A method and system for minimizing resource allocation in a low-orbit remote sensing satellite network

Abstract

The invention relates to the field of satellite communication technology, and specifically discloses a method and system for allocating resources with minimal delay in a low-orbit remote sensing satellite network, comprising: obtaining the configuration of a satellite constellation and information of each satellite, and determining a satellite scheduling plan; establishing a network model and a delay model according to the satellite scheduling plan, and collecting information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send instructions for unloading computing tasks to other satellites; according to the instructions for unloading computing tasks, making multiple computing satellites form a satellite computing cluster and determining the shortest computing time of the remote sensing satellite; using a dual method and a KKT condition to calculate a feasible solution for optimal resource allocation of satellite communication and computing resources within a satellite cluster under the shortest computing time, and determining the delay of the satellite computing cluster according to the feasible solution; and setting a reward function to minimize the sum of the delays of each satellite computing cluster to determine the optimal resource allocation plan.

Description

Technical Field

[0001] The present invention relates to the field of satellite communication technology, and in particular to a method and system for allocating resources with minimal delay in a low-orbit remote sensing satellite network. Background Art

[0002] With the rapid development and widespread application of space technology, the deployment of low-orbit remote sensing satellite systems has become the only way for my country to move from a space power to a space power. The low-orbit remote sensing satellite system is equipped with a variety of remote sensing business payloads such as high-resolution cameras, receives and analyzes electromagnetic wave signals emitted by earth targets, and relies on global all-weather observation capabilities to accurately monitor the physical characteristics and changing laws of ground targets. As an important means of obtaining near-Earth space information, the low-orbit remote sensing satellite system is not only the trend of remote sensing technology and industrial development, but also the planning focus of many countries.

[0003] Current demand for remote sensing services: With the continuous improvement of remote sensing satellite sensor technology and the improvement of image resolution, the demand for high-resolution remote sensing services has grown rapidly, and the amount of remote sensing data has grown exponentially. A high-resolution remote sensing image can even reach tens or hundreds of GB. The amount of data generated by high-resolution remote sensing satellites every day is as high as tens of TB. Especially in emergency situations such as disaster relief, remote sensing satellite systems urgently need to achieve real-time data collection and rapid response within minutes to meet real-time remote sensing needs. Therefore, remote sensing satellite systems must not only have high-precision data perception capabilities, but also build efficient and massive emergency remote sensing data transmission and processing infrastructure to ensure that data can be collected, transmitted and processed in the shortest time, thereby realizing the remote sensing vision of "immediate sensing and transmission" of remote sensing data.

[0004] The dilemma of traditional remote sensing systems: Traditional remote sensing models are isolated in perception, transmission and processing, the global layout of gateways is limited, and the remote sensing model of "first upload, then process" and other difficulties have led to cumbersome and inefficient processing links for traditional remote sensing services, and the response delay cannot meet the needs of modern remote sensing. Compared with the US emergency remote sensing system, which only takes 12 minutes to complete the response, the current remote sensing system in my country takes up to 20 to 30 hours to respond to overseas emergency remote sensing, and there is an urgent need for remote sensing breakthroughs. Summary of the invention

[0005] To achieve the purpose of the present invention, the present application provides a method for allocating resources with minimal delay in a low-orbit remote sensing satellite network, comprising: Step S1: Obtain the configuration of the satellite constellation and the location information, communication resource information and computing resource information of each satellite, and determine the satellite scheduling plan; Step S2: establishing a network model and a delay model according to the satellite scheduling scheme, and collecting information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send an instruction to offload computing tasks to the other satellites; Step S3: according to the instruction of unloading the computing task, multiple computing satellites are made to form a satellite computing cluster and the shortest computing time of the remote sensing satellite is determined; Step S4: using the Lagrange duality method and the Karush-Kuhn-Tucker condition to calculate and derive a feasible solution for optimal resource allocation of satellite communication and computing resources within the satellite cluster with the shortest computing time, and determining the delay of the satellite computing cluster according to the feasible solution; Step S5: By setting the reward function to minimize the sum of the delays of each satellite computing cluster, and repeating steps S1-S4 until the algorithm converges, the optimal solution for optimal resource allocation is obtained, and the optimal resource allocation plan is determined based on the optimal solution.

[0006] In some specific embodiments, in step S1, The location information includes: the latitude and longitude coordinates of the satellite, the Cartesian coordinates of the satellite, and the adjacent satellites in the same orbit and adjacent orbits of each satellite; The communication resource information includes: the task allocation ratio of satellites to transmit information to other satellites, the power of transmitters and the transmission rate of tasks; The computing resource information includes: the computing resources of the CPU allocated by each computing satellite to each subtask of the low-orbit remote sensing satellite.

[0007] In some specific embodiments, step S2 further includes: The parameters of the satellite constellation are determined based on the neural network model, including: the total number of satellites, orbital plane, and orbital inclination; and the orbital parameters of each satellite are determined based on the orbital characteristics of the satellite constellation, including: the satellite's position, coverage, and communication link.

[0008] In some specific embodiments, step S2 further includes: Step S21: define hyperparameters, including: number of multi-agents, state space dimension, action space dimension, maximum number of steps, maximum number of epochs, discount factor, learning rate, batch size, number of parallel computing cores, clipping parameters and action entropy coefficient; Step S22: Initialize the policy network of the network model, including the number of network layers, input and output dimensions, and masks; Step S23: Initialize the experience replay pool of the network model to store the state, action, reward and new state of the agent during the training process; Step S24: creating multiple independent environments, each environment corresponding to a low-orbit remote sensing satellite; Step S25: Using the satellite's observation data as input, each low-orbit remote sensing satellite generates an action according to the current strategy; Step S26: Execute the action, obtain the reward and the new state, standardize the reward to reduce the variance in the gradient calculation, improve the stability of the training, and store the state, action, reward and new state into the experience replay pool; Step S27: Calculate the loss and update the policy network, using the loss function and policy update method in the PPO algorithm; Step S28: Based on the calculated loss, use the Adam optimizer to update the parameters of the policy network; Step S29: storing the data of each low-orbit remote sensing satellite during the training process into the experience playback pool for subsequent batch processing and strategy update; Step S30: Repeat steps S25-S29 until the network model traversal times reaches the maximum epoch number.

[0009] In some specific embodiments, in step S3: the shortest computing time of the remote sensing satellite is: when the unloading decision of the remote sensing satellite is determined, when the computing time of all computing satellites in the satellite cluster is the same, the computing time is the shortest.

[0010] To achieve the same invention purpose, the present application also provides a resource allocation system for minimizing delay in a low-orbit remote sensing satellite network, comprising: Satellite information acquisition module: used to obtain the configuration of the satellite constellation and the location information, communication resource information and computing resource information of each satellite, and determine the satellite scheduling plan; Instruction issuing module: used to establish a network model and a delay model according to the satellite scheduling plan, and collect information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send instructions to unload computing tasks to the other satellites; Satellite computing module: used for making a plurality of computing satellites form a satellite computing cluster and determining the shortest computing time of the remote sensing satellite according to the instruction of unloading computing tasks; The delay determination module is used to use the Lagrange duality method and the Karush-Kuhn-Tucker condition to calculate and derive a feasible solution for optimal resource allocation of satellite communication and computing resources within a satellite cluster with the shortest computing time of the delay, and determine the delay of the satellite computing cluster according to the feasible solution; Delay minimization module: used to minimize the sum of delays of each satellite computing cluster by setting a reward function, and repeatedly execute the satellite information acquisition module, instruction issuing module, satellite computing module, and delay determination module in sequence until the algorithm converges, obtains the optimal solution for optimal resource allocation, and determines the optimal resource allocation plan based on the optimal solution.

[0011] In some specific embodiments, the satellite information acquisition module is used to: The location information includes: the latitude and longitude coordinates of the satellite, the Cartesian coordinates of the satellite, and the adjacent satellites in the same orbit and adjacent orbits of each satellite; The communication resource information includes: the task allocation ratio of satellites to transmit information to other satellites, the power of transmitters and the transmission rate of tasks; The computing resource information includes: the computing resources of the CPU allocated by each computing satellite to each subtask of the low-orbit remote sensing satellite.

[0012] In some specific embodiments, the instruction issuing module is also used to: The parameters of the satellite constellation are determined based on the neural network model, including: the total number of satellites, orbital plane, and orbital inclination; and the orbital parameters of each satellite are determined based on the orbital characteristics of the satellite constellation, including: the satellite's position, coverage, and communication link.

[0013] In some specific embodiments, the instruction issuing module is further used to perform the following steps: Step S21: define hyperparameters, including: number of multi-agents, state space dimension, action space dimension, maximum number of steps, maximum number of epochs, discount factor, learning rate, batch size, number of parallel computing cores, clipping parameters and action entropy coefficient; Step S22: Initialize the policy network of the network model, including the number of network layers, input and output dimensions, and masks; Step S23: Initialize the experience replay pool of the network model to store the state, action, reward and new state of the agent during the training process; Step S24: creating multiple independent environments, each environment corresponding to a low-orbit remote sensing satellite; Step S25: Using the satellite's observation data as input, each low-orbit remote sensing satellite generates an action according to the current strategy; Step S26: Execute the action, obtain the reward and the new state, standardize the reward to reduce the variance in the gradient calculation, improve the stability of the training, and store the state, action, reward and new state into the experience replay pool; Step S27: Calculate the loss and update the policy network, using the loss function and policy update method in the PPO algorithm; Step S28: Based on the calculated loss, use the Adam optimizer to update the parameters of the policy network; Step S29: storing the data of each low-orbit remote sensing satellite during the training process into the experience playback pool for subsequent batch processing and strategy update; Step S30: Repeat steps S25-S29 until the network model traversal times reaches the maximum epoch number.

[0014] In some specific embodiments, in the satellite computing module: The shortest computing time of the remote sensing satellite is: when the unloading decision of the remote sensing satellite is determined, when the computing time of all computing satellites in the satellite cluster is the same, the computing time is the shortest.

[0015] Beneficial effects of the above technical solution: In order to overcome the limitations of traditional remote sensing satellites, this application explores a method and system for deep integration of remote sensing with satellite communications and edge computing, relying on global low-orbit satellite communications, integrating remote sensing, communication and computing resources, and building a remote-computing fusion system. Specifically, remote sensing satellites act as perception nodes in this system, and quickly forward remote sensing instructions and data by accessing the satellite communication network, so that remote sensing operations can be freed from the limitations of the global ground station layout, and remote sensing tasks can be executed "anytime, anywhere", effectively improving the response efficiency. The computing satellite is responsible for processing remote sensing satellite data, and the on-board processing of remote sensing tasks such as remote sensing data target identification and change detection can be completed by using multi-satellite distributed collaboration. By prepending computing resources to the edge of the network, the impact of massive remote sensing data on the gateway is avoided, the bandwidth of satellite-to-ground communications is saved, the system's dependence on central bandwidth is reduced, and the processing link is simplified. The generated remote sensing information products can be forwarded to remote sensing users via communication satellites, thereby significantly improving the response efficiency of remote sensing data. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A schematic flow chart of a method for allocating resources with minimal delay in a low-orbit remote sensing satellite network provided by an embodiment of the present invention; Figure 2 A schematic structural diagram of a delay minimization resource allocation system in a low-orbit remote sensing satellite network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0019] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar symbols throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0020] Embodiment 1 An embodiment of the present invention provides a method for allocating resources with minimal delay in a low-orbit remote sensing satellite network, referring to Figure 1 As shown, including: Step S1: Obtain the configuration of the satellite constellation and the location information, communication resource information and computing resource information of each satellite, and determine the satellite scheduling plan; In a specific embodiment of the present invention, in step S1, The location information includes: the latitude and longitude coordinates of the satellite, the Cartesian coordinates of the satellite, and the adjacent satellites in the same orbit and adjacent orbits of each satellite; The communication resource information includes: the task allocation ratio of satellites to transmit information to other satellites, the power of transmitters and the transmission rate of tasks; The computing resource information includes: the computing resources of the CPU allocated by each computing satellite to each subtask of the low-orbit remote sensing satellite.

[0021] Step S2: establishing a network model and a delay model according to the satellite scheduling scheme, and collecting information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send an instruction to offload computing tasks to the other satellites; In a specific embodiment of the present invention, step S2 further includes: The parameters of the satellite constellation are determined based on the neural network model, including: the total number of satellites, orbital plane, and orbital inclination; and the orbital parameters of each satellite are determined based on the orbital characteristics of the satellite constellation, including: the satellite's position, coverage, and communication link.

[0022] In a specific embodiment of the present invention, step S2 further includes: Step S21: define hyperparameters, including: number of multi-agents, state space dimension, action space dimension, maximum number of steps, maximum number of epochs, discount factor, learning rate, batch size, number of parallel computing cores, clipping parameters and action entropy coefficient; Step S22: Initialize the policy network of the network model, including the number of network layers, input and output dimensions, and masks; Step S23: Initialize the experience replay pool of the network model to store the state, action, reward and new state of the agent during the training process; Step S24: creating multiple independent environments, each environment corresponding to a low-orbit remote sensing satellite; Step S25: Using the satellite's observation data as input, each low-orbit remote sensing satellite generates an action according to the current strategy; Step S26: Execute the action, obtain the reward and the new state, standardize the reward to reduce the variance in the gradient calculation, improve the stability of the training, and store the state, action, reward and new state into the experience replay pool; Step S27: Calculate the loss and update the policy network, using the loss function and policy update method in the PPO algorithm; Step S28: Based on the calculated loss, use the Adam optimizer to update the parameters of the policy network; Step S29: storing the data of each low-orbit remote sensing satellite during the training process into the experience playback pool for subsequent batch processing and strategy update; Step S30: Repeat steps S25-S29 until the network model traversal times reaches the maximum epoch number.

[0023] Specifically, the MAPPO algorithm is used to enable multiple satellites to simultaneously learn the decision-making strategy learning problem. Each satellite can only collect local observation data from the environment. This application provides a partially observable Markov decision process model with state space, action space and reward, using a tuple (S, A, R) to represent an MDP (Markov Decision Process) decision process. The specific content of the tuple is as follows: Action space: The action space is defined as the set of edge satellites from which the master can offload data. Each edge satellite m can be a set of Any satellite in can provide offloading or not, and a set of arrays is used to represent the action space, which is mathematically expressed as: ; State space: Since the power of the satellite is limited, the capacity of the edge satellite is also limited. Therefore, the state of the edge satellite m can be regarded as a multidimensional array, defined as: ; in, F m represents the satellite’s computing resources, d m Represents the position of the satellite in the coordinate system.

[0024] Reward function: The core purpose of the reward function is to minimize the maximum delay of the system while satisfying the power and delay constraints. If the constraints are met, the shorter the time, the greater the reward. If the output does not meet the constraints, the reward function is a negative number, which can be regarded as a penalty. The mathematical expression of the reward function is: ; in and are two predetermined constants, a Indicates the action of satellite selection. S Indicates the current state. T Indicates the delay required to complete the task.

[0025] The MAPPO algorithm is a distributed learning algorithm that uses DRL to optimize the unloading strategy of the agent in the unloading decision-making process. In this method, each edge satellite is regarded as an agent, and the agent receives reward signals and local state observation data from the network controller. Under the guidance of the MAPPO algorithm, the agent learns its strategy. This algorithm uses two neural networks: the policy network and the value network. The former is responsible for the decision-making of the agent, and the latter is responsible for the value assessment function. The data generated during the transition process, including state, action, reward, and next state, are stored in the experience pool for subsequent parameter updates of the neural network. In order to effectively reduce the computational complexity, this application introduces an attention mechanism, which only gives higher weights to inputs that provide key information to the agent, ensuring efficient use of information.

[0026] Step S3: according to the instruction of unloading the computing task, multiple computing satellites are made to form a satellite computing cluster and the shortest computing time of the remote sensing satellite is determined; In a specific embodiment of the present invention, step S3 includes: the shortest computing time of the remote sensing satellite is: when the unloading decision of the remote sensing satellite is determined, when the computing time of all computing satellites in the satellite cluster is the same, the computing time is the shortest.

[0027] Step S4: using the Lagrange duality method and the Karush-Kuhn-Tucker condition to calculate and derive a feasible solution for optimal resource allocation of satellite communication and computing resources within the satellite cluster with the shortest computing time, and determining the delay of the satellite computing cluster according to the feasible solution; Step S5: By setting the reward function to minimize the sum of the delays of each satellite computing cluster, and repeating steps S1-S4 until the algorithm converges, the optimal solution for optimal resource allocation is obtained, and the optimal resource allocation plan is determined based on the optimal solution.

[0028] Embodiment 2 An embodiment of the present invention provides a system for allocating resources with minimal delay in a low-orbit remote sensing satellite network, referring to Figure 2 As shown, including: Satellite information acquisition module 10: used to obtain the configuration of the satellite constellation and the location information, communication resource information and computing resource information of each satellite, and determine the satellite scheduling plan; The instruction issuing module 20 is used to establish a network model and a delay model according to the satellite scheduling plan, and collect information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send an instruction to unload the computing task to the other satellites; Satellite computing module 30: used for making a plurality of computing satellites form a satellite computing cluster and determining the shortest computing time of the remote sensing satellite according to the instruction of unloading the computing task; The delay determination module 40 is used to derive a feasible solution for optimal resource allocation of satellite communication and computing resources within a satellite cluster under the shortest delay computing time by using the Lagrange duality method and the Karush-Kuhn-Tucker computing condition, and determine the delay of the satellite computing cluster according to the feasible solution; The delay minimization module 50 is used to minimize the sum of the delays of each satellite computing cluster by setting a reward function, and repeatedly execute the satellite information acquisition module, the instruction issuing module, the satellite computing module, and the delay determination module in sequence until the algorithm converges, obtains the optimal solution for the optimal resource allocation, and determines the optimal resource allocation plan based on the optimal solution.

[0029] In a specific embodiment of the present invention, the satellite information acquisition module 10 is used to: The location information includes: the latitude and longitude coordinates of the satellite, the Cartesian coordinates of the satellite, and the adjacent satellites in the same orbit and adjacent orbits of each satellite; The communication resource information includes: the task allocation ratio of satellites to transmit information to other satellites, the power of transmitters and the transmission rate of tasks; The computing resource information includes: the computing resources of the CPU allocated by each computing satellite to each subtask of the low-orbit remote sensing satellite.

[0030] In a specific embodiment of the present invention, the instruction issuing module 20 is further used for: The parameters of the satellite constellation are determined based on the neural network model, including: the total number of satellites, orbital plane, and orbital inclination; and the orbital parameters of each satellite are determined based on the orbital characteristics of the satellite constellation, including: the satellite's position, coverage, and communication link.

[0031] In a specific embodiment of the present invention, the instruction issuing module 20 is further configured to perform the following steps: Step S21: define hyperparameters, including: number of multi-agents, state space dimension, action space dimension, maximum number of steps, maximum number of epochs, discount factor, learning rate, batch size, number of parallel computing cores, clipping parameters and action entropy coefficient; Step S22: Initialize the policy network of the network model, including the number of network layers, input and output dimensions, and masks; Step S23: Initialize the experience replay pool of the network model to store the state, action, reward and new state of the agent during the training process; Step S24: creating multiple independent environments, each environment corresponding to a low-orbit remote sensing satellite; Step S25: Using the satellite's observation data as input, each low-orbit remote sensing satellite generates an action according to the current strategy; Step S26: Execute the action, obtain the reward and the new state, standardize the reward to reduce the variance in the gradient calculation, improve the stability of the training, and store the state, action, reward and new state into the experience replay pool; Step S27: Calculate the loss and update the policy network, using the loss function and policy update method in the PPO algorithm; Step S28: Based on the calculated loss, use the Adam optimizer to update the parameters of the policy network; Step S29: storing the data of each low-orbit remote sensing satellite during the training process into the experience playback pool for subsequent batch processing and strategy update; Step S30: Repeat steps S25-S29 until the network model traversal times reaches the maximum epoch number.

[0032] In a specific embodiment of the present invention, in the satellite computing module 30: The shortest computing time of the remote sensing satellite is: when the unloading decision of the remote sensing satellite is determined, when the computing time of all computing satellites in the satellite cluster is the same, the computing time is the shortest.

[0033] This application explores a method and system for deep integration of remote sensing, satellite communications, and edge computing. It relies on global low-orbit satellite communications, integrates remote sensing, communication, and computing resources, and builds a remote-computing fusion system. Specifically, remote sensing satellites act as perception nodes in this system. By accessing the satellite communication network, they quickly forward remote sensing instructions and data, freeing remote sensing operations from the limitations of global ground station layouts, and realizing "anytime, anywhere" remote sensing task execution, effectively improving response efficiency. Computing satellites are responsible for processing remote sensing satellite data, and can use multi-satellite distributed collaboration to complete on-board processing of remote sensing tasks such as remote sensing data target identification and change detection. By prepending computing resources to the edge of the network, the impact of massive remote sensing data on the gateway is avoided, the bandwidth of satellite-to-ground communications is saved, the system's dependence on central bandwidth is reduced, and the processing link is simplified. The generated remote sensing information products can be forwarded to remote sensing users via communication satellites, thereby significantly improving the response efficiency of remote sensing data.

[0034] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0035] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referenced to each other. The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing terminal device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes. Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the attached claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention. Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0036] The method and device provided by the present invention are introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

[0037] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", "one specific embodiment" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0038] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for allocating resources with minimal delay in a low-orbit remote sensing satellite network, characterized in that: include: Step S1: Obtain the configuration of the satellite constellation and the location information, communication resource information and computing resource information of each satellite, and determine the satellite scheduling plan; Step S2: establishing a network model and a delay model according to the satellite scheduling scheme, and collecting information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send an instruction to offload computing tasks to the other satellites; Step S3: according to the instruction of unloading the computing task, multiple computing satellites are made to form a satellite computing cluster, and the shortest computing time of the remote sensing satellite is determined; Step S4: using the Lagrange dual method and the Karush-Kuhn-Tucker condition to calculate a feasible solution for optimal resource allocation of satellite communication and computing resources within the satellite cluster with the shortest computing time of the delay, and determining the delay of the satellite computing cluster according to the feasible solution; Step S5: By setting the reward function to minimize the sum of the delays of each satellite computing cluster, and repeating steps S1-S4 until the algorithm converges, the optimal solution for optimal resource allocation is obtained, and the optimal resource allocation plan is determined based on the optimal solution.

2. The method for allocating resources with minimal delay in a low-orbit remote sensing satellite network according to claim 1, characterized in that: In step S1, The location information includes: the latitude and longitude coordinates of the satellite, the Cartesian coordinates of the satellite, and the adjacent satellites in the same orbit and adjacent orbits of each satellite; The communication resource information includes: the task allocation ratio of satellites to transmit information to other satellites, the power of transmitters and the transmission rate of tasks; The computing resource information includes: the computing resources of the CPU allocated by each computing satellite to each subtask of the low-orbit remote sensing satellite.

3. The method for allocating resources with minimal delay in a low-orbit remote sensing satellite network according to claim 1, characterized in that: Step S2 also includes: The parameters of the satellite constellation are determined based on the neural network model, including: the total number of satellites, orbital plane, and orbital inclination; and based on the orbital characteristics of the satellite constellation, the orbital parameters of each satellite are determined, including the satellite's position, coverage, and communication link.

4. The method for allocating resources with minimal delay in a low-orbit remote sensing satellite network according to claim 1, characterized in that: Step S2 also includes: Step S21: define hyperparameters, including: number of multi-agents, state space dimension, action space dimension, maximum number of steps, maximum number of epochs, discount factor, learning rate, batch size, number of parallel computing cores, clipping parameters and action entropy coefficient; Step S22: Initialize the policy network of the network model, including the number of network layers, input and output dimensions, and masks; Step S23: Initialize the experience replay pool of the network model to store the state, action, reward and new state of the agent during the training process; Step S24: creating multiple independent environments, each environment corresponding to a low-orbit remote sensing satellite; Step S25: Using the satellite's observation data as input, each low-orbit remote sensing satellite generates an action according to the current strategy; Step S26: Execute the action, obtain the reward and the new state, standardize the reward to reduce the variance in the gradient calculation, improve the stability of the training, and store the state, action, reward and new state into the experience replay pool; Step S27: Calculate the loss and update the policy network, using the loss function and policy update method in the PPO algorithm; Step S28: Based on the calculated loss, use the Adam optimizer to update the parameters of the policy network; Step S29: storing the data of each low-orbit remote sensing satellite during the training process into the experience playback pool for subsequent batch processing and strategy update; Step S30: Repeat steps S25-S29 until the network model traversal times reaches the maximum epoch number.

5. The method for allocating resources with minimal delay in a low-orbit remote sensing satellite network according to claim 1, characterized in that: In step S3, the shortest computing time of the remote sensing satellite is: when the unloading decision of the remote sensing satellite is determined, when the computing time of all computing satellites in the satellite cluster is the same, the computing time is the shortest.

6. A delay minimization resource allocation system in a low-orbit remote sensing satellite network, characterized in that: include: Satellite information acquisition module: used to obtain the configuration of the satellite constellation and the location information, communication resource information and computing resource information of each satellite, and determine the satellite scheduling plan; Instruction issuing module: used to establish a network model and a delay model according to the satellite scheduling plan, and collect information of other satellites in the network through each low-orbit remote sensing satellite carrying remote sensing data to determine whether to send instructions to unload computing tasks to the other satellites; Satellite computing module: used for making a plurality of computing satellites form a satellite computing cluster and determining the shortest computing time of the remote sensing satellite according to the instruction of unloading computing tasks; The delay determination module is used to use the Lagrange duality method and the Karush-Kuhn-Tucker condition to calculate and derive a feasible solution for optimal resource allocation of satellite communication and computing resources within a satellite cluster with the shortest computing time of the delay, and determine the delay of the satellite computing cluster according to the feasible solution; Delay minimization module: used to minimize the sum of delays of each satellite computing cluster by setting a reward function, and repeatedly execute the satellite information acquisition module, instruction issuing module, satellite computing module, and delay determination module in sequence until the algorithm converges, obtains the optimal solution for optimal resource allocation, and determines the optimal resource allocation plan based on the optimal solution.

7. The delay minimization resource allocation system in a low-orbit remote sensing satellite network according to claim 6, characterized in that: The satellite information acquisition module is used to: The location information includes: the latitude and longitude coordinates of the satellite, the Cartesian coordinates of the satellite, and the adjacent satellites in the same orbit and adjacent orbits of each satellite; The communication resource information includes: the task allocation ratio of satellites to transmit information to other satellites, the power of transmitters and the transmission rate of tasks; The computing resource information includes: the computing resources of the CPU allocated by each computing satellite to each subtask of the low-orbit remote sensing satellite.

8. The delay minimization resource allocation system in a low-orbit remote sensing satellite network according to claim 6, characterized in that: The instruction issuing module is also used for: The parameters of the satellite constellation are determined based on the neural network model, including: the total number of satellites, orbital plane, and orbital inclination; and the orbital parameters of each satellite are determined based on the orbital characteristics of the satellite constellation, including: the satellite's position, coverage, and communication link.

9. The delay minimization resource allocation system in a low-orbit remote sensing satellite network according to claim 6, characterized in that: The instruction issuing module is also used to perform the following steps: Step S21: define hyperparameters, including: number of multi-agents, state space dimension, action space dimension, maximum number of steps, maximum number of epochs, discount factor, learning rate, batch size, number of parallel computing cores, clipping parameters and action entropy coefficient; Step S22: Initialize the policy network of the network model, including the number of network layers, input and output dimensions, and masks; Step S23: Initialize the experience replay pool of the network model to store the state, action, reward and new state of the agent during the training process; Step S24: creating multiple independent environments, each environment corresponding to a low-orbit remote sensing satellite; Step S25: Using the satellite's observation data as input, each low-orbit remote sensing satellite generates an action according to the current strategy; Step S26: Execute the action, obtain the reward and the new state, standardize the reward to reduce the variance in the gradient calculation, improve the stability of the training, and store the state, action, reward and new state into the experience replay pool; Step S27: Calculate the loss and update the policy network, using the loss function and policy update method in the PPO algorithm; Step S28: Based on the calculated loss, use the Adam optimizer to update the parameters of the policy network; Step S29: storing the data of each low-orbit remote sensing satellite during the training process into the experience playback pool for subsequent batch processing and strategy update; Step S30: Repeat steps S25-S29 until the network model traversal times reaches the maximum epoch number.

10. The delay minimization resource allocation system in a low-orbit remote sensing satellite network according to claim 6, characterized in that: In the satellite computing module: The shortest computing time of the remote sensing satellite is: when the unloading decision of the remote sensing satellite is determined, when the computing time of all computing satellites in the satellite cluster is the same, the computing time is the shortest.

Images