A method for multi-dimensional resource management in a LEO satellite network

Abstract

The application provides a multi-dimensional resource management method in a LEO satellite network, comprising the following steps: step S1, calculating long-term average total power consumption of the LESC network, and constructing an optimization problem P1 for minimizing the long-term average total power consumption; step S2, decoupling the optimization problem P1 to obtain a problem P2; step S3, finding an optimal calculation offloading decision for minimizing the problem P2; step S4, respectively obtaining an optimal CPU frequency of a ground user, an optimal transmission power of the ground user, an optimal wideband allocation of a LEO satellite to the ground user, and an optimal calculation resource allocation of the LEO satellite; and step S5, iteratively performing the optimization problem P1 based on the step S3 and the step S4 until convergence, so as to obtain an optimal resource allocation result of the LEO satellite network with fusion mobile edge computing. The application can quickly determine an optimal resource allocation scheme according to a current network environment, and has high flexibility.

Description

Technical Field

[0001] This invention relates to a multi-dimensional resource management method in LEO satellite networks. Background Technology

[0002] Currently, traditional terrestrial networks are struggling to meet the demands of ubiquitous global connectivity because they cannot fully cover complex terrains such as mountains and oceans. Furthermore, terrestrial network infrastructure is susceptible to damage, which can disrupt user communications. In contrast, Low Earth Orbit (LEO) satellite networks offer unparalleled advantages over terrestrial networks. They not only provide communication services for aircraft, ships, and terrestrial networks but also achieve truly seamless global coverage, becoming an indispensable part of daily life for people in both urban and rural areas.

[0003] However, the development of emerging services such as interactive games and high-definition video has spurred the growth of many computationally intensive applications, which often have high demands on computing efficiency. Therefore, LEO satellite networks not only need to provide network access services to ground users but also assist them in computing various applications. However, the computing power of LEO satellites is limited and may not be able to meet the computing needs of ground users. To address this issue, LEO satellite (LSEC) networks integrating Mobile Edge Computing (MEC) technology have emerged. By deploying MEC servers on LEO satellites, the computing tasks of ground users can be directly offloaded to the LEO satellites for processing, effectively reducing latency and energy consumption in completing computing tasks.

[0004] Currently, some studies have addressed resource allocation in LSEC networks. However, existing research often assumes that the channel environment between LEO satellites and ground users is constant, leading to relatively simple algorithm designs. In reality, the complex communication environment between LEO satellites and ground users, such as time-varying channel states, significantly impacts resource allocation in LSEC networks. Therefore, researching resource allocation in LSEC networks with time-varying communication environments is of great importance. Summary of the Invention

[0005] This invention proposes a multi-dimensional resource management method for LEO satellite networks, which can quickly determine the optimal resource allocation scheme based on the current network environment, and is highly flexible.

[0006] This invention is achieved through the following technical solution:

[0007] A multi-dimensional resource management method in LEO satellite networks includes the following steps:

[0008] Step S1: Calculate the long-term average total power consumption of the LEO satellite network that integrates mobile edge computing technology, and construct an optimization problem P1 that minimizes the long-term average total power consumption. The optimization problem P1 is constrained by the CPU frequency and transmission power of ground users, the CPU frequency and bandwidth allocation ratio of LEO satellites, the queue stability of ground users and LEO satellites, and the computation offloading decision of ground users.

[0009] Step S2: Decouple the optimization problem P1 using the Lyapunov optimization algorithm to obtain problem P2;

[0010] Step S3: Based on the current network environment, use a deep neural network to generate computation offloading decisions for ground users, augment the computation offloading decisions, and bring the augmented computation offloading decisions into problem P2 to find the optimal computation offloading decision that minimizes problem P2.

[0011] Step S4: Given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under these conditions, the optimal CPU frequency, optimal transmission power, optimal bandwidth allocation to ground users, and optimal computing resource allocation to LEO satellites are obtained respectively.

[0012] Step S5: Iterate the optimization problem P1 based on steps S3 and S4 until convergence, so as to obtain the optimal resource allocation result of the LEO satellite network integrating mobile edge computing.

[0013] in, Indicates the first Each time slot.

[0014] Furthermore, in step S1, the long-term average total power consumption of the LEO satellite network integrating mobile edge computing technology is... ,in, For the LEO satellite network to integrate mobile edge computing in the 19th century Total power consumption per time slot Expressing expectations, This represents the total length of all time slots.

[0015] Furthermore, in step S1, the optimization problem P1 is... , in, For the set of CPU frequencies of ground users, For the first CPU frequency of each ground user For the first Maximum CPU frequency for each ground user For the first Transmission power for each ground user For the first Maximum transmission power for each ground user Allocate LEO satellites to the first CPU frequency of each ground user This is the maximum CPU frequency of the LEO satellite. Allocate LEO satellites to the first Bandwidth allocation ratio for each ground user For the first The ground user in the first The queue length for each time slot, For LEO satellites in the The queue length for each time slot, For the first The computational offloading decision for each ground user This refers to the number of ground users.

[0016] Furthermore, in step S2, problem P2 is... ,in, For the first The ground user in the first The queue length for each time slot, For the first Task size for each ground user For LEO satellites in the The queue length for each time slot, The mission size calculated for the LEO satellite, For the first The size of a mission transmitted from a ground user to a LEO satellite. These are control parameters.

[0017] Furthermore, in step S4, given the set of unloading decisions for ground users... , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the optimal CPU frequency for ground users is obtained by solving problem P3, which is: The obtained ground users Optimal CPU frequency for: Among them, middle The set is denoted as , For the first The ground user in the first The queue length for each time slot, For time slot interval, For the first CPU frequency of each ground user For the first Maximum CPU frequency for each ground user For the first The number of CPU revolutions required for a ground user to process a 1-bit task For the first Energy coefficient for each ground user.

[0018] Furthermore, in step S4, given the set of unloading decisions for ground users... , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the optimal transmission power for ground users is obtained by solving problem P4, which is: The optimal transmission rate obtained for ground users for: ,in, ,Will middle The set is denoted as , For bandwidth size, For channel gain, This represents noise density.

[0019] Furthermore, in step S4, given the set of unloading decisions for ground users... , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the CVX solver is used to solve problem P5 to obtain the optimal bandwidth allocation for LEO satellites to ground users. Problem P5 is: .

[0020] Furthermore, in step S4, given the set of unloading decisions for ground users... , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the CVX solver is used to solve problem P6 to obtain the optimal allocation of computational resources for the LEO satellite. Problem P6 is: ,in, For the first The number of CPU revolutions required for a ground user to process a 1-bit task This represents the energy coefficient of the LEO satellite.

[0021] Furthermore, in step S1, the LEO satellite network integrating mobile edge computing is in the... Total power consumption of each time slot for ,in, , .

[0022] Furthermore, in step S3, the order-preserving quantization method is used to augment the computational unloading decision.

[0023] The present invention has the following beneficial effects:

[0024] This invention first constructs an optimization problem P1 to minimize the long-term average total power consumption of an LSEC network. Then, it uses the Lyapunov optimization algorithm to decouple the non-convex optimization problem P1. Next, it generates computation offloading decisions for ground users based on a deep neural network, thereby obtaining the optimal computation offloading decision for minimizing problem P2. Given a set of offloading decisions, a set of transmission power for ground users, a set of CPU frequencies for LEO satellites, and a set of bandwidth allocation ratios allocated to ground users, it calculates the optimal CPU frequency for ground users, the optimal transmission power for ground users, the optimal bandwidth allocation for LEO satellites, and the optimal computational resource allocation for LEO satellites. Based on these results, it iterates on optimization problem P1 until it converges, ultimately obtaining the optimal resource allocation result. This enables rapid computation offloading decisions and determination of the optimal resource allocation scheme based on the current network environment, offering high flexibility, a more reasonable allocation scheme, and minimizing the long-term average total power consumption of the LSEC network while ensuring long-term queue stability. Therefore, compared to existing optimization methods, this invention has better performance. Attached Figure Description

[0025] The present invention will now be described in further detail with reference to the accompanying drawings.

[0026] Figure 1 This is a schematic diagram of a scenario according to the present invention.

[0027] Figure 2 This is a flowchart of the present invention.

[0028] Figure 3 This chart compares the average total power consumption performance of the present invention with that of existing algorithms under different numbers of ground users.

[0029] Figure 4 This chart compares the average total power consumption performance of the present invention with that of existing algorithms under different task arrival rates. Detailed Implementation

[0030] like Figure 1 As shown, the LEO satellite network (LESC network) integrating mobile edge computing technology includes one LEO satellite and... The first ground user, the LEO satellite is at an altitude of 1200 km, the first Maximum transmission power for each ground user , No. Maximum CPU frequency for each ground user , No. The number of CPU revolutions required for a ground user to process a 1-bit task Trans / bit, the Energy coefficient of a ground user The maximum CPU frequency of LEO satellites Time slot interval LEO satellite energy coefficient Control parameters noise density .

[0031] like Figure 2 As shown, the multi-dimensional resource management method in the LEO satellite network includes the following steps:

[0032] Step S1: Calculate the long-term average total power consumption of the LEO satellite network (LESC network) that integrates mobile edge computing technology, and construct an optimization problem P1 that minimizes the long-term average total power consumption. The optimization problem P1 is constrained by the CPU frequency and transmission power of ground users, the CPU frequency and bandwidth allocation ratio of LEO satellites, the queue stability of ground users and LEO satellites, and the computation offloading decision of ground users.

[0033] The long-term average total power consumption of the LESC network is ,in, For LSEC networks in the first Total power consumption per time slot For the first The ground user in the first The power consumption of processing computing tasks in each time slot. For LEO satellites in the The power consumption of processing computing tasks in each time slot. For the first The computational offloading decision for each ground user For the first Transmission power for each ground user Expressing expectations, The total length of all time slots;

[0034] The optimization problem P1 is specifically as follows: ,in, For the set of unloading decisions for ground users, For the set of CPU frequencies of ground users, For the set of transmission power for ground users, This is a set of CPU frequencies for LEO satellites. The set of bandwidth allocation ratios assigned to terrestrial users, where "st" represents a constraint condition, and the symbol " "Indicates arbitrary, (C1) is the CPU frequency constraint for ground users, For the first CPU frequency of each ground user For the first The maximum CPU frequency for each ground user, (C2) represents the transmission power constraint for each ground user. For the first Transmission power for each ground user For the first The maximum transmission power for each ground user, (C3) is the CPU frequency constraint of the LEO satellite. Allocate LEO satellites to the first CPU frequency of each ground user Where (C4) is the maximum CPU frequency of the LEO satellite, and (C4) is the bandwidth allocation ratio constraint for the LEO satellite. Allocate LEO satellites to the first The bandwidth allocation ratio for each ground user, (C5) represents the queue stability constraint for ground users. For the first The ground user in the first The queue length for each time slot, (C6) is the queue stability constraint for LEO satellites. For LEO satellites in the The queue length for each time slot, (C7) represents the computational offloading decision constraint for ground users. For the first The computational offloading decision for each ground user This refers to the number of ground users.

[0035] Step S2: Since the objective function and constraints of optimization problem P1 both contain binary and continuous variables, this problem is non-convex. Therefore, the Lyapunov optimization algorithm is used to decouple optimization problem P1, resulting in problem P2: ,in, For the first The ground user in the first The queue length for each time slot, For the first Task size for each ground user For LEO satellites in the The queue length for each time slot, The mission size calculated for the LEO satellite, For the first The size of a mission transmitted from a ground user to a LEO satellite. For control parameters, The constraints (C1), (C2), (C3), and (C4) in the optimization problem P1, and (C7) refers to the constraint (C7) in the optimization problem P1.

[0036] Step S3: First, based on the current network environment, use a deep neural network to generate computational offloading decisions for ground users. To obtain a high-quality computational offloading decision, the computational offloading decision is augmented using the order-preserving quantization method (i.e.,... Expand to Group computing unloading decision ), and then bring the augmented computational unloading decision into problem P2, fix other optimization variables, and find the optimal computational unloading decision that minimizes problem P2;

[0037] Step S4: Given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under these conditions, the optimal CPU frequency, optimal transmission power, optimal bandwidth allocation to ground users, and optimal computing resource allocation to LEO satellites are obtained respectively.

[0038] More specifically, the optimal CPU frequency for ground users is obtained by solving problem P3, which is: ,Will , Trans / bit , Substituting into problem P3, we obtain the optimal CPU frequency for ground user k. for: , among which, will middle The set is denoted as , For the first The ground user in the first The queue length for each time slot, For time slot interval, For the first CPU frequency of each ground user For the first CPU frequency of each ground user For the first Maximum CPU frequency for each ground user For the first The number of CPU revolutions required for a ground user to process a 1-bit task For the first Energy coefficient for each ground user;

[0039] The optimal transmission power for ground users is obtained by solving problem P4, which is: ,Will , , Substituting into problem P4, we obtain the optimal transmission rate for ground users. for: ,in, ,Will middle The set is denoted as , For bandwidth size, For channel gain, Noise density;

[0040] The CVX solver is used to solve problem P5 to obtain the optimal bandwidth allocation for LEO satellites to ground users. Problem P5 is: ;

[0041] The CVX solver is used to solve problem P6 to obtain the optimal allocation of computational resources for the LEO satellite. Problem P6 is: ,in, For the first The number of CPU revolutions required for a ground user to process a 1-bit task This represents the energy coefficient of the LEO satellite.

[0042] Step S5: Based on steps S3 and S4, iterate the optimization problem P1 until convergence to obtain the optimal resource allocation result of the LEO satellite network integrating mobile edge computing. The specific iterative process is the existing technology.

[0043] Figure 3In this study, under varying numbers of ground users, the average total power consumption performance of the proposed method (DRLCO method) is compared with that of existing methods such as NONA (non-augmented method), DQN (deep Q-network method), and RRAA (fixed resource allocation method). It can be seen that the average total power consumption of the four methods increases with the increase in the number of ground users. However, the DRLCO algorithm proposed in this invention has the lowest average total power consumption, indicating that the proposed DRLCO algorithm has better performance.

[0044] Figure 4 In this study, the average total power consumption performance of the DRLCO method proposed in this invention is compared with that of existing methods such as NONA, DQN, and RRAA under different task arrival rates. It can be seen that the average total power consumption of all four methods increases with the increase of the task arrival rate. This is because the increased task arrival rate increases the number of tasks to be processed in the LSEC network, thus requiring more power from ground users and LEO satellites to process the tasks in the queue. However, compared with the four methods, the DRLCO algorithm proposed in this invention has the lowest average total power consumption, indicating that the proposed DRLCO algorithm can effectively reduce the average total power consumption of the LSEC network.

[0045] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made in accordance with the scope of the patent application and the contents of the specification of the present invention should still fall within the scope of the patent of the present invention.

Claims

1. A multi-dimensional resource management method in a LEO satellite network, characterized in that: Includes the following steps: Step S1: Calculate the long-term average total power consumption of the LEO satellite network that integrates mobile edge computing technology, and construct an optimization problem P1 that minimizes the long-term average total power consumption. The optimization problem P1 is constrained by the CPU frequency and transmission power of ground users, the CPU frequency and bandwidth allocation ratio of LEO satellites, the queue stability of ground users and LEO satellites, and the computation offloading decision of ground users. Step S2: Decouple the optimization problem P1 using the Lyapunov optimization algorithm to obtain problem P2; Step S3: Based on the current network environment, use a deep neural network to generate computation offloading decisions for ground users, augment the computation offloading decisions, and bring the augmented computation offloading decisions into problem P2 to find the optimal computation offloading decision that minimizes problem P2. Step S4: Given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under these conditions, the optimal CPU frequency, optimal transmission power, optimal bandwidth allocation to ground users, and optimal computing resource allocation to LEO satellites are obtained respectively. Step S5: Iterate the optimization problem P1 based on steps S3 and S4 until convergence to obtain the optimal resource allocation result of the LEO satellite network integrating mobile edge computing. Indicates the first One time slot; In step S1, the long-term average total power consumption of the LEO satellite network integrating mobile edge computing technology is: ,in, For the LEO satellite network to integrate mobile edge computing in the 19th century Total power consumption per time slot Expressing expectations, The total length of all time slots; In step S1, the optimization problem P1 is: ， in, For the set of CPU frequencies of ground users, For the first CPU frequency of each ground user For the first Maximum CPU frequency for each ground user For the first Transmission power for each ground user For the first Maximum transmission power for each ground user Allocate LEO satellites to the first CPU frequency of each ground user This is the maximum CPU frequency of the LEO satellite. Allocate LEO satellites to the first Bandwidth allocation ratio for each ground user For the first The ground user in the first The queue length for each time slot, For LEO satellites in the The queue length for each time slot, For the first The computational offloading decision for each ground user This refers to the number of ground users.

2. The multi-dimensional resource management method in a LEO satellite network according to claim 1, characterized in that: In step S2, problem P2 is ,in, For the first The ground user in the first The queue length for each time slot, For the first Task size for each ground user For LEO satellites in the The queue length for each time slot, The mission size calculated for the LEO satellite, For the first The size of a mission transmitted from a ground user to a LEO satellite. These are control parameters.

3. The multi-dimensional resource management method in a LEO satellite network according to claim 1, characterized in that: In step S4, given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the optimal CPU frequency for ground users is obtained by solving problem P3, which is: The obtained ground users Optimal CPU frequency for: Among them, middle The set is denoted as , For the first The ground user in the first The queue length for each time slot, For time slot interval, For the first CPU frequency of each ground user For the first Maximum CPU frequency for each ground user For the first The number of CPU revolutions required for a ground user to process a 1-bit task For the first Energy coefficient for each ground user.

4. A multi-dimensional resource management method in a LEO satellite network according to claim 1, 2, or 3, characterized in that: In step S4, given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the optimal transmission power for ground users is obtained by solving problem P4, which is: The optimal transmission rate obtained for ground users for: ,in, ,Will middle The set is denoted as , For bandwidth size, For channel gain, This represents noise density.

5. A multi-dimensional resource management method in a LEO satellite network according to claim 1, 2, or 3, characterized in that: In step S4, given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the CVX solver is used to solve problem P5 to obtain the optimal bandwidth allocation for LEO satellites to ground users. Problem P5 is: .

6. A multi-dimensional resource management method in a LEO satellite network according to claim 1, 2, or 3, characterized in that: In step S4, given the set of unloading decisions for ground users , set of transmission power for ground users CPU frequency set of LEO satellites and the set of bandwidth allocation ratios allocated to ground users Under the given conditions, the CVX solver is used to solve problem P6 to obtain the optimal allocation of computational resources for the LEO satellite. Problem P6 is: ,in, For the first The number of CPU revolutions required for a ground user to process a 1-bit task This represents the energy coefficient of the LEO satellite.

7. A multi-dimensional resource management method in a LEO satellite network according to claim 2 or 3, characterized in that: In step S1, the LEO satellite network integrating mobile edge computing is in the... Total power consumption of each time slot for ,in, , .

8. A multi-dimensional resource management method in a LEO satellite network according to claim 1, 2, or 3, characterized in that: In step S3, the order-preserving quantization method is used to augment the computational unloading decision.

Images