CN116068990A - Star group intelligent fault diagnosis interactive virtual simulation platform verification method - Google Patents
Star group intelligent fault diagnosis interactive virtual simulation platform verification method Download PDFInfo
- Publication number
- CN116068990A CN116068990A CN202211637144.9A CN202211637144A CN116068990A CN 116068990 A CN116068990 A CN 116068990A CN 202211637144 A CN202211637144 A CN 202211637144A CN 116068990 A CN116068990 A CN 116068990A
- Authority
- CN
- China
- Prior art keywords
- fault diagnosis
- satellite
- star
- simulation
- real
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 179
- 238000000034 method Methods 0.000 title claims abstract description 97
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 74
- 238000012795 verification Methods 0.000 title claims abstract description 64
- 238000004088 simulation Methods 0.000 claims abstract description 95
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 32
- 230000006870 function Effects 0.000 claims abstract description 30
- 238000005516 engineering process Methods 0.000 claims abstract description 22
- 230000003993 interaction Effects 0.000 claims abstract description 22
- 238000013461 design Methods 0.000 claims abstract description 18
- 230000004927 fusion Effects 0.000 claims abstract description 16
- 238000011161 development Methods 0.000 claims abstract description 7
- 238000013135 deep learning Methods 0.000 claims abstract description 4
- 239000003795 chemical substances by application Substances 0.000 claims description 82
- 230000008569 process Effects 0.000 claims description 39
- 238000012549 training Methods 0.000 claims description 34
- 210000004556 brain Anatomy 0.000 claims description 26
- 230000000007 visual effect Effects 0.000 claims description 24
- 238000004891 communication Methods 0.000 claims description 21
- 230000002787 reinforcement Effects 0.000 claims description 17
- 238000002347 injection Methods 0.000 claims description 13
- 239000007924 injection Substances 0.000 claims description 13
- 230000015572 biosynthetic process Effects 0.000 claims description 11
- 238000004519 manufacturing process Methods 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 11
- 230000035945 sensitivity Effects 0.000 claims description 11
- 230000005540 biological transmission Effects 0.000 claims description 10
- 238000005286 illumination Methods 0.000 claims description 10
- 230000007613 environmental effect Effects 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000004886 process control Methods 0.000 claims description 4
- 230000010354 integration Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000000712 assembly Effects 0.000 claims description 2
- 238000000429 assembly Methods 0.000 claims description 2
- 238000009826 distribution Methods 0.000 claims description 2
- 238000012800 visualization Methods 0.000 abstract description 2
- 230000008901 benefit Effects 0.000 description 15
- 238000010586 diagram Methods 0.000 description 12
- 238000005755 formation reaction Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 6
- 230000036541 health Effects 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 240000000543 Pentas lanceolata Species 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000003550 marker Substances 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000009877 rendering Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000012631 diagnostic technique Methods 0.000 description 1
- 238000005538 encapsulation Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0208—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the configuration of the monitoring system
- G05B23/0213—Modular or universal configuration of the monitoring system, e.g. monitoring system having modules that may be combined to build monitoring program; monitoring system that can be applied to legacy systems; adaptable monitoring system; using different communication protocols
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Radio Relay Systems (AREA)
Abstract
The invention discloses a verification method of an intelligent fault diagnosis interactive virtual simulation platform of a star group, which comprises the following steps: s1, designing a star intelligent fault diagnosis technology based on multi-information fusion: the fault diagnosis technology is online fault diagnosis based on a Python deep learning algorithm; s2, designing and realizing a real-time simulation framework of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group: the real-time interactive virtual simulation verification platform utilizes a Unity3D design, and the real-time simulation framework comprises a main control computer, a real-time data interface and a view computer; s3, designing and realizing star intelligent fault diagnosis interactive virtual simulation verification platform vision software: the view software is based on Unity3D development, a real satellite three-dimensional model is built by adopting 3D Max, space environment building is completed by utilizing Skybox components, and the view software comprises star group parameter configuration, man-machine interaction functions and fault diagnosis result software. By adopting the steps, the invention realizes real-time display of fault results and three-dimensional visualization of running states.
Description
Technical Field
The invention relates to the fields of satellite group fault diagnosis, virtual simulation verification, real-time simulation interaction and the like, in particular to a satellite group intelligent fault diagnosis interactive virtual simulation platform verification method.
Background
The small satellite has the advantages of low cost, small volume, high flexibility and the like, and can replace a large spacecraft to complete tasks in a formation flying mode. Satellite groups have become a hot spot for research at home and abroad due to the advantages of low system cost, good system performance, high reliability, strong adaptability and the like. However, as the number of the involved satellites is large, the decoupling relationship is complex, and the operation environment is special, the fault diagnosis of the satellite group is more difficult; in addition, when a certain satellite in the satellite group breaks down, if diagnosis and processing are not performed in time, the loss is more serious, and not only one satellite is likely to be involved, so that the method has important practical significance for carrying out fault diagnosis research on the satellite group. In the single star level fault diagnosis process, if the satellite has the problems of underallocation, incomplete measurement information and the like, accurate fault diagnosis is difficult to realize, and at the moment, the lead satellite is required to carry out multi-information fusion on star level intelligent fault diagnosis. Besides the self position, the star sensitivity and the actuating mechanism information, the measurement information such as the relative position, the star sensitivity and the like of the neighbor satellites are further considered, and the fault diagnosis method based on the centralized multi-agent reinforcement learning is different from the single-star fault diagnosis method, the multi-agent reinforcement learning method can simultaneously consider global information, is more comprehensive than single-star local information, and the number change of the star satellites does not influence the use of a network. Therefore, in order to improve the real-time performance and accuracy of the self-healing process of the spacecraft control system, and provide an effective way for prolonging the service life of the on-orbit spacecraft, it is necessary to develop a multi-information fusion fault diagnosis technology applied to the constellation formation.
In addition, in order to verify the effectiveness of diagnostic techniques, a set of effective simulation verification techniques are required. The traditional digital simulation technology generally adopts a simulation module provided by MATLAB/Simulink to realize simulation verification of the whole physical model and the diagnosis technology. However, simulink can only perform offline simulation, and cannot perform real-time simulation interaction with an algorithm end, so that a worker cannot timely judge the on-orbit running state and fault condition of the satellite. Based on the traditional digital simulation technology, the virtual visual simulation technology outputs digital simulation data in a graphical form, and can perform real-time data transmission work with an algorithm end, so that the virtual visual simulation technology is stronger in real-time performance and higher in readability, and is popular. The Unity3D is used as the dominant software of the free open source game engine to gradually become the virtual visual simulation, so that a plurality of complex problems in the development of the virtual visual simulation scene are solved, and the unique UGUI system provides convenience for human-computer interaction, so that the method is suitable for building the fault diagnosis technology interactive virtual simulation verification platform.
Therefore, it is necessary to provide a verification method for a star intelligent fault diagnosis interactive virtual simulation platform to solve the above problems.
Disclosure of Invention
The invention aims to provide a verification method of an intelligent fault diagnosis interactive virtual simulation platform for a constellation, which realizes intelligent fault diagnosis of the constellation by taking measurement information such as relative positions and star sensitivity among adjacent satellites into consideration, develops the interactive virtual simulation verification platform and realizes visual demonstration of interactive fault diagnosis and constellation configuration reconstruction process.
In order to achieve the above purpose, the invention provides a verification method of an intelligent fault diagnosis interactive virtual simulation platform for a star group, which comprises the following steps:
s1, designing a star intelligent fault diagnosis technology based on multi-information fusion: the fault diagnosis technology is online fault diagnosis based on a Python deep learning algorithm;
s2, designing and realizing a real-time simulation framework of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group: the real-time interactive virtual simulation verification platform utilizes a Unity3D design, and the real-time simulation architecture of the star intelligent fault diagnosis interactive virtual simulation verification platform comprises a real-time simulation architecture of a main control computer, a real-time data interface and a vision computer;
s3, designing and realizing star intelligent fault diagnosis interactive virtual simulation verification platform vision software: the view software is based on Unity3D development, a real satellite three-dimensional model is built by adopting 3D Max, space environment building is completed by utilizing Skybox components, and the view software comprises star group parameter configuration, man-machine interaction functions and fault diagnosis result software.
Preferably, in the step S1, the fault diagnosis technology based on multiple information uses the position and the star sensitivity and mechanism information of the Leader satellite, and then performs global information integration by combining the relative position and the star sensitivity information of the neighboring satellites, and performs decoupling processing on the whole data information.
Preferably, the intelligent fault diagnosis of the constellation level with multiple information fusion is specifically as follows:
(1) Establishing a random game model oriented to fault diagnosis: summarizing the fault diagnosis problem of the satellite group level satellite into a random game model;
(2) Offline network training: the centralized multi-agent reinforcement learning task decision method algorithm has N satellite data as network input, and comprises a diagnosis result of the satellite to the satellite and position and speed information of the satellite, wherein the network is composed of a full-connection layer and a Bi-LSTM layer, so that the N satellite data are subjected to information communication, decision is made based on global information, meanwhile, the fault diagnosis network is not required to be changed and learned again when the number of satellites in formation changes, the Bi-LSTM-based multi-agent network decouples the data, and finally, the fault diagnosis condition of the N satellites is output after the whole information is fused;
(3) On-line fault diagnosis: after the off-line training process is completed, network parameters of an evaluation network and a decision network are determined, a central decision structure is adopted for a track planning task, and multi-satellite real-time fault diagnosis decision is finally realized in a real task environment with uncertain environment influence.
Preferably, in the step S2, the design and implementation of the real-time simulation architecture of the intelligent fault diagnosis interactive virtual simulation verification platform for the star group transmits the simulation data calculated by the main control computer end to the visual computer end for visual simulation verification, and an interactive interface with controllable data transmission time sequence is required between the algorithm software Python and the visual software Unity.
Preferably, the real-time simulation architecture of the main control computer, the real-time data interface and the vision computer is communicated through the ML-Agents real-time data interface, and the Socket is used for communication between the vision computer and the main control computer in the model training stage.
Preferably, the design and implementation steps of the real-time simulation architecture of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group are as follows:
(1) ML-Agents real-time simulation architecture: the relationship of the three high-level components of the ML-Agents framework is: the learning environment communicates with the Python API through External Communicator, and the communication function of ML-Agents is based on a client/server architecture, and Socket sockets are used for communicating a Unity process and a Python process in a model training stage;
(2) Building a real-time training environment: building a training environment, namely expanding three assemblies around the Academy, the Brain and the Agent, and building a Brain assembly under the Academy;
In the Academy component, configuring an IP address and a port number of an algorithm software end;
the Brain is downwards responsible for distributing Agent control quantity, and upwards responsible for summarizing the environment state to the algorithm end, and according to specific tasks, the Brain is required to appoint the dimension of the environment state and the dimension of the control quantity;
binding an Agent component for each Agent in a scene, wherein the Agent component is responsible for acquiring specific environmental data and executing control quantity of Brain distribution;
in the built virtual simulation environment, the configuration of three components of ML-Agents and the realization of scene logic are completed, a real-time simulation environment required by reinforcement learning is built, and the maximum waiting time T of a timer is set in a program wait If no data is received beyond the time, the TCP connection is disconnected.
Preferably, the design and implementation of the star intelligent fault diagnosis interactive virtual simulation verification platform vision software are implemented, the construction work of the star working environment is carried out based on the Unity3D physical engine, the satellite group gesture and data change simulation data transmitted by the main control computer are received, the earth and satellite virtual simulation model in the three-dimensional scene is driven, the demonstration of the satellite group operation environment and the gesture adjustment process is carried out, and the three-dimensional virtual scene is divided into three parts, namely a scene resource module, a graphical interface module and a vision demonstration module.
Preferably, the design and implementation specific three-dimensional virtual simulation scene of the star intelligent fault diagnosis interactive virtual simulation verification platform vision software is as follows:
(1) The scene resource module relates to the simulation of space scene resources and can be concretely divided into the production of satellite models, the production of earth models, the realization of starry sky backgrounds and the simulation of solar illumination; 3dmax is used for manufacturing the satellite and the earth model, and then the built models such as the earth, the satellite and the like are imported into the Unity 3D; the sky background is realized through a sky box in Unity 3D; the sky box is a panoramic view and is composed of upper, lower, left, right, front, back and pictures along the main axis direction, and the six pictures are combined into a sphere;
(2) And a graphical interface module: the text box and button part is developed by adopting UGGUI and TextMeshPro components built in the Unity3D, and the chart part is realized by adopting Xcharts components;
(3) A view demonstration module: after the simulation is started, the vision software enters an interface of the star group when the star group normally works in orbit, and the test personnel checks the star group configuration, starts the simulation or exits the simulation platform function by clicking any satellite to check the state information.
Preferably, the graphical interface is divided into three parts, namely a satellite state information panel, an injection fault panel and a fault diagnosis result panel.
Preferably, the man-machine interaction function comprises simulation process control, satellite state display and fault injection panel function.
Therefore, the invention provides a verification method of an intelligent fault diagnosis interactive virtual simulation platform for a star group, which has the following beneficial effects:
(1) The intelligent fault diagnosis of the star group and the interactive virtual simulation verification platform are designed and realized, and the functions of data monitoring, real-time simulation, visual display and the like are integrated into a whole, so that the intelligent fault diagnosis algorithm of the star group is used for verifying the effectiveness of the intelligent fault diagnosis algorithm of the star group.
(2) The intelligent fault diagnosis of the star group and the design and implementation of the interactive virtual simulation verification platform are realized, the architecture of the main control computer-vision computer is adopted to build real-time simulation love, the real-time requirement is met, and the cost is lower.
(3) The intelligent fault diagnosis of the star group and the design and realization of the interactive virtual simulation verification platform are realized, the interactive virtual simulation platform is built by adopting the Unity3D, the running state of the satellite is subjected to three-dimensional visual display through the animation effect, and the complex chart data analysis in the past is eliminated.
(4) The intelligent fault diagnosis of the star group and the design and implementation of the interactive virtual simulation verification platform develop a real-time simulation interaction function through an autonomous design UI interface, a tester can randomly select a satellite to check the on-orbit running state through the UI interface, randomly select a fault satellite to inject various fault types at any time, interact with an intelligent fault diagnosis algorithm, and can display the fault diagnosis result in real time.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
FIG. 1 is a general structure diagram of a simulation verification platform of a star intelligent fault diagnosis interactive virtual simulation platform verification method of the invention;
FIG. 2 is a diagram of a satellite group fault diagnosis training network for a satellite group intelligent fault diagnosis interactive virtual simulation platform verification method of the invention;
FIG. 3 is a real-time simulation structure diagram of a satellite constellation ML-Agent of the intelligent fault diagnosis interactive virtual simulation platform verification method of the constellation;
FIG. 4 is a logical diagram of ML-Agent code block call of a star intelligent fault diagnosis interactive virtual simulation platform verification method;
FIG. 5 is a diagram of a satellite constellation view demonstration architecture of a constellation intelligent fault diagnosis interactive virtual simulation platform verification method of the present invention;
FIG. 6 is a functional diagram of a visual software of a verification method of an intelligent fault diagnosis interactive virtual simulation platform for a star group;
FIG. 7 is a diagram of a simulation platform start interface of a method for verifying a star intelligent fault diagnosis interactive virtual simulation platform according to the present invention;
FIG. 8 is a diagram of a star fault diagnosis interface of a verification method of a star intelligent fault diagnosis interactive virtual simulation platform;
Fig. 9 is a diagram of a star formation configuration reconstruction interface of a star intelligent fault diagnosis interactive virtual simulation platform verification method.
Detailed Description
The technical scheme of the invention is further described below through the attached drawings and the embodiments.
Unless defined otherwise, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this invention belongs. The terms "first," "second," and the like, as used herein, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "upper", "lower", "left", "right", etc. are used merely to indicate relative positional relationships, which may also be changed when the absolute position of the object to be described is changed.
Examples
In this embodiment, the invention provides a verification method for a star intelligent fault diagnosis interactive virtual simulation platform, which comprises the following steps:
s1, designing a star intelligent fault diagnosis technology based on multi-information fusion: the fault diagnosis technology is online fault diagnosis based on a Python deep learning algorithm;
in the step S1, the fault diagnosis technology based on multiple information utilizes the position and the star sensitivity and mechanism information of the Leader satellite, and then carries out global information integration by combining the relative position and the star sensitivity information of the neighbor satellite, and carries out decoupling processing on the whole data information.
The intelligent fault diagnosis of the star group level of the multi-information fusion is specifically as follows:
(1) Establishing a random game model oriented to fault diagnosis: summarizing the fault diagnosis problem of the satellite group level satellite into a random game model;
(2) Offline network training: the centralized multi-agent reinforcement learning task decision method algorithm has N satellite data as network input, and comprises a diagnosis result of the satellite to the satellite and position and speed information of the satellite, wherein the network is composed of a full-connection layer and a Bi-LSTM layer, so that the N satellite data are subjected to information communication, decision is made based on global information, meanwhile, the fault diagnosis network is not required to be changed and learned again when the number of satellites in formation changes, the Bi-LSTM-based multi-agent network decouples the data, and finally, the fault diagnosis condition of the N satellites is output after the whole information is fused;
(3) On-line fault diagnosis: after the off-line training process is completed, network parameters of an evaluation network and a decision network are determined, a central decision structure is adopted for a track planning task, and multi-satellite real-time fault diagnosis decision is finally realized in a real task environment with uncertain environment influence.
S2, designing and realizing a real-time simulation framework of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group: the real-time interactive virtual simulation verification platform utilizes a Unity3D design, and the real-time simulation architecture of the star intelligent fault diagnosis interactive virtual simulation verification platform comprises a real-time simulation architecture of a main control computer, a real-time data interface and a vision computer; the man-machine interaction function comprises simulation process control, satellite state display and fault injection panel function.
In step S2, the design and implementation of the real-time simulation architecture of the star intelligent fault diagnosis interactive virtual simulation verification platform needs to transmit the simulation data calculated by the main control computer end to the visual computer end for visual simulation verification, and an interactive interface with controllable data transmission time sequence is needed between the algorithm software Python and the visual software Unity.
The real-time simulation architecture of the main control computer, the real-time data interface and the vision computer is communicated through the ML-Agents real-time data interface, and the Socket is used for communication between the vision computer and the main control computer in the model training stage.
The design and implementation steps of the real-time simulation architecture of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group are as follows:
(1) ML-Agents real-time simulation architecture: the relationship of the three high-level components of the ML-Agents framework is: the learning environment communicates with the Python API through External Communicator, and the communication function of ML-Agents is based on a client/server architecture, and Socket sockets are used for communicating a Unity process and a Python process in a model training stage;
(2) Building a real-time training environment: the training environment is built around the three components of Academy, brain and Agent described above. Creating a Brain component under Academy;
in the Academy component, configuring an IP address and a port number of an algorithm software end;
the Brain is downwards responsible for distributing Agent control quantity, and upwards responsible for summarizing environmental states to an algorithm end. According to specific tasks, the dimension of the environmental state and the dimension of the control quantity are required to be specified in the Brain;
an Agent component is bound for each Agent in the scene. The Agent component is responsible for acquiring specific environmental data and executing control quantity distributed by Brain;
in the built virtual simulation environment, the configuration of the three ML-Agents components and the realization of scene logic are completed, and the real-time simulation environment required by reinforcement learning can be built efficiently. Setting a timer in the program ensures the stability of the communication, i.e. setting a maximum waiting time T wait If the data is not received in the past, the TCP connection is disconnected, and resource waste is avoided.
S3, designing and realizing star intelligent fault diagnosis interactive virtual simulation verification platform vision software: the view software is based on Unity3D development, a real satellite three-dimensional model is built by adopting 3D Max, space environment building is completed by utilizing Skybox components, and the view software comprises star group parameter configuration, man-machine interaction functions and fault diagnosis result software.
The intelligent satellite group fault diagnosis interactive virtual simulation verification platform vision software is designed and realized, the construction work of a satellite group working environment is carried out based on a Unity3D physical engine, the satellite group gesture and data change simulation data transmitted by a main control computer are received, an earth and satellite virtual simulation model in a three-dimensional scene is driven, the demonstration of the satellite group operation environment and the gesture adjustment process is carried out, and the three-dimensional virtual scene is divided into a scene resource module, a graphic interface module and a vision demonstration module.
The specific three-dimensional virtual simulation scene of the design and implementation of the star intelligent fault diagnosis interactive virtual simulation verification platform vision software is as follows:
(1) The scene resource module mainly relates to the simulation of space scene resources, and can be concretely divided into the production of satellite models, the production of earth models, the realization of starry sky backgrounds and the simulation of solar illumination; 3dmax is used for manufacturing the satellite and the earth model, and then the built models such as the earth, the satellite and the like are imported into the Unity 3D; the sky background is realized through a sky box in Unity 3D; the space box is a panoramic view and is composed of upper, lower, left, right, front, back and front maps along the main axis direction, and the six maps are combined into a sphere. The Unity3D has a strong illumination system, and the global illumination system realizes the simulation of solar illumination by simulating direct light and indirect light
(2) And a graphical interface module: the text box and button part is developed by adopting UGGUI and TextMeshPro components built in the Unity 3D, and the chart part is realized by adopting Xcharts components; the graphic interface is divided into three parts of a satellite state information panel, an injection fault panel and a fault diagnosis result panel.
(3) A view demonstration module: the experimenter can perform man-machine interaction with the virtual simulation platform, after simulation is started, the vision software enters an interface of the star group when the star group normally works in orbit, and the experimenter checks the star group configuration, starts simulation or exits the simulation platform function by clicking any satellite to check state information.
The multi-agent reinforcement learning method is different from a single-star level fault diagnosis method, global information can be considered at the same time, the multi-agent reinforcement learning method is more comprehensive than single-star local information, and the number change of the satellite group satellites does not influence the use of a network; in addition, the input data of the single-agent network is too much, the decoupling performance is too strong, the learning effect is poor, and the data can be decoupled by the multi-agent network.
The health state diagnosis result of the Leader satellite on each satellite is given out based on the multi-agent reinforcement learning method, the fault type is given out, the satellite group level intelligent fault diagnosis with multi-information fusion is realized, and the structure diagram is shown in the figure.
The real-time simulation architecture of the star fault diagnosis interactive virtual simulation verification platform comprises a real-time simulation architecture of a main control computer, a real-time data interface and a vision computer. The real-time data interface communicates through the ML-Agents interface. The communication function of ML-Agents is based on a client/server architecture. Communication between the vision computer and the main control computer is carried out in the simulation stage by using Socket sockets. The vision software comprises star group parameter configuration, man-machine interaction function and fault diagnosis result software. The man-machine interaction function comprises simulation process control, satellite state display and fault injection panel function.
The vision computer builds a real star model and a space environment, and sends information such as the position, the gesture and the like of the star to the main control computer through the real-time data interface. The main control computer carries out fault diagnosis on the satellite group in the view computer based on the satellite kinematics and dynamic model and carries out intelligent fault diagnosis algorithm on the satellite group, and the fault diagnosis result is sent to the view computer for display through the real-time data interface. In order to solve the problem of real-time data interaction between the main control computer and the vision computer, the invention further provides an ML-Agents real-time data interface for communication. The communication function of ML-Agents is based on a client/server architecture, and Socket sockets are used for communication between a view computer and a main control computer in a model training stage. The real-time simulation architecture designed by the invention can realize real-time data interaction between the fault diagnosis algorithm and the virtual visual simulation.
And establishing a real satellite three-dimensional model by adopting 3D Max based on the Unity3D development by using the constellation fault diagnosis room interactive virtual simulation verification platform vision software, and completing space environment establishment by using a Skybox assembly.
The experimenter can click any satellite to check the state information such as the position, the gesture, the speed and the like, and meanwhile, the virtual visual simulation software can perform real-time data interaction with the intelligent fault diagnosis algorithm, and the satellite health condition is reflected on the right fault diagnosis result panel. In addition, the testers can select any satellite at any time and inject faults so as to simulate the state of the satellite when the satellite breaks down, the fault diagnosis algorithm can immediately diagnose the faults according to the received satellite group state information by further relying on interconnection of the vision software and the intelligent fault diagnosis algorithm, and the diagnosis result is sent to the vision software in real time for display, so that the visual demonstration of the interactive satellite group fault diagnosis flow is finally realized.
Firstly, designing a star group intelligent fault diagnosis technology based on multi-information fusion.
The invention provides a multi-information fusion star group intelligent fault diagnosis technology based on the health state of the satellite and the health state of the neighbor satellite, realizes the online real-time autonomous diagnosis of fault information, ensures the safety and reliability of the spacecraft and the smooth proceeding of the aerospace task, and provides an effective way for prolonging the service life of the on-orbit spacecraft.
1. Multi-information fusion star group level intelligent fault diagnosis method
In the single star level fault diagnosis process, if the satellite has the problems of underallocation, incomplete measurement information and the like, accurate fault diagnosis is difficult to realize, and at the moment, the lead satellite is required to carry out multi-information fusion on star level intelligent fault diagnosis. Besides the self position, the star sensitivity and the actuating mechanism information, the measurement information such as the relative position, the star sensitivity and the like of the neighbor satellites are further considered, and the fault diagnosis method based on the centralized multi-agent reinforcement learning is different from a single star level fault diagnosis method, the multi-agent reinforcement learning method can simultaneously consider global information, is more comprehensive than single star local information, and the number change of the star satellites does not influence the use of a network; in addition, the input data of the single-agent network is too much, the coupling is too strong, the learning effect is poor, and the data can be decoupled by the multi-agent network.
The health state diagnosis result of the Leader satellite on each satellite is given out based on the multi-agent reinforcement learning method, the fault type is given out, the satellite group level intelligent fault diagnosis with multi-information fusion is realized, and the structure diagram is shown in the figure.
(1) Random game model establishment for fault diagnosis
The environment of Multi-agent reinforcement learning (MARL) is a random gaming framework based on a markov decision process. The fault diagnosis problem of the satellite group level satellite is summarized into a random game model, and the meanings of all symbols are as follows:
1) State set S
For the ith satellite, i ε [1, N]The position p of the self at the time t is calculated i Velocity v i Attitude angle theta i Angular velocity w i Output information f of executing mechanism of flywheel i And corresponding observations x for neighbors i As state quantity s, i.e.
s＝(p i ,v i ,θ i ,w i ,f i ,x i )∈S (1)
2) Observation state set O
For the ith satellite, i ε [1, N]The position p of the self at the time t is calculated i Velocity v i Attitude angle theta i Angular velocity w i Output information f of executing mechanism of flywheel i And the corresponding observations x of the neighbors i As the state quantity o,
o＝(p i ,v i ,θ i ,w i ,f i ,x i )∈O (2)
3) Action set A
And diagnosing which part of the satellite has a fault as action a.
A＝{a 1 ,a 2 ...a N } (3)
a i Is the judgment of whether the satellite i has faults or not, a i =0 is no fault, a i =1 is baseline drift failure, a i =2 is a star-sensitive fault, a i =3 is the failure of flywheel 1, a i =4 is the failure of the flywheel 2,a i =5 is the failure of flywheel 3, a i =6 is the failure of flywheel 4, a i Failure of =7 gyro 1, a i Failure of =8 gyro 2, a i Failure of =9 gyro 3, a i Fault of gyro 4 =10.
4) Immediate benefit value
And accurately taking a satellite fault diagnosis result as a final target, establishing a reward and punishment mechanism, and determining a benefit value obtained by single-step diagnosis. Assuming that each satellite receives the same value of benefit, i.e
For each satellite, gain value +.>
The main steps are as follows:
for single step diagnostics, the observed benefit is obtained when there is a match with the marker information, and the benefit is 0 when there is no match with the marker information.
5) Discount factor gamma
Gamma represents the importance of the future benefit value relative to the current benefit value. When γ=0, which is equivalent to considering only the current benefit and not considering the future benefit, γ=1, the future benefit and the current benefit are considered to be equally important.
By definition of the random game model symbols, the satellite fault diagnosis process can be described as: when each satellite is in the environment, the actual position, speed, attitude angle and angular speedWhen the output information of the executing mechanism of the flywheel and the information such as the corresponding observed value of the neighbor form a state s (t), the satellite obtains the position, the speed and the attitude angle of the satellite according to the sensor, the output information of the executing mechanism of the wheel and the observed information o such as the observed state of the neighbor on the satellite i (t) outputting whether the fault is the type a of the fault i And obtaining a corresponding immediate benefit value according to equation (4)
This process is repeated until the goal of forming the accuracy of the desired fault diagnosis is reached.
(2) Offline network training
The centralized multi-agent reinforcement learning task decision method algorithm (Bicnet) takes data of N satellites as network input, and comprises diagnostic results of the satellites on the satellites and information such as the positions, the speeds and the like of the satellites. The network is composed of a full-connection layer and a Bi-LSTM layer, so that data of N satellites can be communicated with each other, decisions can be made based on global information, and meanwhile, the fault diagnosis network does not need to be changed and learned again when the number of satellites in formation changes; in addition, the input data of the single agent network is too much, the coupling is too strong, the learning effect is poor, and the multi-agent network based on Bi-LSTM can be used for decoupling the data, so that the learning speed is improved. Based on Bi-LSTM network, the fault diagnosis condition of N satellites after the fusion of the whole information is finally output is represented by 0-10. 0 represents no fault of the satellite, 1 represents baseline drift fault of the satellite, 2 represents satellite-sensitive fault of the satellite, 3 represents flywheel 1 fault of the satellite, 4 represents flywheel 2 fault of the satellite, 5 represents flywheel 3 fault of the satellite, 6 represents flywheel 4 fault of the satellite, 7 represents gyro 1 fault of the satellite, 8 represents gyro 2 fault of the satellite, 9 represents gyro 3 fault of the satellite, and 10 represents gyro 4 fault of the satellite.
And judging all the fault information and the sensor information of the satellite input by the network, outputting the fault information and the sensor information as an overall judging Q value, judging the quality of the satellite fault diagnosis result by using the overall Q value, and guiding the decision network to carry out gradient update towards the maximum profit direction.
The complete training process of the fault diagnosis network comprises a data acquisition process and a parameter training process. Two training procedures will be described below.
1) Data acquisition process
In the data acquisition process, each satellite takes the position, speed, attitude angle, angular speed of the satellite, output information of an executing mechanism of a flywheel and corresponding observed values of neighbors as o, and all observed values o of N satellites 1 ,…,o N Sending the operation a corresponding to the output in the decision network 1 ,…,a N Obtain a return
After the above operations are performed, each iteration round stores data in the following form: />
This process requires a large number of rounds to obtain enough experience vectors to be all in the experience pool. The method of collecting and training the network is adopted in this section, and the updated decision network is used for collecting data, so that the independence of the collected data is ensured.
2) Parameter training process
Taking the data of N satellites into consideration, using theta= [ theta ] 1 ,…,θ N ]Representing a set of satellite decision parameters, using μ= [ μ ] 1 ,…μ N ]Representing a set of satellite decisions. In the process of evaluating the network training, the network updating scheme is shown as a formula (6), and an Adam optimizer is adopted to update the evaluation network weight:
in the method, in the process of the invention,
is a target value generated by the target network, which is defined as follows:
where i subscript denotes the ith satellite and j superscript denotes the jth random sample.
Adopting an Adam optimizer to update the weight of the decision network:
wherein D is an experience storage pool, and the element composition is as follows
(3) On-line fault diagnosis
After the offline training process is completed, the network parameters of the evaluation network and the decision network are also determined. For a track planning task, a main star adopts a central decision structure, and in a real task environment with uncertain influence on the environment, the real-time fault diagnosis decision of multiple satellites is finally realized.
The decision process of the online diagnosis link, algorithm 1, is given below.
Secondly, designing and realizing a real-time simulation framework of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group.
The intelligent fault diagnosis technology of the star group designed in the first step only carries out fault diagnosis on the star group in an algorithm layer, but cannot intuitively reflect the fault type of the star group and the influence generated after the fault, so that simulation data calculated by a main control computer end are required to be transmitted to a visual computer end in real time for visual simulation verification, and an interactive interface with controllable data transmission time sequence is required between algorithm software Python and visual software Unity. The ML-Agents plugin by the Unity3D authorities in 2018 made possible real-time interactions of Python and Unity 3D. Because the new version of ML-Agents has stronger encapsulation, especially the algorithm end has poorer openness, the earliest version of ML-Agents is selected. The ML-Agents structure framework is shown in FIG. 3.
ML-Agents real-time simulation architecture
The relationship of the three high-level components of the ML-Agents framework is: the learning environment (Unity 3D) communicates with the Python API through External Communicator. The following describes the composition of the learning environment and the transmission principle of External Communicator.
The learning environment contains three additional components that can help organize the Unity3D scene: agent component, brain component, academy component. The Agent component may be attached to one Unity Game Object in a one-to-one correspondence with satellites in the scene, responsible for gathering environmental information and enforcing control strategies. The Brain component encapsulates the Agent's decision logic, is responsible for receiving the information that the Agent gathered, and distributes tactics and rewards value to the Agent. Agents with similar behavior actions can share a Brain component, and in order to reduce the coupling degree and the convenience of data processing, a one-to-one correspondence exists between the agents and the Brain. The Academy component is a control center of the whole interaction environment, is responsible for commanding the observation and decision process of the Agent and coordinating data among different Brain, and can specify a plurality of environment parameters, such as rendering quality and environment running speed parameters, and a communication interface with Python is also positioned in the component.
The ML-Agents communication function is based on a Client/Server (C/S) architecture. And in the model training stage, socket sockets are used for communication of the Unity process and the Python process. Wherein the Python end is a Socket Server, and the Unity Environment end is a Socket Client. The transport layer protocols currently mainly include the transmission control protocol (Transmission Control Protocol, TCP) and the user datagram protocol (User Datagram Protocol, UDP). Where the TCP protocol provides reliable, time-sequential data transmission based on byte streams, while the UDP protocol provides transport services based on datagrams, although the services are lighter weight, but less reliable. Because of the requirements of the real-time emulation interface for transmission reliability and data ordering, the ML-Agents transport layer employs the TCP protocol.
Socket is a software abstract layer between application layer and TCP/IP protocol family, adopts design idea of door-to-door mode, conceals complex protocol family from user, programmer can work for development of service logic, and let Socket be responsible for organizing data to meet specified communication protocol. A combination of IP and port (port) can uniquely identify a Socket. Firstly, a server initializes Socket parameters, binds with a designated port (bind), and then calls an accept blocking process to wait for connection of a client. When the client has a service request, the client initializes a Socket, designates a target IP and a port number, and then calls a connect connection server. If the connection is successfully established, the server can process the request of the client, and the two parties begin to perform data interaction. Because a training algorithm trains task scenes in a limited way, usually in a one-to-one relation, the ML-Agents server side processes the request of the client side in a blocking way. That is, the main thread blocks the connection request of the waiting client, when the client is connected, the main thread immediately builds a sub thread for data communication with the connection of the client, and the main thread still waits for a new client connection.
2. Real-time training environment construction
The training environment is built around the three components of Academy, brain and Agent described above. The scene construction steps are as follows:
in the Academy component, an IP address and a port number of an algorithm software end are configured. The rendering level of the virtual scene may also be set here during training, typically with lower levels selected to reduce overhead. All Brain models in the scene are uniformly managed under Academy.
The Brain component is created under Academy based on the number of models run at the algorithm end (typically only one model is training).
The Brain is downwards responsible for distributing Agent control quantity, and upwards responsible for summarizing environmental states to an algorithm end. Thus, depending on the particular task, dimensions in Brain specifying environmental states and dimensions in control volume are needed. The calculation formula is as follows: dimension x number of states.
An Agent component is bound for each Agent in the scene. The Agent component is responsible for acquiring specific environmental data and executing control quantity distributed by Brain.
The implementation of the function in the Agent component is described with emphasis. The method call logic diagram is shown. When the scene receives a reset instruction transmitted by the algorithm end, an AgentReset () function is called, and the attributes such as the position of the satellite are reset according to a preset rule. The scene calls the collectionservationbs () function once in each time step, adds information one by one through built-in AddVectorObs (), and sends after combination. The scene call agentenaction () function executes the control amounts saved in the action array vectorAction.
In summary, in the built virtual simulation environment, the configuration of the three ML-Agents and the realization of scene logic are completed, and the real-time simulation environment required by reinforcement learning can be built efficiently. Setting a timer in the program ensures the stability of the communication, i.e. setting a maximum waiting time T wait If the data is not received in the past, the TCP connection is disconnected, and resource waste is avoided.
Thirdly, designing and realizing the star intelligent fault diagnosis interactive virtual simulation verification platform vision software.
The overall structure of the vision software is shown in fig. 5, and the satellite group operation environment and the posture adjustment process are demonstrated by receiving the satellite group posture and data change simulation data transmitted by the main control computer and driving the earth and satellite virtual simulation model in the three-dimensional scene. The invention builds the working environment of the satellite group based on the Unity3D physical engine. The Unity is a real-time 3D interactive content creation and operation platform, has strong cross-platform characteristics, a gorgeous special effect system, a complete basic framework, a fine performance analysis tool, an extensible editor, a convenient resource management system and rich plug-ins, and is suitable for building and demonstrating the vision of the working environment.
The main interaction function of the vision software is shown in fig. 6, the star group parameter configuration can be checked at the starting interface, and the start and stop of the simulation process are controlled; after the simulation is started, the state information of any satellite, such as position, gesture, speed and the like, can be checked by clicking the satellite; the injection fault panel can select the satellite number and the type of the injection fault to be injected, and then click an injection fault button to simulate the satellite to generate faults; the diagnosis algorithm judges whether the satellites have faults according to the state information of the in-orbit satellites, and displays the diagnosis result of each satellite on a fault diagnosis result panel, wherein the diagnosis result comprises information such as whether the faults exist, the number of the fault satellites, the names of fault components and the like.
Aiming at the task environment and the demonstration requirement of the satellite group, the three-dimensional virtual scene can be divided into a scene resource module, a graphic interface module and a scene demonstration module, and the technical scheme is explained as follows.
1. Scene resource module: the invention mainly relates to the simulation of space scene resources, and particularly can be divided into the production of satellite models, the production of earth models, the realization of starry sky backgrounds and the simulation of solar illumination. The satellite and earth model is manufactured by using 3dmax, and then the built models of the earth, the satellite and the like are imported into the Unity 3D. The starry sky background is mainly implemented by using sky boxes (skyboxes) in Unity 3D. The space box is a panoramic view and consists of upper, lower, left, right, front, back and mapping along the main axis direction, and the six mapping maps are combined into a sphere, so that a continuous picture can be seen from any angle. The Unity3D has a powerful illumination system, and in fact, the global illumination system can simulate solar illumination by simulating direct light and indirect light. The specific implementation method is that the solar energy fairy (Sprite) which is manufactured by adding the light in the light attribute is calculated once per frame in real time, so that the simulated sunlight is more natural. Finally, a script for controlling the sunlight angle is mounted on the light source to simulate the day and night alternation effect brought by earth self-transmission.
2. And a graphical interface module: the text box and button part is developed by adopting UGGUI and TextMeshPro components built in the Unity 3D, and the chart part is realized by adopting Xcharts components. The graphic interface is mainly divided into three parts, namely a satellite state information panel, an injection fault panel and a fault diagnosis result panel. The satellite state information panel comprises the number, the position information and the curve graph thereof, the gesture information and the curve graph thereof, and the speed information and the curve graph thereof of the selected satellite, and a tester can interact in a mode of clicking the satellite to randomly select the satellite to check the in-orbit state; the fault injection panel comprises a fault satellite number and a fault type, a tester can select and set a satellite fault through a drop-down menu, and is provided with a random fault button to test the reliability of an algorithm, and the constellation configuration reconstruction can be realized through a configuration reconstruction button; the fault diagnosis result panel mainly displays satellite state results diagnosed by the intelligent fault diagnosis algorithm of the star group, the diagnosis result is white when no fault exists, the corresponding satellite diagnosis result becomes red and flashes to remind testers when the fault exists, and the fault diagnosis result is clear at a glance, so that the testers can record and debug conveniently.
3. A view demonstration module: the satellite group visual demonstration flow is as follows, fig. 7 is a simulation platform starting interface, and a tester can perform man-machine interaction with a virtual simulation platform to select to view the functions of satellite group configuration, start simulation or exit the simulation platform; when a tester selects to start simulation, the vision software enters an interface when a star group normally works in orbit, the star group formation keeps concentric round formations, and the tester can check the state information by clicking any satellite; above the virtual simulation platform interface is a fault simulation functional area, and a tester can interact with the simulation platform in a drop-down menu mode to select and simulate a certain satellite to generate a certain fault. FIG. 8 is a diagnostic interface after fault injection, which can show that a tester injects a star-sensitive fault, then turns red from the star 1 and changes the attitude, simulates the influence of the star-sensitive fault on the attitude of a satellite, and meanwhile, the fluctuation abnormality of the attitude curve of the satellite can be obviously seen from the state information panel of the star 1 on the left side, the intelligent fault diagnosis technology of the star cluster is displayed on the right side fault diagnosis result panel, and after the star cluster is detected to be faulty, a virtual simulation platform can pop a window to prompt the tester whether to select to reconstruct a fault configuration; fig. 9 is an interface after the constellation failure configuration is reconstructed, and it can be seen that the failed satellites have flown away from the constellation formation, and the remaining satellites of the constellation are subjected to formation reconstruction to continue to execute tasks. The full-view demonstration process verifies the effectiveness of the intelligent fault diagnosis algorithm for the star group.
Specific examples are given below:
system software and hardware configuration
Therefore, the verification method of the intelligent fault diagnosis interactive virtual simulation platform for the star group can realize real-time display of fault results, three-dimensional visualization of running states, and has the advantages of strong engineering applicability, low cost and wide application range.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention and not for limiting it, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that: the technical scheme of the invention can be modified or replaced by the same, and the modified technical scheme cannot deviate from the spirit and scope of the technical scheme of the invention.
Claims (10)
1. The verification method of the intelligent fault diagnosis interactive virtual simulation platform for the star group is characterized by comprising the following steps of:
s1, designing a star intelligent fault diagnosis technology based on multi-information fusion: the fault diagnosis technology is online fault diagnosis based on a Python deep learning algorithm;
s2, designing and realizing a real-time simulation framework of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group: the real-time interactive virtual simulation verification platform utilizes a Unity3D design, and the real-time simulation architecture of the star intelligent fault diagnosis interactive virtual simulation verification platform comprises a real-time simulation architecture of a main control computer, a real-time data interface and a vision computer;
S3, designing and realizing star intelligent fault diagnosis interactive virtual simulation verification platform vision software: the view software is based on Unity3D development, a real satellite three-dimensional model is built by adopting 3D Max, space environment building is completed by utilizing Skybox components, and the view software comprises star group parameter configuration, man-machine interaction functions and fault diagnosis result software.
2. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 1, wherein the method is characterized by comprising the following steps of: in the step S1, the fault diagnosis technology based on multiple information utilizes the position and the star sensitivity and mechanism information of the Leader satellite, and then combines the relative position and the star sensitivity information of the neighbor satellite to perform global information integration, and performs decoupling processing on the whole data information.
3. The verification method of the intelligent fault diagnosis interactive virtual simulation platform for the star group of claim 2, wherein the intelligent fault diagnosis for the star group with multiple information fusion is specifically as follows:
(1) Establishing a random game model oriented to fault diagnosis: summarizing the fault diagnosis problem of the satellite group level satellite into a random game model;
(2) Offline network training: the centralized multi-agent reinforcement learning task decision method algorithm has N satellite data as network input, and comprises a diagnosis result of the satellite to the satellite and position and speed information of the satellite, wherein the network is composed of a full-connection layer and a Bi-LSTM layer, so that the N satellite data are subjected to information communication, decision is made based on global information, meanwhile, the fault diagnosis network is not required to be changed and learned again when the number of satellites in formation changes, the Bi-LSTM-based multi-agent network decouples the data, and finally, the fault diagnosis condition of the N satellites is output after the whole information is fused;
(3) On-line fault diagnosis: after the off-line training process is completed, network parameters of an evaluation network and a decision network are determined, a central decision structure is adopted for a track planning task, and multi-satellite real-time fault diagnosis decision is finally realized in a real task environment with uncertain environment influence.
4. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 1, wherein the method is characterized by comprising the following steps of: in the step S2, the design and implementation of the real-time simulation architecture of the intelligent fault diagnosis interactive virtual simulation verification platform for the star group transmits the simulation data calculated by the main control computer end to the visual computer end for visual simulation verification, and an interactive interface with controllable data transmission time sequence is needed between the algorithm software Python and the visual software Unity.
5. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 4, which is characterized in that: the real-time simulation architecture of the main control computer, the real-time data interface and the vision computer is communicated through the ML-Agents real-time data interface, and the Socket is used for communication between the vision computer and the main control computer in the model training stage.
6. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 4, wherein the designing and implementing steps of the real-time simulation architecture of the intelligent fault diagnosis interactive virtual simulation platform of the star group are as follows:
(1) ML-Agents real-time simulation architecture: the relationship of the three high-level components of the ML-Agents framework is: the learning environment communicates with the Python API through External Communicator, and the communication function of ML-Agents is based on a client/server architecture, and Socket sockets are used for communicating a Unity process and a Python process in a model training stage;
(2) Building a real-time training environment: building a training environment, namely expanding three assemblies around the Academy, the Brain and the Agent, and building a Brain assembly under the Academy;
in the Academy component, configuring an IP address and a port number of an algorithm software end;
the Brain is downwards responsible for distributing Agent control quantity, and upwards responsible for summarizing the environment state to the algorithm end, and according to specific tasks, the Brain is required to appoint the dimension of the environment state and the dimension of the control quantity;
binding an Agent component for each Agent in a scene, wherein the Agent component is responsible for acquiring specific environmental data and executing control quantity of Brain distribution;
In the built virtual simulation environment, the configuration of three components of ML-Agents and the realization of scene logic are completed, a real-time simulation environment required by reinforcement learning is built, and the maximum waiting time T of a timer is set in a program wait If no data is received beyond the time, the TCP connection is disconnected.
7. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 1, wherein the method is characterized by comprising the following steps of: the intelligent satellite group fault diagnosis interactive virtual simulation verification platform vision software is designed and realized, the construction work of a satellite group working environment is carried out based on a Unity3D physical engine, the satellite group gesture and data change simulation data transmitted by a main control computer are received, an earth and satellite virtual simulation model in a three-dimensional scene is driven, the demonstration of the satellite group operation environment and the gesture adjustment process is carried out, and the three-dimensional virtual scene is divided into a scene resource module, a graphic interface module and a vision demonstration module.
8. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 7, wherein the specific three-dimensional virtual simulation scene of the design and implementation of the intelligent fault diagnosis interactive virtual simulation verification platform of the star group is:
(1) The scene resource module relates to the simulation of space scene resources and can be concretely divided into the production of satellite models, the production of earth models, the realization of starry sky backgrounds and the simulation of solar illumination; 3dmax is used for manufacturing the satellite and the earth model, and then the built models such as the earth, the satellite and the like are imported into the Unity 3D; the sky background is realized through a sky box in Unity 3D; the sky box is a panoramic view and is composed of upper, lower, left, right, front, back and pictures along the main axis direction, and the six pictures are combined into a sphere;
(2) And a graphical interface module: the text box and button part is developed by adopting UGGUI and TextMeshPro components built in the Unity3D, and the chart part is realized by adopting Xcharts components;
(3) A view demonstration module: after the simulation is started, the vision software enters an interface of the star group when the star group normally works in orbit, and the test personnel checks the star group configuration, starts the simulation or exits the simulation platform function by clicking any satellite to check the state information.
9. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 8, wherein the method is characterized by comprising the following steps of: the graphic interface is divided into three parts, namely a satellite state information panel, an injection fault panel and a fault diagnosis result panel.
10. The method for verifying the intelligent fault diagnosis interactive virtual simulation platform of the star group according to claim 1, wherein the method is characterized by comprising the following steps of: the man-machine interaction function comprises simulation process control, satellite state display and fault injection panel function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211637144.9A CN116068990B (en) | 2022-12-16 | 2022-12-16 | Star group intelligent fault diagnosis interactive virtual simulation platform verification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211637144.9A CN116068990B (en) | 2022-12-16 | 2022-12-16 | Star group intelligent fault diagnosis interactive virtual simulation platform verification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116068990A true CN116068990A (en) | 2023-05-05 |
| CN116068990B CN116068990B (en) | 2023-11-10 |
Family
ID=86175987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211637144.9A Active CN116068990B (en) | 2022-12-16 | 2022-12-16 | Star group intelligent fault diagnosis interactive virtual simulation platform verification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116068990B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116736883A (en) * | 2023-05-23 | 2023-09-12 | 天津大学 | Unmanned aerial vehicle cluster intelligent cooperative motion planning method |
| CN117068393A (en) * | 2023-08-21 | 2023-11-17 | 天津大学 | Star group collaborative task planning method based on mixed expert experience playback |
| CN117609908A (en) * | 2023-10-23 | 2024-02-27 | 天津大学 | Star group fault diagnosis method based on multi-information fusion |
| CN117068393B (en) * | 2023-08-21 | 2024-07-26 | 天津大学 | Star group collaborative task planning method based on mixed expert experience playback |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105511295A (en) * | 2015-11-30 | 2016-04-20 | 中国科学院光电研究院 | STKX component-based satellite vision real-time simulation system and method thereof |
| CN107204871A (en) * | 2017-04-19 | 2017-09-26 | 天津大学 | Wireless sensor network biological treatability appraisal procedure based on Evolutionary Game Model |
| CN108845802A (en) * | 2018-05-15 | 2018-11-20 | 天津大学 | Unmanned plane cluster formation interactive simulation verifies system and implementation method |
| CN109934130A (en) * | 2019-02-28 | 2019-06-25 | 中国空间技术研究院 | The in-orbit real-time fault diagnosis method of satellite failure and system based on deep learning |
| CN112650076A (en) * | 2020-11-27 | 2021-04-13 | 上海航天控制技术研究所 | Constellation cooperative control ground simulation system |
| CN113051776A (en) * | 2021-04-25 | 2021-06-29 | 电子科技大学 | Satellite attitude and orbit simulation system and method based on Unity3D |
| US11091280B1 (en) * | 2018-06-05 | 2021-08-17 | United States Of America As Represented By The Administrator Of Nasa | Modelling and analyzing inter-satellite relative motion |
| CN115408866A (en) * | 2022-09-02 | 2022-11-29 | 南京航空航天大学 | Time-varying topological structure-based constellation on-orbit health assessment method and terminal |
-
2022
- 2022-12-16 CN CN202211637144.9A patent/CN116068990B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105511295A (en) * | 2015-11-30 | 2016-04-20 | 中国科学院光电研究院 | STKX component-based satellite vision real-time simulation system and method thereof |
| CN107204871A (en) * | 2017-04-19 | 2017-09-26 | 天津大学 | Wireless sensor network biological treatability appraisal procedure based on Evolutionary Game Model |
| CN108845802A (en) * | 2018-05-15 | 2018-11-20 | 天津大学 | Unmanned plane cluster formation interactive simulation verifies system and implementation method |
| US11091280B1 (en) * | 2018-06-05 | 2021-08-17 | United States Of America As Represented By The Administrator Of Nasa | Modelling and analyzing inter-satellite relative motion |
| CN109934130A (en) * | 2019-02-28 | 2019-06-25 | 中国空间技术研究院 | The in-orbit real-time fault diagnosis method of satellite failure and system based on deep learning |
| CN112650076A (en) * | 2020-11-27 | 2021-04-13 | 上海航天控制技术研究所 | Constellation cooperative control ground simulation system |
| CN113051776A (en) * | 2021-04-25 | 2021-06-29 | 电子科技大学 | Satellite attitude and orbit simulation system and method based on Unity3D |
| CN115408866A (en) * | 2022-09-02 | 2022-11-29 | 南京航空航天大学 | Time-varying topological structure-based constellation on-orbit health assessment method and terminal |
Non-Patent Citations (2)
| Title |
|---|
| 宗群: "角速度约束卫星编队控制与虚拟演示验证", 《哈尔滨工业大学学报》, vol. 53, no. 3, pages 193 - 200 * |
| 李雪梅;: "互联网安全性能优化评估数学建模仿真研究", 计算机仿真, no. 07 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116736883A (en) * | 2023-05-23 | 2023-09-12 | 天津大学 | Unmanned aerial vehicle cluster intelligent cooperative motion planning method |
| CN116736883B (en) * | 2023-05-23 | 2024-03-08 | 天津大学 | Unmanned aerial vehicle cluster intelligent cooperative motion planning method |
| CN117068393A (en) * | 2023-08-21 | 2023-11-17 | 天津大学 | Star group collaborative task planning method based on mixed expert experience playback |
| CN117068393B (en) * | 2023-08-21 | 2024-07-26 | 天津大学 | Star group collaborative task planning method based on mixed expert experience playback |
| CN117609908A (en) * | 2023-10-23 | 2024-02-27 | 天津大学 | Star group fault diagnosis method based on multi-information fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116068990B (en) | 2023-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116068990B (en) | Star group intelligent fault diagnosis interactive virtual simulation platform verification method | |
| CN111949523B (en) | Ground closed loop simulation verification system and method for multi-satellite collaborative satellite-borne autonomous planning software | |
| CN106647335A (en) | Digital satellite attitude and orbit control algorithm ground simulation verification system | |
| CN109814478B (en) | Virtual debugging system based on iOpenWorks | |
| CN114063474B (en) | Simulation method of semi-physical simulation system based on unmanned aerial vehicle cluster | |
| CN111931371B (en) | Multi-star collaborative ground verification system application mode design method | |
| CN114415630B (en) | Comprehensive test platform and method for airplane management system | |
| CN109634141A (en) | A kind of medium-and-large-sized unmanned plane semi-physical simulation method and system that Open-closed-loop combines | |
| CN113836754A (en) | Multi-agent simulation modeling oriented simulation method, device, equipment and medium | |
| CN114338418B (en) | Virtual-real combined information network verification platform | |
| CN110134025A (en) | A kind of small distributed hypersonic aircraft real-time emulation system | |
| CN110399698A (en) | A kind of visualization conceptual design method of modularization spacecraft | |
| WO2024131078A1 (en) | Scenario self-adaptive unmanned driving control method for monorail crane transport robot | |
| CN115390585A (en) | Attitude and orbit control digital twin system based on spacecraft cluster and construction method thereof | |
| CN114186347A (en) | Multi-aircraft cooperative application simulation system | |
| CN110225100A (en) | A kind of actual situation mapped system towards Intelligent assembly production line | |
| CN113741511A (en) | Unmanned aerial vehicle cluster deduction and fault diagnosis method and system | |
| CN114578712B (en) | Multifunctional underwater autonomous vehicle cluster simulation system | |
| CN114090432A (en) | Method, system, device, electronic equipment and storage medium for simulation test | |
| CN115098941A (en) | Unmanned aerial vehicle digital twin control method and platform for agile deployment of intelligent algorithm | |
| CN110705021B (en) | Data-driven test driving method | |
| CN113848750A (en) | Two-wheeled robot simulation system and robot system | |
| Agrawal et al. | A Requirements-Driven Platform for Validating Field Operations of Small Uncrewed Aerial Vehicles | |
| Agrawal et al. | Coupled Requirements-driven Testing of CPS: From Simulation To Reality | |
| Lopez et al. | Modeling a uav surveillance scenario-an applied mbse approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |