CN114269019A - Resource allocation method based on double-shooting game in aerospace information network - Google Patents

Abstract

The invention discloses a resource allocation method based on a double-beat game in an aerospace information network, which is characterized by comprising a low-orbit satellite system based on a High Altitude Platform (HAP) and an EWA algorithm mechanism based on the double-beat game, wherein the low-orbit satellite system can directly access computing resources of an adjacent High Altitude Platform, the computing resources are purchased from the High Altitude Platform by a low-orbit satellite, a resource trading market is established, and a buyer and a seller can dynamically adjust the bidding and asking price strategies through interaction with the environment based on the EWA algorithm. The invention introduces a high-altitude platform to unload the computing task of the low-orbit satellite so as to reduce the transmission delay. And calculating the resource allocation problem between the near-orbit satellite and the high-altitude platform based on a double-shot selling mechanism. Then, a nash equilibrium search algorithm based on Empirical Weight Attraction (EWA) was designed, which combines the advantages of reinforcement learning and belief learning, and is performed on each participant.

Description

Resource allocation method based on double-shooting game in aerospace information network

Technical Field

The invention relates to the technical field, in particular to a resource allocation method based on double-beat game selling in an aerospace information network.

Background

In recent years, satellite communication technology has gradually changed our lives and is an important communication technology that is indispensable. Satellite communication has many advantages, such as long communication distance, no geographical limitation, high system reliability, wide coverage, etc. Compared with a high Orbit Earth (GEO) satellite and a Medium Orbit Earth (MEO) satellite, the Low Orbit Earth (LEO) satellite is close to the ground, communication delay is short, data transmission rate is high, the weight and the volume of the mobile terminal are almost the same as those of personal mobile equipment, and the mobile terminal is more suitable for popularization.

However, satellite communication also has its disadvantages, due to limitations in computing power and insufficient energy, the satellite has to offload the data to the ground station for further processing. Long-range transmission has a large impact on many real-time services. Therefore, effective offloading mechanisms have received extensive attention from both academia and industry.

In this context, a High Altitude Platform (HAP) was introduced to offload the computational tasks of low earth orbit satellites. The high altitude platform with the height of 17-30km can be used as a relay station to transfer resources between the low orbit satellite system and the ground station. The introduction of the high-altitude platform can reduce transmission delay and effectively avoid interference to the troposphere. Meanwhile, the high-altitude platform has strong computing power and can help process computing resources unloaded by the satellite. With the increase of high-altitude platforms, lower transmission delay and lower cost can be obtained in the real-time remote sensing event. Nevertheless, low earth orbit satellites and high altitude platforms remain separate, self-administered individuals. Improving the resource allocation mechanism between low earth orbit satellites and high altitude platforms remains an urgent issue.

Disclosure of Invention

The invention aims to solve the technical problem that data has to be unloaded to a ground station for further processing due to the limitation of the computing capacity and insufficient energy of a low-orbit satellite at present.

A double-beat game-based resource allocation method in an aerospace information network comprises the steps of designing a High Altitude Platform (HAP) -based low orbit satellite system and an EWA algorithm mechanism based on a double-beat game, wherein the low orbit satellite system can directly access computing resources of adjacent High Altitude platforms, the computing resources are purchased from the High Altitude platforms through low orbit satellites, a resource trading market is established, and a buyer and a seller can dynamically adjust bidding strategies and asking price strategies through interaction with the environment based on the EWA algorithm.

Preferably, an M/M/1 queuing model is adopted in the low-orbit satellite system to quantify the urgency degree of low-orbit satellite tasks, and lambda is used_iRepresenting LEO_iTask arrival rate of, and HAP_jWill be expressed as mu_jIn addition, since the distance between the high altitude platform and the low orbit satellite is short, the main service latency can be attributed to the processing delay, and the total delay of the mission can be expressed as:

equal to the wait time plus the service time; then, the average latency can be calculated as:

according to the formula L_q＝λ_iT_wThe average length of the queue can be expressed as:

each low earth orbit satellite has different traffic demands, a higher λ indicates a higher frequency of arrival of the task, and a smaller μ indicates that the task requires a large amount of computation, particularly when L_qVery large means that the low earth orbit satellites have very urgent access to additional computing resources.

Preferably, a double-beat game mechanism is adopted in the low-orbit satellite system, and A ═ a is used_j1,2,3 … N to describe the asking price strategy for high-altitude platforms, where a_jRepresents SS_jAcceptable minimum price; use of b_i＝{b_i,j1,2,3 … N to describe LEO_iBidding prices for different high-altitude platforms; using the matrix B ═ B _i1,2,3 … M represents the bidding strategy for all low orbit satellites; beta is a_i,jFor representing LEO_iPaying HAP for computing resources_jThe amount of (c); an aerial platform has several items available for sale to N buyers, so that β is a matrix whose columns record the HAPs_jA trade price that sells its computing resources to a certain low earth orbit satellite, such that:

is HAP_jTotal reward in all successful transactions.

Preferably, the EWA algorithm mechanism based on the double-auction game introduces Sagents and Bagents which are agents traded by buyers and sellers; the bids and asks range from 0 to P_maxDiscrete and the number of bids and asks P (P ═ P)_max+1)；

Is Sagent_jFor HAP_jThe set of inquiry policies of (1),

meaning that it will select a certain asking price, obviously S_sjIs equal to a_j. Also, in the same manner as above,

denotes Bagent_iThe policy of Bagent is to bid b_iOf the vector of (1), the element b thereof_i,jIs for Sagent_jThe number of Sagents, the number of elements, etc.; one seller has P bids and n sellers, so the number of strategies is PⁿAnd, S_biIs equal to

Using S_s＝{S_s1,S_s2,…,S_snAs a set of inquiry strategies for n sellers, i.e. S_sDefine S as A_s-j＝{S_s1,S_s2,…,S_s(j-1),S_s(j+1),…,S_snDescription except Sagent_jOr equal A- { a_jS of }_sAnd (4) other strategies. For batents, S_b＝{S₁,S₂,…,S_pIs its set of bidding strategies, i.e. S_b＝B。S_b-i＝{S_b1,S_b2,…,S_b(i-1),S_b(i+1),…,S_bnS or S_s＝B-{b_iIs except Bagent_i(ii) a policy other than;

key parameters in EWA:

for Bagents and Sagents, pi is used_bi(S_bi,S_b-i)＝U_iAnd pi_sj(S_sj,S_s-j)＝U_jIndicating the rewards of Bagents and Sagents, the rewards and the policy of the Agent itself and its opponents.

The invention introduces a high-altitude platform to unload the computing task of the low-orbit satellite so as to reduce the transmission delay. And calculating the resource allocation problem between the near-orbit satellite and the high-altitude platform based on a double-shot selling mechanism. Then, a nash equilibrium search algorithm based on Empirical Weight Attraction (EWA) was designed, which combines the advantages of reinforcement learning and belief learning, and is performed on each participant. Auction participants can obtain information from other opponents and accumulate experience and thought. Thus, they can complete transactions with other participants in a fuzzy environment and maximize the overall interest of the participants. The convergence of this EWA algorithm is verified by simulation.

Drawings

The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.

FIG. 1 is a schematic diagram of a low-orbit satellite system based on a high-altitude platform in a double-beat-sale-game-based resource allocation method in an aerospace information network according to the invention;

fig. 2 is a schematic diagram of a double-auction process in a resource allocation method based on a double-auction game in an aerospace information network according to the present invention;

FIG. 3 is a schematic diagram of convergence of an EWA-based resource allocation algorithm in a double-beat-sale-game-based resource allocation method in an aerospace information network according to the present invention;

FIG. 4 is a schematic diagram illustrating the average cost of a seller and the average asking price of the seller in the method for allocating resources based on double-auction game in the air-space information network according to the present invention;

fig. 5 is a schematic diagram of average buyer value and average buyer bid price in a double auction game-based resource allocation method in an aerospace information network according to the present invention.

Detailed Description

In order that the objects and advantages of the invention will be more clearly understood, the invention is further described below with reference to examples; it should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only for explaining the technical principle of the present invention, and do not limit the scope of the present invention.

The invention provides a resource allocation method based on a double-beat selling game in an aerospace information network, which comprises the following steps: please refer to fig. 1: a low-earth satellite system based on High Altitude Platform (HAP) is designed, wherein the low-earth satellite can directly access to the computing resources of the adjacent High Altitude Platform, and the computing resources are purchased from the High Altitude Platform by the low-earth satellite so as to reduce the task burden of the low-earth satellite, thereby establishing a resource trading market.

Suppose there are M low-orbit satellites labeled X ═ M_i},i＝1,2,3…M。C＝{ c _i1,2,3 … M denotes a low-earth satellite M_iThe resource requirements of (1). Suppose there are N aerial platforms, denoted Y ═ N respectively_j},j＝1,2,3…N。R＝{r_jJ is 1, …, N represents HAP_jThe remaining capacity of the resource is calculated.

To evaluate the estimated revenue of the additional computing resources of different low earth orbit satellites, it is assumed that the estimated revenue is positively correlated with the urgency of the mission. In the present invention, an M/M/1 queuing model is used to quantify the urgency of low-earth-orbit satellite missions. Using λ_iRepresenting LEO_iTask arrival rate of, and HAP_jWill be expressed as mu_j. In addition, the main service latency can be attributed to processing delays due to the short distance between the high altitude platform and the low orbit satellite. The total delay of a task can be expressed as:

equal to the latency plus the service time. Then, the average latency can be calculated as:

according to the formula L_q＝λ_iT_wThe average length of the queue can be expressed as:

each low earth satellite has different traffic demands. A higher lambda indicates a higher frequency of arrival of the task, while a smaller mu indicates that the task requires a larger amount of computation. In particular, when L_qVery large means that the low earth orbit satellites have very urgent access to additional computing resources.

How to properly allocate limited high altitude platform computing resources among low earth orbit satellites is a complex problem. In the context of this trading market model, on the one hand, the overhead platform should acquire a fee that at least matches its total fee. Low earth orbit satellites, on the other hand, are willing to purchase as much additional computing resources as possible to satisfy different types of tasks. For example, the car networking task is time sensitive and computationally intensive, and therefore it is critical whether these low earth orbit satellites can accept computing resources in a timely manner.

The double auction gaming mechanism:

in the present invention, a ═ a is used_j1,2,3 … N to describe the asking price strategy for high-altitude platforms, where a_jRepresents SS_jThe lowest price that is acceptable. For simplicity, it is assumed that the computing tasks from low earth orbit satellites are indivisible, which means that the computing tasks can only be offloaded on one high altitude platform. Then, using b_i＝{ b _i,j1,2,3 … N to describe LEO_iAnd (4) bidding prices for different high-altitude platforms. Due to quality of goods (μ) from different vendors_j) Different, LEO_iTending to different bid prices. In other words, for LEO_iHAP of_jQuality of service determines LEO_iThe preference for high altitude platforms is different. Using the matrix B ═ B _i1,2,3 … M for all low earth orbit satellitesAnd (4) bidding strategies.

As shown in FIG. 2, in a two-way auction, each buyer and seller submit an ask price and a bid price simultaneously. The broker then determines the HAP_jWhether to accept LEO_iThe bid of (1). In the present invention, β_i,jFor representing LEO_iPaying HAP for computing resources_jThe amount of (c). If LEO_iAnd HAP_jWithout transaction between, then β_i,jEqual to 0. Since a high-altitude platform has several commodities available to several buyers, β is a matrix whose columns record HAPs_jA trade price that sells its computing resources to a certain low-orbit satellite. Thus, it is possible to prevent the occurrence of,

is HAP_jTotal reward in all successful transactions. Similarly, another matrix α ═ α is defined_j,iIts element represents HAP_jTo LEO_iThe amount of computing resources of. The sum of the j row of alpha records the HAP_jTotal transaction amount of (1). If LEO_iAnd HAP_jNo transaction occurs between, then α_j,iEqual to 0. Agent can calculate HAP by the following method_jHow many resources remain in:

if the amount of computing resources remaining is insufficient, the transaction will not be allowed.

Game analysis:

since the transactions are conducted by brokers, the private information is not known to each other. In such an opaque information environment, the seller or buyer can only know whether he has completed the transaction. Therefore, to improve their asking price and bidding strategy, it is necessary to define a utility function to simulate the participant's profit in a round of trading by:

U_j＝P_j-Cost_j

represents HAP_jA utility function of, wherein P_jIs the total reward for a round of double auctions, and Cost_jIncluding construction costs and maintenance costs. Also, LEO_iThe utility function of (a) is:

if LEO_iIf no transaction is reached with all sellers, U_iIs equal to-V_i(L_q)。V_i(L_q) Is LEO_iAn estimate or value of how much profit it can make from a successful transaction. Performing its computational tasks on a high-altitude platform depends on importance or urgency. Furthermore, this importance and the estimation of computational resources may be reflected in the need for computational delay.

Thus, using L_qTo estimate the return of low earth orbit satellites in receiving high altitude platform computing resources. A large amount of L_qMeaning that the low earth orbit satellite has accumulated sufficient mission and LEO_iThere is now an urgent need to obtain resources from high altitude platforms. As a result, purchasing additional computing resources is more attractive to them.

In double auction, brokers strive to maximize the sum of the utility functions of both buyers and sellers. This problem can be described as follows:

in the above maximization problem, there are some constraints, one of which is the expected amount of resources, and the remaining resources must be greater than 0. In addition, the bid price of the buyer and the ask price of the seller must be between 0 and the maximum price, which means that there is a limitation in the price given by the buyer and the seller.

Therefore, we have formulated a double auction model for the resource allocation problem described above. Both buyers and sellers can continually adjust their policies to maximize their profits. On the one hand, buyers desire to obtain resources at the lowest price that sellers can accept. On the other hand, buyers need to compete with other buyers for limited computing resources. When the buyer and seller adjust the strategy after each auction round, this is a game that the two parties are not working. Specifically, the goal of this game is to find nash equilibrium.

Nash equilibrium is the best result of gaming, with each participant in the auction finding the best strategy to obtain the maximum profit, and no participant has an incentive to change his strategy after considering the choice of opponents. In our question, nash equilibrium for double-beat game is defined as follows:

definition 1: in our question, let A^*And B^*Optimal bidding strategies and interrogation strategies for low earth orbit satellites and high altitude platforms. Then, if the following condition is satisfied, point (A)^*,B^*) Is the nash equilibrium point:

U_i(A^*,B^*)≥U_i(A^*,B)

and the number of the first and second groups,

U_j(A^*,B^*)≥U_j(A,B^*)

the EWA mechanism:

in the present invention, in order to search Nash balance, a multi-agent machine learning algorithm, called EWA algorithm, is introduced in the system model.

An EWA mechanism based on double-beat selling game:

the EWA algorithm combines belief learning and reinforcement learning. Belief learning enables agents to select their next strategy based on observations of other Agent selections, reinforcement learning enables agents to learn from experience. Thus, based on the EWA algorithm, buyers and sellers can dynamically adjust their bid and ask strategies through interaction with the environment.

For simplicity, Sagents and Bagents, which are agents traded by buyers and sellers, were introduced. In the present invention, the bid and ask ranges from 0 to P_maxDiscrete and the number of bids and asks P (P ═ P)_max+1)。

Is Sagent_jFor HAP_jThe set of interrogation strategies of.

Meaning that it will select a certain asking price. Obviously, S_sjIs equal to a_j. Also, in the same manner as above,

denotes Bagent_iThe bidding strategy of (1). The strategy of Bagent is to bid b_iOf the vector of (1), the element b thereof_i,jIs for Sagent_jThe number of elements, etc. Since one seller has P bids and n sellers have P policiesⁿ. And, S_biIs equal to

Next, S is used_s＝{S_s1,S_s2,…,S_snAs a set of inquiry strategies for n sellers, i.e. S_sA. Definition of S_s-j＝{S_s1,S_s2,…,S_s(j-1),S_s(j+1),…,S_snDescription except Sagent_jOr equal A- { a_jS of }_sAnd (4) other strategies. For batents, S_b＝{S₁,S₂,…,S_pIs its set of bidding strategies, i.e. S_b＝B。S_b-i＝{S_b1,S_b2,…,S_b(i-1),S_b(i+1),…,S_bnS or S_s＝B-{b_iIs except Bagent_iAnd (4) other policies.

Key parameters in EWA:

for theBagents and Sagents, using π_bi(S_bi,S_b-i)＝U_iAnd pi_sj(S_sj,S_s-j)＝U_jIndicating the rewards of Bagents and Sagents, the rewards and the policy of the Agent itself and its opponents. The reward of one broker depends not only on its behavior but also on the bidding and asking strategies of other brokers in the same auction. In the EWA mechanism, there are two basic parameters that can help an Agent update its policy.

1) N (t): n (t) is the "empirical equivalent" of past experience. It is usually not equal to the number of past observations because the importance of experience decreases as the number of auction rounds increases. The update rule is as follows:

N(t)＝ρN(t-1)+1

ρ is the influence factor, which acts to attenuate the influence of past strategies. t is a time factor equal to the iteration or auction round.

2)A(t)：

And

is Bagent_iStrategies of tau and Sagent_jAttraction value of strategy τ. Use of

And

to represent Bagent_iAnd Sagent_jThe significance of the historical experience in selecting a policy. The update rule of τ is:

also, in the same manner as above,

i (x, y) is 1 only when x is equal to y, otherwise I (x, y) is 0. Phi is a parameter as the learning rate.

δ in A (t) is a notional factor that encourages an Agent to explore the environment and attempt to select an unknown strategy. Depending on the award amount, the Agent may adjust its policy to find the most appropriate policy in the double-auction game.

Each policy or price has a corresponding A (t) indicating a desire to select that price as either a bid or ask price. This equation includes experience and observation of others' behavior, which is characteristic of reinforcement learning and belief learning. The EWA algorithm is an intermediate state between reinforcement learning and belief-based learning.

EWA and reinforcement learning:

in reinforcement learning, three parts are involved: state S, action a and reward R. In each step, the Agent observes the current state s (t) s and makes a decision a (t) a according to the current policy. Following this action, the principal gets an immediate reward R (a, S) from the environment, and the state transitions to the next state S (t +1) e S.

The reward update equation for the RL is:

from this equation, the relationship between RL and EWA can be found, i.e. when δ is 0, ρ is 0,

and N (0) ═ 1, a (t) will degrade into an accumulation of r (t) to obtain some features that enhance learning.

EWA and confidence learning:

in the belief learning model, the history of the adversary strategy set is composed of

Representation, which is a set of policies

Is compared to a hypothetical count of observed values (e.g., buyer). The sum of these frequencies is:

the initial a priori beliefs are:

the beliefs are updated by adding a depreciation coefficient p to the actions actually selected by the other participants.

Expressing beliefs with historical beliefs:

the expected return over time period T by choosing strategy τ is:

combining the last two equations, the update rule of the expectation function in belief learning can be obtained as follows:

strategy selection:

the EWA mechanism can switch from belief learning to reinforcement learning, or combine the two methods into one, depending on the different allocations of δ, ρ, Φ. Furthermore, we can find these parameters important.

Each participant who executes the EWA algorithm submits its bid or ask priceTo the broker. The price is calculated by the policy. Thus, the solution to the double beat game is a distributed approach where each low orbit satellite or high altitude platform separately runs its specific EWA algorithm. Each Agent is based on A (t) and is Bagent in the following manner_iSelecting a strategy:

for Sagent_j：

P (t +1) is the probability of selecting strategy P for the next cycle. After receiving all bidding and asking price strategies, the Agent can finally decide how to effectively allocate the computing resources of all high-altitude platforms.

It is crucial that the calculations of the proposed algorithm cannot be too complex due to the limited computational power of low earth orbit satellites and high altitude platforms. What an Agent does in an auction is to select one of 0 to P_{max}The price of (c). Further, after the broker has allocated an allocation plan, the broker resumes the attractiveness of each price. Therefore, the computational complexity of low-earth orbit satellites and high-altitude platforms is o (n). In addition, brokers need to determine a trade between buyers and sellers. The computational complexity of the transaction decision process is O (n)²)。

The double-clap selling algorithm based on the EWA in the air-space information network specifically comprises the following steps:

step 1: input V ═ V_i(L_q)},Cost＝{Cost_j},R,C；

Step 2: broker initializes β, P and α, Conv ═ 0, Flag ═ 0;

and step 3: for each low earth orbit satellite and high altitude platform, if HAP_jOr LEO_iIf the new addition is made, the broker allocates the agents with the initial parameters of A (0) and N (0) for the new addition, otherwise, the former agents and the former parameters are reserved;

and 4, step 4: starting to determine a resource allocation plan in T time;

and 5: repeating the following substeps until Conv ═ 1;

a: based on (18), (19), calculating A and B of all agents;

b: in ascending order a_j∈A；

C: descending order of arrangement b_i,j∈B；

D: let Flag equal to 0, find the intersection (i, j) of two price sequences equal to argmin_(i,j){b_i,j-a_j(ii) removing (i, j); if r is_j≥c_iThen update beta_i＝b_i,j,P_j＝P_j+b_i,j,α_j,i＝c_iFlag 1, price in descending order

Removal of B from sequence B_i,jRepeating the above steps until Flag is 0;

e: if a and B are equal in the last round of transaction, Conv is 1;

f: all agents calculated Cost, β, V, P according to (9), (10);

g: updating A (t) of all the agents;

step 6: assigning α to HAPs, LEOs;

and 7: returning to alpha.

Simulation result and analysis:

convergence performance of double clap selling algorithm based on EWA:

as shown in fig. 3, the EWA algorithm exhibits excellent convergence performance. After approximately 800 double-beat games, the strategy converges for all participants. Furthermore, we find that the total reward does not reach a maximum until the algorithm converges, which indicates that the EWA converges to an optimal solution.

Average cost of seller and average ask of seller:

as shown in fig. 4, it can be seen that the seller gradually increases the average asking price as the average cost increases. As the cost increases, the seller will raise the ask slightly, as shown, although the average cost increases from 8 to 32, the change in ask is not extreme, as the broker will protect private information and the seller will not be able to give an aggressive price. In addition, as the average capacity becomes higher, the seller will lower the average ask, with higher capacity meaning that more low-orbit satellites can be provided with computing resources and earn more profits to offset the cost. In this market, all participants trade under the broker's limits, and the broker's actions can avoid corruption and direct the market win-win.

Average buyer value and average buyer bid:

as shown in FIG. 5, we observed a change in bid price as the average buyer value increased, with little fluctuation, and a smooth increase. While computing resources may bring more profit or value to the buyer, they are willing to pay more money to acquire computing resources to help them beat other resources less valuable than their buyer. However, when their demand is higher, they must also bid higher to compete with others. While the large resource demands generally mean that they can make more profits, the total resources of high altitude platforms are also limited. The high bid for a unit of computing resource means that their task can be completed before other tasks. Thus, they may tie up computing resources to meet their needs in advance.

The invention introduces a high-altitude platform to unload the computing task of the low-orbit satellite so as to reduce the transmission delay. And calculating the resource allocation problem between the near-orbit satellite and the high-altitude platform based on a double-shot selling mechanism. Then, a nash equilibrium search algorithm based on Empirical Weight Attraction (EWA) was designed, which combines the advantages of reinforcement learning and belief learning, and is performed on each participant. Auction participants can obtain information from other opponents and accumulate experience and thought. Thus, they can complete transactions with other participants in a fuzzy environment and maximize the overall interest of the participants. The convergence of this EWA algorithm is verified by simulation.

So far, the technical solutions of the present invention have been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of the present invention is obviously not limited to these specific embodiments. Equivalent changes or substitutions of related technical features can be made by those skilled in the art without departing from the principle of the invention, and the technical scheme after the changes or substitutions can fall into the protection scope of the invention.

Claims (4)

1. A resource allocation method based on double auction game in an aerospace information network is characterized by comprising a low orbit satellite system based on a High Altitude Platform (HAP) and an EWA algorithm mechanism based on double auction game, wherein the low orbit satellite system can directly access computing resources of an adjacent High Altitude Platform, the computing resources are purchased from the High Altitude Platform by a low orbit satellite, a resource trading market is established, and a buyer and a seller can dynamically adjust the bidding and asking price strategies through interaction with the environment based on the EWA algorithm.

2. The method for allocating resources based on the double-beat game in the aerospace information network as claimed in claim 1, wherein M/1 queuing model is adopted in the low-earth orbit satellite system to quantify the urgency degree of low-earth orbit satellite tasks, and λ is used_iRepresenting LEO_iTask arrival rate of, and HAP_jWill be expressed as mu_jIn addition, since the distance between the high altitude platform and the low orbit satellite is short, the main service latency can be attributed to the processing delay, and the total delay of the mission can be expressed as:

equal to the wait time plus the service time; then, the average latency can be calculated as:

according to the formula L_q＝λ_iT_wThe average length of the queue can be expressed as:

each low earth orbit satellite has different traffic demands, a higher λ indicates a higher frequency of arrival of the task, and a smaller μ indicates that the task requires a large amount of computation, particularly when L_qVery large means that the low earth orbit satellites have very urgent access to additional computing resources.

3. The method for allocating resources based on the double-auction game in the aerospace information network according to claim 1, wherein a double-auction game mechanism is adopted in the low-earth orbit satellite system, and a ═ is used_j1,2,3 … N to describe the asking price strategy for high-altitude platforms, where a_jRepresents SS_jAcceptable minimum price; use of b_i＝{b_i,j1,2,3 … N to describe LEO_iBidding prices for different high-altitude platforms; using the matrix B ═ B_i1,2,3 … M represents the bidding strategy for all low orbit satellites; beta is a_i,jFor representing LEO_iPaying HAP for computing resources_jThe amount of (c); an aerial platform has several items available for sale to N buyers, so that β is a matrix whose columns record the HAPs_jA trade price that sells its computing resources to a certain low earth orbit satellite, such that:

is HAP_jTotal reward in all successful transactions.

4. The resource allocation method based on the double-auction game in the aerospace information network according to claim 1, wherein the EWA algorithm mechanism based on the double-auction game introduces Sagents and Bagents which are agents traded by buyers and sellers; the bids and asks range from 0 to P_maxDiscrete and the number of bids and asks P (P ═ P)_max+1)；

Is Sagent_jFor HAP_jThe set of inquiry policies of (1),

meaning that it will select a certain asking price, obviously S_sjIs equal to a_j. Also, in the same manner as above,

denotes Bagent_iThe policy of Bagent is to bid b_iOf the vector of (1), the element b thereof_i,jIs for Sagent_jThe number of Sagents, the number of elements, etc.; one seller has P bids and n sellers, so the number of strategies is PⁿAnd, S_biIs equal to

Using S_s＝{S_s1,S_s2,…,S_snAs a set of inquiry strategies for n sellers, i.e. S_sDefine S as A_s-j＝{S_s1,S_s2,…,S_s(j-1),S_s(j+1),…,S_snDescription except Sagent_jOr equal A- { a_jS of }_sAnd (4) other strategies. For batents, S_b＝{S₁,S₂,…,S_pIs its set of bidding strategies, i.e. S_b＝B。S_b-i＝{S_b1,S_b2,…,S_b(i-1),S_b(i+1),…,S_bnS or S_s＝B-{b_iIs except Bagent_i(ii) a policy other than;

key parameters in EWA:

for Bagents and Sagents, pi is used_bi(S_bi,S_b-i)＝U_iAnd pi_sj(S_sj,S_s-j)＝U_jIndicating the rewards of Bagents and Sagents, the rewards and the policy of the Agent itself and its opponents.

Images