CN-121979627-A - Remote sensing satellite constellation parallel simulation scheduling method and system based on container cloud

Abstract

The invention discloses a remote sensing satellite constellation parallel simulation scheduling method and system based on container cloud, comprising the steps of carrying out large sample design on remote sensing satellite constellation simulation to generate a large sample design scene; the method comprises the steps of distributing different large sample expected scenes into different containers, scheduling operation, recording operation indexes, performing experience evaluation according to the operation indexes, determining initial weight scores to obtain operation evaluation values of different scene expected information in different containers, building a BP neural network, training the BP neural network to obtain a trained BP neural network, optimizing the weight scores of the operation evaluation values by using the trained BP neural network to obtain a large sample container scheduling model, evaluating the generated samples by using the large sample container scheduling model to obtain an adaptive container selection scheme, and performing scheduling execution according to the container selection scheme to obtain simulation data.

Inventors

• LI JIN
• HU QIYUAN
• JIAO KE
• Lin pengda
• SUN HEJIE
• LI JINDONG
• QIU JIAWEN
• TANG HAITAO
• CAO RUI
• LIU YIXIN
• CUI YONGJUN
• ZHANG WEI
• WU PINGPING
• PENG KUI
• GAO HE
• YU ZHONGJIANG
• YU LONGJIANG
• AN LIANG
• ZHANG XIAO

Assignees

• 中国空间技术研究院

Dates

Publication Date

20260505

Application Date

20251212

Claims (9)

1. 1. The remote sensing satellite constellation large sample parallel simulation scheduling method based on the container cloud is characterized by comprising the following steps of: Carrying out large sample design on remote sensing satellite constellation simulation to generate a large sample design scene; different large sample design scenes are distributed into different containers, the containers are scheduled to run, and running indexes are recorded; According to the operation index, experience evaluation is carried out, initial weight scores are determined, and operation evaluation values of different scene setting information in different containers are obtained; establishing a BP neural network, and training the BP neural network to obtain a trained BP neural network; Optimizing the weight score of the operation evaluation value by using the trained BP neural network to obtain a large sample container scheduling model; evaluating the generated samples by using a large sample container scheduling model to obtain an adaptive container selection scheme; and scheduling and executing according to the container selection scheme to obtain simulation data.
2. 2. The container cloud-based remote sensing satellite constellation large sample parallel simulation scheduling method is characterized in that the samples comprise constellation number, constellation configuration, communication relation and task mode.
3. 3. The container cloud-based remote sensing satellite constellation large sample parallel simulation scheduling method according to claim 1, wherein the operation indexes comprise operation time, maximum acceleration ratio and hardware resource utilization rate.
4. 4. The container cloud-based remote sensing satellite constellation large sample parallel simulation scheduling method according to claim 1, wherein the BP neural network comprises an input layer, a hidden layer and an output layer, wherein, The input layer has n nodes, the hidden layer has q nodes, and the output layer has m nodes; taking the large sample scene vector as a node of an input layer; taking a large sample scene as input; The evaluation values of the indexes including the running time, the maximum speed-up ratio and the utilization of hardware resources are taken as expected output values of an output layer.
5. 5. The container cloud-based remote sensing satellite constellation large sample parallel simulation scheduling method is characterized by comprising the following steps of: n=2i+2 m=i wherein i is the number of nodes of the input layer, m is the number of nodes of the output layer, and a is a mediation constant between 1 and 10.
6. 6. The method for parallel simulation scheduling of remote sensing satellite constellation large samples based on container cloud as claimed in claim 4, wherein the activation function of BP neural network is Wherein x is an input value of the neural network.
7. 7. The method for parallel simulation scheduling of remote sensing satellite constellation large samples based on container cloud as claimed in claim 5, wherein the output of the j-th node of the hidden layer of BP neural network The method comprises the following steps: , wherein, For the input of the j-th node of the hidden layer, Output of the ith node of the input layer; and v ij is the connection weight between the BP neural network input layer and hidden layer nodes, and n is the total number of the input layer nodes.
8. 8. The method for parallel simulation scheduling of remote sensing satellite constellation large samples based on container cloud as claimed in claim 5, wherein the method is characterized in that the output of the g node of the output layer of BP neural network The method comprises the following steps: , wherein, W jg represents the connection weight between the hidden layer of the BP neural network and the node of the output layer; The output of the j node of the hidden layer; And q is the threshold value of the g-th neuron in the BP neural network output layer, and q is the total number of nodes of the hidden layer.
9. 9. A container cloud based remote sensing satellite constellation parallel simulation system, comprising: the design management subsystem is used for carrying out large sample design on remote sensing satellite constellation simulation and generating samples and sending the samples to the simulation data management subsystem; The simulation control management subsystem distributes scene planning information into different containers, schedules operation, records operation indexes, performs weight scores according to the operation indexes to obtain operation evaluation values of the different scene planning information in the different containers, establishes a BP neural network, trains the BP neural network to obtain a trained BP neural network, performs weight optimization on the operation evaluation values by using the trained BP neural network to obtain a large sample container scheduling model, evaluates the generated samples by using the large sample container scheduling model to obtain an adaptive container selection scheme, performs scheduling execution according to the container selection scheme to obtain simulation data, and sends the simulation data to the simulation data management subsystem; and the simulation data management subsystem is used for storing and distributing the samples and the simulation data.

Description

Remote sensing satellite constellation parallel simulation scheduling method and system based on container cloud Technical Field The invention relates to a remote sensing satellite constellation parallel simulation scheduling method and system based on a container cloud, in particular to a remote sensing satellite constellation large sample parallel simulation scheduling method and system based on a container cloud, and belongs to the technical field of simulation. Background The current domestic and foreign remote sensing satellite constellation simulation system is mainly based on a ground digital simulation platform, and a single simulation result, such as information of an observation task chain, a time chain, an accuracy chain and the like, is obtained by manually editing a desired scene (parameter setting) operation simulation flow. Such systems typically employ serial simulation flows that can only be performed one by one for combined scenarios of elements such as different satellite orbits, different detectors, different simulation times, etc. When verifying the performance boundary of a constellation system, simulation verification of a large number of scenes is required, and the existing system cannot efficiently execute simulation tasks under the condition of generating wanted data in a large number in an automatic manner, so that the simulation tasks have long execution period and low resource utilization rate, and the quantitative data requirements of the shape shaping of subsequent equipment are difficult to effectively support. Meanwhile, the system lacks an effective parallel processing mechanism, and abnormality caused by resource contention is easy to occur in the task execution process, so that the overall simulation efficiency and reliability are affected. Specifically, in the prior art, systems such as SATELLITE MISSION PLANNING SYSTEM (SMPS) and SATELLITE MISSION PLANNING TOOL (SMPT) have a core architecture based on a conventional serial processing mode, although they have a certain degree of maturity in terms of single scene simulation. In addition, the system lacks effective containerization and distributed processing capacity and cannot fully utilize modern cloud computing resources, so that the response time of the system is obviously increased in a high concurrency scene. These defects severely restrict the efficiency and accuracy of remote sensing satellite constellation system efficiency evaluation, and are difficult to meet the requirements of current satellite constellation rapid iteration and complex scene verification. In order to effectively verify the performance boundary of a constellation system, quantitative data support is provided for shaping the following equipment form, parallel large-sample simulation verification is required for a typical scene, the simulation scene is supposed to comprise various factors such as different satellite orbits, different detectors, different simulation times and the like, the coverage range is wide, the data volume is huge, simulation tasks are required to be executed in parallel, and a solution for improving the task execution efficiency is provided. Disclosure of Invention In order to solve the technical problems, the invention provides a remote sensing satellite constellation large sample parallel simulation scheduling method and system based on a container cloud, which realize intelligent container allocation of remote sensing satellite constellation simulation tasks. The technical scheme of the invention is as follows: The invention discloses a remote sensing satellite constellation large sample parallel simulation scheduling method based on container cloud, which comprises the following steps: Carrying out large sample design on remote sensing satellite constellation simulation to generate a large sample design scene; different large sample design scenes are distributed into different containers, the containers are scheduled to run, and running indexes are recorded; According to the operation index, experience evaluation is carried out, initial weight scores are determined, and operation evaluation values of different scene setting information in different containers are obtained; establishing a BP neural network, and training the BP neural network to obtain a trained BP neural network; Optimizing the weight score of the operation evaluation value by using the trained BP neural network to obtain a large sample container scheduling model; evaluating the generated samples by using a large sample container scheduling model to obtain an adaptive container selection scheme; and scheduling and executing according to the container selection scheme to obtain simulation data. Further, in the method, the samples comprise constellation number, constellation configuration, communication relation and task mode. Further, in the above method, the operation indexes include operation time, maximum speed-up ratio, and hardware resource utilizati