task is defined according to NOMA technology _n,m The signal-to-interference-and-noise ratio of the uplink transmission to the HAP over subcarrier k can be expressed as:

in the middle ofRepresenting subcarrier selection factors, when task _n,m When subcarrier k is selected, < >>Otherwise, the reverse is performed; />Representing the transmission power, σ, of user q uploading data over subcarrier K ² Is background noise; the task can be obtained according to the formula _n,m The transmission rate from the terrestrial uplink to the HAP is:

in B of _k Representing the transmission bandwidth of subcarrier k;

(2) Establishing a calculation model: under the SAGE disaster relief network, the tasks generated by GUE have three unloading processing modes, namely o is defined _n,m E {0,1,2,3} is task _n,m Is a factor for offloading decisions:

when o _n,m When=0, the task is processed locally, and the time delay is representedWherein->The computing power of GUE; energy consumptionWhere κ is a constant related to the CPU architecture;

when o _n,m When=1, the MEC server process for unloading the task to the HAP is represented, the time delay isWherein->Assigning a task to an MEC server on a HAP _n,m Is a computing resource of (a); energy consumption->

When o _n,m When=2, it means that the task is offloaded to the satelliteProcessing by MEC server, delayWherein->Assigning a task to an MEC server on an LEO satellite _n,m L of the computing resource of (1) _n,s The distance from HAP n to LEO satellite, c is the speed of light; energy consumption->

When o _n,m When=3, the task is unloaded to the cloud data center for processing, and the time delay is representedWherein->Representing a latency of a task from the LEO satellite to a cloud data center for processing; energy consumption->

Obtaining the task _n,m The time delay and energy consumption after the scheduling are as follows:

z in _{*} If the formula is true, Z _{*} =1, otherwise Z _{*} ＝0；

(3) Modeling of optimization problems: task _n,m The scheduled QoE function is defined as:

wherein the method comprises the steps ofWeights respectively representing time delay and energy consumption; the problem is modeled as maximizing the QoE function values for all the gus, namely:

s.t C1:

C2:

C3:

C4:

C5:

C6:

C7:

wherein F represents a resource allocation scheme, S represents a subcarrier selection scheme, and W represents an unloading decision scheme; constraint condition C1 represents the maximum tolerance time delay of a task, C2 represents the task unloading decision selection space, C3 represents the subcarrier decision factor of the task, C4-C5 represent that the total calculation resources allocated by the unloading task cannot exceed the total calculation resources of the corresponding MEC server, and C6-C7 represent the minimum calculation resources required by the task to be unloaded to the corresponding edge server;

(4) Constructing priority according to two indexes of maximum time delay constraint of task and CPU period number required by calculating task, wherein task _n,m The mathematical model of the priority score can be expressed as:

wherein 0 is less than or equal to lambda _n,m Weights of two indexes which are less than or equal to 1 can be obtained by an entropy method; after the score of each task is obtained, the tasks are classified into four grades from high to low by adopting a sorting algorithm, and the four grades are stored in a matrix A= [ A ] ₁ ,A ₂ ,A ₃ ,A ₄ ]In (a) and (b); the task can be obtained _n,m The initialization pre-unloading scheme of (1) is as follows: when score _n,m ∈A ₁ ，o _n,m =1; when score _n,m ∈A ₂ ，o _n,m =2; when score _n,m ∈A ₃ ，o _n,m =3; when a task score _n,m ∈A ₄ ，o _n,m ＝0；

(5) A game model phi= { Y, S, G } is constructed for the allocation of subcarrier resources, wherein Y represents all task sets needing uplink transmission, and S= { S ₁ ,s ₂ ,...s _k The packet set of subcarriers, G the utility function of the task, proportional to the uplink transmission rate, subcarriers s _k The calculation formula of the utility function is as follows:in->Representing the transmission rate of task q over subcarrier k:

definition 1: { s _ι ,s _k }→{s _ι ∪q,s _k 'q' tableIndicator group s _k Task q in (a) leaves and joins group s _ι Wherein s is _ι S is E S, and S _ι ≠s _k ；

Definition 2: > _q Representing task q versus group s _k Sum s _t Is the degree of preference of group s _k Task q preference group s in (a) _t Can be expressed as:

definition 3: if group s _k Task q and group s in (a) _ι If the task q' in (a) satisfies the definition 2, a switching behavior is generated, and the update packet is as follows: { s _ι ,s _k }→{s _ι ∪{q}\{q'},s _k ∪{q'}\{q}}；

Initializing subcarrier random distribution when GUE task uplink transmission, after iteration of cooperative game is carried out, randomly selecting subcarriers different from the last subcarrier to carry out grouping operation in the definition until each grouping is stable, thereby realizing highest efficiency of NOMA system;

for the problem of allocation of edge server computing resources, the mathematical model can be derived from equation (9) as:

(6) Modeling the offload decision problem as game ψ= { Z, W, H _n,m (where Z represents a set of gus, w= { (o) _n,m ):o _n,m E {0,1,2,3}, n E I, m E J } represent the selection policy space for each GUE, H _n,m The represented QoE function is called a potential game when a function χ exists in the game ψ that satisfies the following conditions:

and demonstrates that the potential function in game ψ is: