Cloud edge end collaborative satellite data management method and system

Abstract

The application discloses a cloud-edge-end cooperative satellite data management method and system, and belongs to the technical field of satellite computing. The method comprises the following steps: collecting a physical parameter signal on the end side, and converting the physical parameter signal into sensing data; processing the sensing data by combining auxiliary data on the edge side to obtain semantic data, and performing distributed storage on the sensing data, the auxiliary data and the semantic data; managing spatiotemporal big data containing the semantic data on the cloud side, and performing unified fusion expression, query and analysis processing on the spatiotemporal big data. The application can support multi-satellite cloud cooperation intelligent computing, and finally realizes real-time, intelligent, networked and holographic satellite data processing of the earth space environment, and provides basic support for data processing and calculation processing of an intelligent satellite brain.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite computing, in particular to a cloud-edge-end collaborative satellite data management method and system. BACKGROUND

[0002] With the continuous integration and development of aerospace technology, Internet of Things, big data and artificial intelligence technology, intelligent satellite brain technology has become an important development direction. In the intelligent satellite real-time data processing mode, the system not only needs to complete the real-time acquisition of satellite data, but also needs to analyze and intelligently calculate the sampled data in real time, and through the summary calculation and analysis processing of massive data, realizes the efficient distributed intelligent processing of data.

[0003] The existing mainstream satellite on-board computing architecture is mostly a special architecture, mainly focusing on on-board high-performance special computing systems and on-board edge computing platforms. The first disadvantage of these methods is that due to the weak on-board data processing capability, it is difficult to adapt to the application requirements of on-board intelligent processing of space data; in addition, it cannot be linked in real time with the ground, and cannot effectively manage satellite data in real time on the ground.

[0004] For a few existing works on cloud-edge-end collaborative satellite data processing, for example, Chinese patent application 202211673001.3 discloses a satellite cluster-oriented cloud-edge collaborative computing system and its management method, which can integrate the computing resources of satellite clusters, realize computing collaboration between satellites and between satellites and the ground, and improve data processing efficiency and computing resource utilization, but there are the following significant deficiencies in data management:

[0005] 1. The existing work only briefly mentions remote sensing and telemetry data, but does not detail the structure of the data itself and does not explicitly explain and design the data processing method.

[0006] 2. The existing work only briefly involves data transmission from satellite edge to ground cloud side, but does not consider the characteristics of time continuity and discontinuity, limited bandwidth, high delay, and unstable link in satellite-ground and satellite-satellite communication. These characteristics are inherent characteristics of current actual data links and are important considerations for system practicality.

[0007] 3. The existing work does not involve the solution to the edge autonomy problem when the satellite cannot communicate with the ground station due to satellite orbit reasons (or cannot communicate in real time in the satellite relay communication scenario).

[0008] 4. The existing work involves control of on-board loads. But it does not consider the fine management of end-side load device control parameters and the linkage with the cloud-edge-end data management process, such as parameter persistence at the edge to cope with reconnection recovery, specific closed-loop details of on-demand return, etc.

[0009] In summary, there is no systematic work in intelligent satellite computing and multi-satellite cloud collaboration data management at present, which is difficult to meet the real-time, intelligent, networked and holographic requirements of Earth space environment satellite data processing. SUMMARY

[0010] The application discloses a cloud-edge-end collaborative satellite data management method and system, aiming to propose and design a mechanism method and system of cloud-edge-end collaborative data management. The cloud-edge-end collaborative data management can support multi-satellite cloud collaboration intelligent computing, and finally realize real-time, intelligent, networked and holographic Earth space environment satellite data processing, providing basic support for intelligent satellite brain data processing and computing.

[0011] To achieve the above purpose, the technical scheme of the application includes the following contents.

[0012] A cloud-edge-end collaborative satellite data management method, the method comprising:

[0013] Collecting physical parameter signals on the end side, and converting the physical parameter signals into perception data;

[0014] Combining auxiliary data to process the perception data on the edge side to obtain semantic data, and performing distributed storage on the perception data, the auxiliary data and the semantic data; wherein the auxiliary data comprises geographic space data, ionosphere global model data and target recognition model library data;

[0015] The cloud side manages spatio-temporal big data containing the semantic data, and performs unified fusion expression, query and analysis processing on the spatio-temporal big data.

[0016] Further, the perception data comprises numerical perception data and multimedia perception data;

[0017] The edge side combines auxiliary data to process the perception data to obtain semantic data, comprising:

[0018] Extracting numerical perception data in a time period in units of components to obtain a vector representation of the numerical perception data; wherein the components comprise a pos component or a speed component of a spatial position sensor sampling value;

[0019] Intelligently processing the multimedia perception data to obtain semantic description of the multimedia perception data; wherein the intelligent processing comprises orthographic calculation, geographic registration and target recognition of image data;

[0020] According to the sampling time and the sampling location of the perception data, the vector representation of the numerical perception data, the semantic description of the multimedia perception data, and the identification of the edge side, semantic data corresponding to the perception data is generated.

[0021] Further, the perception data is stored in a distributed manner, comprising:

[0022] According to a Hash, Round Robin or regional distribution strategy, the entire set of perception data records is stored in a distributed manner on each storage node; wherein for perception data containing file pointers, the perception data and the corresponding data file are stored on the same storage node.

[0023] Further, the geographic space data and the semantic data are stored in a distributed manner, comprising:

[0024] For geographic space data and semantic data required for spatio-temporal calculation, spatio-temporal query and spatio-temporal data access operation on the edge side, geographic space data and semantic data with a usage frequency greater than a set value are filtered out;

[0025] Using a spatial-based or spatio-temporal-based data distribution algorithm, the geographic space data and the semantic data are stored in a distributed manner, and the geographic space data and semantic data with a usage frequency greater than a set value are concentrated on a specific storage node.

[0026] Further, the cloud side comprises: an interface service, a data catalog management module, an edge side management module, a device twin management module, a command and mode persistence module, a cloud side spatio-temporal big data storage module, a cloud side computing resource pool, a real-time message engine and a first cloud edge collaborative management module.

[0027] The interface service is configured to provide a standardized API interface for authorized users on the edge side or other authorized users to access data resources and computing capabilities of the cloud side and issue control instructions.

[0028] The data catalog management module is configured to perform unified metadata management, cataloging and indexing on spatio-temporal big data stored on the cloud side, and metadata management and cataloging on accessible data on the edge side.

[0029] The edge side management module is configured to register, monitor, configure and manage the accessed edge side.

[0030] The device twin management module is configured to maintain a digital twin for each edge side load device on the cloud side, and the digital twin synchronizes the state, attributes and configuration of the edge side load device.

[0031] The command and mode persistence module is configured to store commands, task configurations, data modes and models issued to the edge side and the end side.

[0032] The cloud-side spatiotemporal big data storage module is configured to store spatiotemporal big data and other service data from the edge side and the end side.

[0033] The cloud-side computing resource pool is configured to provide computing resources for supporting model training and data analysis.

[0034] The real-time message engine is configured to process the message queue of instruction issuing and data uploading between the cloud side and the edge side based on a message bus mechanism.

[0035] The first cloud-edge-end collaborative management module is configured to coordinate the task allocation, data flow and state synchronization among the cloud side, the edge side and the end side.

[0036] Further, the edge side comprises a data service, a task manager, a data catalog manager, a device manager, a data processor, a message bus, a model and mode module, a second cloud-edge-end collaborative management module, an edge-side spatiotemporal underlying storage and an edge-side computing resource pool.

[0037] The data service is configured to perform data query processing, data subscription and data transmission.

[0038] The task manager is configured to receive, schedule and execute computing tasks and business logic defined from the cloud or locally.

[0039] The data catalog manager is configured to perform metadata management and cataloging on the data stored in the edge side, the cloud side and the edge side.

[0040] The device manager is configured to manage the end side connected through a device exporter.

[0041] The data processor is configured to perform preprocessing operations on the raw data collected from the end side or the data received from the cloud side, the preprocessing operations including cleaning, conversion, aggregation and feature extraction.

[0042] The message bus is configured to perform asynchronous communication among the modules in the edge side and with the end side.

[0043] The model and mode module is configured to store model files, data processing modes and business rules locally in the edge side.

[0044] The second cloud-edge-end collaborative management module is configured to be a collaborative management unit in the edge side and cooperate with the corresponding module in the cloud side.

[0045] The edge-side spatiotemporal underlying storage is configured to persistently store raw data, processed data, application states and necessary configuration information locally in the edge side.

[0046] The edge-side computing resource pool is configured to provide local computing resources.

[0047] Further, the edge side includes devices and applications.

[0048] The devices are configured to collect data or receive instructions for execution, and the devices communicate with the device manager and the message bus of the edge side through the device exporter.

[0049] The applications refer to logic running on the intelligent end devices or directly controlled by the edge side.

[0050] Further, the process of collaborative work of the end side, the edge side and the cloud side includes:

[0051] The cloud side transmits relevant instructions to the edge side through the first cloud-edge-end collaborative management module via the interface service and the real-time message engine, wherein the relevant instructions include control instructions, task configurations and model updates.

[0052] After the second cloud-edge-end collaborative management module of the edge side receives the instructions, it schedules local resources to execute tasks through the task manager, and controls the devices of the end side through the device manager and the device exporter.

[0053] After the devices of the end side collect data, the devices upload the data to the message bus of the edge side through the device exporter.

[0054] After the data processor of the edge side processes the data, it uploads the data to the cloud side via the real-time message engine of the cloud side through the second cloud-edge-end collaborative management module and the data service.

[0055] The device twin management module of the cloud side, the device manager of the edge side and the related storage module jointly maintain the consistency of the state of the devices and the applications of the end side, so that when the network is interrupted, the edge side can run autonomously based on the model and mode module, the edge side space-time underlying storage and the data catalog management module.

[0056] Further, the control of the cloud side on the end side load device is realized by the following steps:

[0057] The user selects a target end side load device, and sets a control parameter to be adjusted and a target value of the control parameter.

[0058] The cloud side encapsulates the target end side load device, the control parameter to be adjusted and the target value of the control parameter into a declarative instruction conforming to a predefined specification, and transmits the declarative instruction to the target edge side.

[0059] The target edge side parses the declarative instruction, identifies the target side load device, the control parameter to be adjusted and the target value of the control parameter, and encapsulates the control parameter to be adjusted and the target value of the control parameter into a bottom layer instruction recognizable by the target side load device.

[0060] After the target edge side addresses and locates the target side load device by using the device manager, the bottom layer instruction is sent to the target side load device, and the parsing result of the declarative instruction is locally persisted at the target edge side.

[0061] A cloud edge end cooperative satellite data management system, the system comprises:

[0062] The end side is used for collecting physical parameter signals and converting the physical parameter signals into perception data.

[0063] The edge side is used for processing the perception data in combination with auxiliary data to obtain semantic data, and performing distributed storage on the perception data, the auxiliary data and the semantic data; wherein the auxiliary data comprises geographic space data, ionosphere global model data and target recognition model library data.

[0064] The cloud side is used for managing spatio-temporal big data containing the semantic data, and performing unified fusion expression, query and analysis processing on the spatio-temporal big data.

[0065] Compared with the prior art, the present application has at least the following beneficial effects.

[0066] Compared with the prior art (for example, only focusing on space-based edge computing, or lacking a fine data management and cooperative transmission mechanism specific to satellite communication constraints), the present application has significant advantages and beneficial effects in terms of data processing efficiency, system reliability, edge intelligence, network adaptability and resource utilization. These effects are due to the multiple innovative technical features of the present application in data abstraction, semantic extraction, communication mechanism, edge autonomy and system architecture.

[0067] 1. Significantly improve data processing and transmission efficiency, reduce bandwidth pressure:

[0068] Technical means: The present application adopts a step-by-step data abstraction mechanism of "end->edge->cloud", especially a unique semantic data record generation and update method based on vectorization expression. At the edge side, the numerical perception data record is vectorized to fit multiple continuous sampling values into a time function vector=(f(t),t start ,t end ), and an intelligent update strategy of active vector comparison and threshold triggering is adopted; for multimedia sampling data, key semantic descriptions (such as JSON documents) are extracted.

[0069] • Principle and theoretical effect:

[0070] ■In theory, the vector representation can compress a large number of redundant time series data points into a small number of parameterized vector descriptions, and the data compression ratio can reach several times or even dozens of times (depending on the data smoothness and threshold). For example, for slowly changing sensor data, hundreds of sampling points may only need one or a few vectors to accurately represent them.

[0071] ■The active vector update mechanism ensures that only when the data actually changes significantly (beyond the threshold) will a new vector be generated and uploaded to the cloud, greatly reducing the frequency of data upload.

[0072] ■The semantic description of multimedia data extraction allows the cloud to directly obtain key information, avoiding the transmission of large raw multimedia files for preliminary analysis.

[0073] • Beneficial effect: Directly solves the pain point of extremely valuable satellite communication bandwidth, greatly reduces the transmission of invalid and redundant data, saves satellite-ground communication resources, and at the same time improves the value density and processing efficiency of cloud data.

[0074] 2. Enhance the reliability and resilience of the system under extreme network conditions:

[0075] • Technical means: The invention designs a "cloud-edge collaborative data transmission and management network architecture and collaborative subsystem" suitable for satellite communication characteristics, which includes a message bus-based communication mechanism (cloud-side real-time message engine, edge-side message bus), standardized declarative interface services, and edge autonomous mechanisms for applications and data (edge metadata persistence, local storage and models, etc.).

[0076] • Principle and theoretical effect:

[0077] ■The message bus mechanism ensures the reliable and orderly delivery of instructions and data in a network that is connected and disconnected at different times through message queues, acknowledgments, and retransmissions.

[0078] ■Declarative interface defines the target state rather than the process, allowing the system to better handle high latency and intermittent connections. Even if there is a delay in executing instructions, the consistency of the final state can be guaranteed.

[0079] ■The edge autonomous mechanism uses local persistent configurations, metadata, models, and patterns, as well as data, so that edge nodes can independently execute scheduled tasks, process local data, and serve end-side devices when disconnected from the cloud. After the network is restored, it can synchronize the state with the cloud and continue data transmission.

[0080] The beneficial effect is that the working stability and service continuity of the system in a poor satellite communication environment are significantly improved, the reliable interaction of core data and instructions is ensured, and the survivability of the entire system and the resilience of task execution are enhanced.

[0081] 3. Fine closed-loop control and intelligent state monitoring of the end-side load device are realized.

[0082] The technical means are that the control initiated by the cloud-side user is delivered to the edge-side "device manager" for accurate execution through a declarative interface, the control parameters are recorded in the edge-side "persistent configuration management", and the end-side device state data is semantically converted and intelligently returned on demand based on the strategy by the edge-side "data processor".

[0083] The principle and theoretical effect are that the declarative delivery of the control parameters and the edge persistence ensure that the control intent can be accurately executed or restored even after network fluctuation or device restart.

[0084] The edge semantic processing and on-demand return of the state data avoid the blind upload of a large amount of raw state data, only return the information valuable to the cloud decision, and improve the signal-to-noise ratio.

[0085] A complete, efficient and intelligent closed loop is formed from "cloud intent -> edge execution -> end-side response -> edge perception and processing -> cloud feedback".

[0086] The beneficial effect is that the control precision and response speed of the remote end-side load device are improved, the efficiency and intelligent level of state monitoring are optimized, and the entire business process is more agile and efficient.

[0087] The beneficial effect is that the control precision and response speed of the remote end-side load device are improved, the efficiency and intelligent level of state monitoring are optimized, and the entire business process is more agile and efficient. BRIEF DESCRIPTION OF DRAWINGS

[0088] Figure 1 is a cloud-edge-end collaborative data management overall mechanism.

[0089] Figure 2 is a cloud-edge-end collaborative data management system architecture.

[0090] Figure 3 is the extraction of semantic data records. DETAILED DESCRIPTION

[0091] The application will be further described in detail below in combination with the drawings and examples, and it should be pointed out that the following examples are intended to facilitate the understanding of the application and do not limit the application in any way.

[0092] In the intelligent satellite real-time data processing mode, the system not only needs to complete the real-time acquisition of satellite data, but also needs to analyze and intelligently calculate the sampled data in real time, and through the summary calculation and analysis processing of massive data, realizes the efficient distributed intelligent processing of data. Satellite payload equipment generates a large amount of sampling data, these on-board equipment including multi-ionospheric detector, spectrometer, quantum wave instrument, SAR radar, etc.; In addition, the satellite-ground and interstellar communication is unreliable transmission, which needs to reduce the data communication cost as much as possible. In this application scenario, the traditional cloud computing centralized data processing mode and the simple edge computing processing mode cannot effectively meet the functional or performance requirements, and a new type of cloud-edge-end collaborative data management mode needs to be adopted.

[0093] In the intelligent satellite brain system, the on-orbit satellite is not only a data acquisition device, the fleeting space weather phenomenon, ground abnormal event, etc. need the satellite to have the autonomous calculation, analysis and strain ability, at the same time, the key information is transmitted to the ground data center, and the data sharing and computing power sharing are realized through the cloud-edge-end collaborative mechanism. In the above application, the cloud computing centralized data management mode will cause the cloud computing center to be overburdened, and the simple edge computing mode cannot effectively support the summary intelligent analysis involving a large amount of data. It can be seen that the cloud-edge-end collaborative data management is the core key to break through the system functional performance bottleneck and realize the rapid development of satellite industry.

[0094] Through the cloud-edge-end collaborative mechanism, the multi-satellite, star-ground efficient data sharing and computing collaboration are realized in the form of service call, and the space scientific data on-board unified expression modeling is realized.

[0095] Specifically, the overall process of cloud-edge-end collaboration is described as follows, as shown in Figure 1 In the data sampling process of the cloud-edge-end collaborative system, the sampling data is continuously processed in an abstract manner along the direction of "end→edge→cloud". Different levels of computing platforms in "end→edge→cloud" store data of different abstraction levels, and different levels of data do not repeat each other, and together complete the description of the state of the objective world:

[0096] The end computing platform is a special computing platform on the sensing device (such as sensor, multimedia monitoring device, sampling device, etc.). It collects physical parameter signals and converts them into meaningful sensing data records through a special processing module. For example, the on-board image imaging device transmits and receives observation signals, and then obtains the image data of the observed area through the processing of the end computing platform.

[0097] ● An edge computing platform is a general-purpose computing platform that sits between edge computing platforms and cloud computing platforms, acting as a bridge between them. Typically, an edge computing platform can connect to and manage multiple edge computing devices; for example, a satellite edge computing platform can manage various sensor devices and multimedia sensing sampling devices on a satellite. The edge computing platform stores sensing data records from edge devices and processes them to obtain semantic data records, such as IoT state vectors, target recognition information, and geographic modeling data.

[0098] ● The cloud data management platform is also a general platform that manages semantic sampling data and other related data from all edge computing platforms within the system (which together constitute spatiotemporal big data). The cloud computing platform performs unified fusion, expression, querying, and analysis processing on this data.

[0099] The cloud-edge-device collaborative satellite data management method and system mainly includes three sub-methods and sub-systems.

[0100] 1. The first step is the convergence and flow of data along the "device → edge → cloud" path. This requires addressing the management challenges arising from the massive volume and heterogeneous nature of data from satellite-based edge-side payload sensing devices.

[0101] 2. Secondly, given the intermittent connection and bandwidth-limited nature of the data link from the satellite side to the ground cloud side, a special design is required for the network architecture supporting data transmission and management.

[0102] 3. Furthermore, how to effectively coordinate the edge and end sides to better control and manage the status and data sampling parameters of the end-side load sensing devices is crucial for effectively managing the source of data generation.

[0103] Based on the aforementioned overall process of cloud-edge-device collaboration, the architecture diagram of the cloud-edge-device collaborative data management system is as follows: Figure 3 As shown.

[0104] 1. Satellite data collaboration method and subsystem based on cloud-edge-device three-level collaboration

[0105] During the data sampling process of space environment data, such as Figure 2 As shown, data is abstracted step by step along the "device → edge → cloud" path, with different levels of computing platforms storing data at different levels of abstraction. The edge computing platform is a dedicated computing platform on sensing devices (such as various space environment data sampling devices), which is responsible for collecting on-site data.

[0106] An edge computing platform is a general-purpose computing platform located between edge computing platforms and cloud computing platforms. An onboard data processing platform is one such edge computing platform. It manages various space environment data acquisition devices on the satellite, performs corresponding data processing and vector extraction on the acquired sensing data, and generates and outputs semantic data.

[0107] Since the end-computing platform generally performs periodic data collection, a large amount of redundant data can be generated. To reduce the amount of redundancy, the present application adopts a vector-based semantic data record expression method. A vector corresponds to a plurality of continuous sampling values. By using a straight line or curve fitting method, it can be expressed as a function f(t) of time, with the start time t start and the end time t end To reduce the vector redundancy, the difference between the actual sampling value and the vector calculation value is calculated and compared with the size of the specified threshold ε. The two vectors with a difference less than ε are merged, and finally the purpose of reducing the data update frequency and data amount of the cloud computing platform is achieved.

[0108] In the intelligent satellite brain cloud-edge-end collaborative system, the cloud computing platform (usually located on the ground) is the computing center, which collects data of all satellites accessing the system and provides data and program access interfaces. The on-board computing platform realizes data sharing by retrieving global data through the data access interface, and shares intelligent algorithms and computing capabilities by using the computing capabilities of the ground cloud computing platform through the corresponding interface.

[0109] Specifically:

[0110] The data managed by the edge computing platform is perception data records (such as original SAR imaging pictures) and semantic data records (such as geographic model data obtained by orthographic calculation, geographic calibration, and three-dimensional modeling). Due to the limitation of the storage capacity of the edge computing platform, only data within a certain period of time is generally stored. In order to facilitate geospatial computing, the edge computing platform often needs to store geospatial data records (such as map information). Of course, there are also a small amount of other types of auxiliary data necessary for edge computing, such as ionospheric global model data and target recognition model library data. These data can be stored in the form of files or relational tables on one or more fixed processor nodes. The following mainly describes the distributed management method of perception data records, semantic data records, and geospatial data records.

[0111] For the distributed storage of perception data records, since it does not contain geospatial information, a data distribution method based on end device identification can be adopted, that is, according to Hash, Round Robin or regional distribution strategy, the entire set of perception data records is distributed and stored on each storage node (for perception data records containing file pointers, the data file and the perception data record are stored in the same node to enable the file to be accessed locally through the file pointer on the data record during data processing). The above method can ensure that all perception data records generated by the same end device are stored in the same processor node, which is beneficial to the extraction and calculation of semantic data records.

[0112] For the distributed storage of semantic data records and geospatial data records, since they have space-time characteristics, a space-based or space-time-based data distribution method can be adopted. Through a reasonable data distribution strategy, the data records required for frequent space-time calculation, space-time query and space-time data access operations of the edge computing platform can be concentrated in a small number of storage nodes, thereby improving the overall computing performance.

[0113] The data processing tasks of the edge computing platform mainly include two categories: query processing and analysis calculation tasks. For query processing tasks (such as data query through identification, time, space, state value, etc.), the above distributed structure can be used to efficiently complete the tasks. Through analysis tasks (such as statistical analysis of perception sampling data, modeling calculation and target recognition of original SAR images, etc.), a space-time Map / Reduce-based method can be used to divide the data objects to be processed (such as semantic data record set, original SAR image, etc.) according to space or space-time attributes, distribute them to each storage node for parallel processing, and finally aggregate the final processing result through Reduce operation.

[0114] In addition to parallel processing of massive data, the edge computing platform can also perform parallel intelligent analysis on a single multimedia data. For example, in the target recognition processing of SAR images, the image can be cut and distributed to each server for parallel processing, and finally the final result is obtained by aggregating the sub-results.

[0115] Through the cloud-edge-end collaborative mechanism, efficient data sharing and flow between multiple satellites and satellite-ground are realized, the efficiency of space information transmission and processing is improved, and the problem of space information transmission bottleneck is solved.

[0116] Further development:

[0117] Since the sampling data of the perception device corresponding to the end computing platform has extensive heterogeneity, a unified data expression method is needed to represent it. The application designs a perception data record expression method based on Schema+Value. Perception data can be roughly divided into two categories, namely numerical sampling values (such as atmospheric density load, electron density load sensor, etc. sampled data) and multimedia sampling values (such as video monitoring image, high altitude and geological exploration image data, audio monitoring signal, etc.). The perception data record SenseData can be uniformly represented as the following format (the following description sets TimeInstant, Point, String, LowAltitudeFile as time, spatial coordinates, string, low altitude file data type respectively):

[0118] SenseData=(t,schemaValue,termID)

[0119] Wherein, t TimeInstant is the sampling time corresponding to the sampling data; schemaValue String is a structured string composed of schema String and value String, schema and value are the "type" and "value" of the sampling data respectively, wherein "type" describes the format and data type of the sampling data, and "value" is the specific sampling data value; termID String is the identification of the end computing platform.

[0120] The perception data record generated by the end computing platform is uploaded to the edge computing platform, and the edge computing platform generates and outputs semantic data records after data processing and vector extraction, as shown in the following formula: Figure 3

[0121] The end computing platform generally adopts a periodic data sampling method, so the perception data record has a large amount of redundancy. In addition, the perception data record of the multimedia type is only the original sampling information (such as the original image without orthographic calculation and geographical matching). Therefore, the application designs a semantic data record expression method based on vector to minimize redundancy and reduce the amount of data uploaded to the cloud computing platform.

[0122] ​The process of generating semantic data records by the edge computing platform is as follows: for numerical perception data records, the extraction of vectors is performed in units of components (such as the pos component or the speed component of spatial position sensor sampling values). A vector corresponds to a plurality of consecutive sampling values, and corresponds to a curve segment l in a VxT plane (where V and T are the value range of the component sampling values and the value range of the sampling time, respectively). The extraction of the vector can be obtained by linear fitting or curve fitting, and can be expressed as a function f(t) of time, plus the start time t start and the end time t end , that is:

[0123] vector = (f(t), t start , t end )

[0124] For each component of each perception device, the last vector is its current active vector (the t end attribute of the current active vector is an undefined value). When new perception data records arrive, the edge computing platform compares the new sampling values with the current active vector. If the difference between the actual sampling values and the vector calculation values is less than a specified threshold value ε, no processing is required; only when the actual sampling values deviate from the current active vector (i.e., the deviation between the two exceeds the specified threshold value), the fitting calculation of a new vector is triggered, a new active vector is generated, and the new active vector is uploaded to the cloud computing platform (the t end of the original active vector is updated to the current time). Since the update speed of the vector is much lower than the sampling speed of the perception data records, the above method effectively reduces the data update frequency and data volume overhead of the cloud computing platform.

[0125] For multimedia sampling data, the edge computing platform can derive relevant semantic descriptions (such as the location and area of a fire area, and the type and location of suspicious and sensitive targets) through intelligent processing (such as orthographic calculation, geographic registration, target recognition, etc. of image data). These semantic description information can be expressed as a JSON document, and also generates a corresponding semantic data record. The multimedia file (such as a low-altitude image file, a geographic model data file) processed by the edge computing platform and the JSON document are stored in the same storage node through file pointer association, so as to facilitate further file access and processing.

[0126] In summary, the two types of perception data records correspond to semantic data records SemanticData, which can be uniformly expressed in the following format:

[0127] SemanticData = (t, pos, vect, edgeID)

[0128] where t TimeInstant is the sampling time corresponding to the data record; pos Point is the sampling location corresponding to the data record; vect Vector U JSON is the generated vector, and edgeID String is the identification of the edge computing platform.

[0129] In the cloud-edge-end collaborative system, the cloud computing platform (usually located on the ground) acts as the hub, collects data of all satellites accessing the system, and provides data access interfaces and program access interfaces. The edge computing platform can query, retrieve and access the global data of the hub through the data access interface, thereby realizing data sharing.

[0130] In order to realize satellite data collaboration based on three-level collaboration of cloud-edge-end, there are two supporting subsystems: cloud-edge collaborative data transmission and management network architecture and collaboration subsystem, and edge-end collaborative end-side load equipment control parameter management and state monitoring subsystem. The following will be introduced respectively.

[0131] 2. Cloud-edge collaborative data transmission and management network architecture and collaboration subsystem

[0132] The data link between the satellite edge side and the ground cloud side has the characteristics of intermittent connection and limited bandwidth. Specifically, the challenges in building an efficient and reliable satellite edge side and ground cloud side data link include: ①First, the participants of the communication, especially the sending and receiving terminals on the edge side, cannot be guaranteed to be online for a long time. They may enter a sleep or low-power state within a certain time window, which poses a great challenge to the continuity, immediacy and sequence of communication; ②Second, the inherent network limitation of satellite communication, mainly manifested in limited available bandwidth and high delay caused by signal propagation, which directly affects the rate and stability of data transmission; ③Third, the stability of network connection is also difficult to be fully guaranteed. The connection may be unstable due to various factors, and even may be offline at any time. Under this condition, the edge application still needs to have the ability to work continuously to ensure that the business does not be interrupted or can be properly handled in the offline state.

[0133] To address the aforementioned challenges, this invention designs a cloud-edge collaborative data transmission and management network architecture and collaborative subsystem, aiming to build a reliable cloud-edge collaborative data link. ① A message bus-based mechanism is designed. Its core advantage lies in ensuring the sequentiality of information during transmission and effectively addressing communication interruptions caused by intermittent terminal online activity or unstable network connections, ensuring the final reliable delivery of messages. ② This service model is dedicated to handling critical operations such as model parameter updates accurately and stably, while providing good support for remote operations and effectively tolerating execution delays caused by network factors such as high latency. ③ An edge autonomy mechanism for applications and data is designed. Specifically, this includes supporting the autonomous operation of edge devices in offline network conditions. By introducing key technologies such as edge metadata persistence and task management, it ensures that even when disconnected from the central network, the various application loads on the edge side can continue to operate stably and have rapid fault recovery capabilities.

[0134] The cloud-edge-device collaborative data transmission and management subsystem proposed in this invention has the following core architecture: Figure 3 As shown, it mainly comprises three parts: cloud, edge, and device, operating through defined interfaces and coordination mechanisms. The subsystem is dedicated to providing stable, efficient, and edge-autonomous data interaction and management capabilities in satellite and terrestrial network communication environments.

[0135] Subsystem architecture and composition:

[0136] (1) Cloud Side

[0137] As the central control and data aggregation node, it is responsible for global task scheduling, large-scale data storage and analysis, model training, etc. Its main modules (based on...) Figure 2 The Query area and related modules include:

[0138] ● Interface Services (Query - Data, Computation, General Commands): Provides standardized API interfaces for edge users or other authorized users to access cloud-side data resources and computing capabilities, and to issue control commands. Specifically, it includes data services, computing services, and general API services. Data services include data query processing, data subscription, and data transmission functions.

[0139] ● Data Catalog Management: Provides unified metadata management, cataloging, and indexing for various types of data stored in the cloud, facilitating querying and traceability. It also manages and catalogs metadata for data accessible from the edge.

[0140] • Edge Management: responsible for registering, monitoring, configuring, and managing the connected edge nodes.

[0141] • Device Twin Management: maintains a digital twin for each edge-side load device in the cloud, synchronizing its state, attributes, and configurations, and supporting remote monitoring and pre-operation.

[0142] • Command & Schema Persistence: stores commands, task configurations, data schemas, and models issued to the edge and terminal.

[0143] • Cloud-Side Spatiotemporal Big Data Storage: used to store large-scale spatiotemporal data and other business data from the edge and terminal.

[0144] • Cloud Computing Resource Pool: provides elastic computing resources for model training and complex data analysis.

[0145] • Real-time Message Engine: a key component for asynchronous message communication between the cloud and edge, using a message bus mechanism to handle message queues for instruction issuance and data upload.

[0146] • Cloud-Edge-Device Collaboration Management: coordinates the core logic control unit for task allocation, data flow, and state synchronization between the cloud, edge, and terminal. It uses declarative interfaces and combines persistence capabilities to securely and reliably deliver interface instructions.

[0147] (2) Edge Side

[0148] Deployed on edge computing nodes close to data sources or execution points, responsible for preliminary data processing, real-time task execution, and local autonomous management. Its main modules include:

[0149] • Data Service: includes data query processing, data subscription, and data transfer functions.

[0150] • Task Manager: responsible for receiving, scheduling, and executing computing tasks and business logic from the cloud or locally defined.

[0151] • Data Catalog Management: Meta-data management and cataloging of data stored locally on the edge, in the cloud-edge, and edge-to-edge.

[0152] • Device Manager: Manages edge-side devices connected through the "Device Exporter", monitors their status, and forwards control instructions to them.

[0153] • Data Processor: Cleans, transforms, aggregates, and extracts features from raw data collected from edge-side devices or data received from the cloud.

[0154] • Message Bus: Asynchronous communication core component between edge-side modules and edge-side devices (through the Device Exporter), ensuring the orderly flow of data and instructions.

[0155] • Models & Schemas: Model files, data processing schemas, and business rules stored locally on the edge, supporting local application execution.

[0156] • Cloud-Edge-Device Collaboration Management: Edge-side collaboration management unit, working with the corresponding cloud-side modules to execute collaborative tasks.

[0157] • Edge-Side Spatiotemporal Underlying Storage: Local persistent storage of raw data, processed data, application state, and necessary configuration information on the edge.

[0158] • Edge Computing Resource Pool: Local computing resources provided by edge-side nodes.

[0159] (3) Device Side

[0160] Refers to various sensors, actuators, smart devices, etc. connected to the edge.

[0161] • Devices: Actual physical devices responsible for data collection or receiving instructions to perform actions. They communicate with the "Device Manager" and "Message Bus" on the edge through the "Device Exporter".

[0162] • Applications: simple logic that can run on certain smart end devices or be directly controlled by the edge side.

[0163] Brief description of collaborative workflow:

[0164] A) Instruction and task issuance: the cloud side issues control instructions, task configurations, model updates, etc. through the "interface service" and "real-time message engine" to the edge side through the "cloud-edge-end collaborative management" module.

[0165] B) Edge side processing and execution: the "cloud-edge-end collaborative management" module of the edge side receives instructions, schedules local resources to execute tasks through the "task manager"

[0166] , controls end side devices through the "device manager" and "device exporter".

[0167] Data processing is completed by the "data processor".

[0168] C) Data collection and upload: end side devices collect data and upload them to the "message bus" of the edge side through the "device exporter". After processing by the "data processor", data are uploaded to the cloud side through the "cloud-edge-end collaborative management"

[0169] module and "data service" of the edge side via the "real-time message engine".

[0170] D) State synchronization and autonomy: the "device twin management" of the cloud side and the "device manager" and related storage modules of the edge side jointly maintain the consistency of device and application states. In the event of network interruption, the edge side utilizes the locally stored "models and patterns", "edge side space-time underlying storage", and "data catalog management" to achieve autonomous operation.

[0171] 3. End side load device control parameter management and state monitoring subsystem of edge-end collaboration

[0172] An end side load device control parameter management and state monitoring subsystem and method of edge-end collaboration are provided, which combine the "cloud-edge collaborative data transmission and management network architecture and collaborative subsystem", utilize the basic capabilities provided thereby, and achieve precise control and intelligent state monitoring of end side load devices.

[0173] The working process mainly includes the following aspects:

[0174] (1) End side load device control parameter issuance and management process:

[0175] A) Cloud side user initiates control request:

[0176] • The user selects the target edge device and sets the control parameters (e.g., sensor sampling rate, camera exposure time, actuator action threshold, etc.) and their target values through the cloud-side application interface.

[0177] B) Declarative packaging and issuing of control instructions:

[0178] • The cloud-side encapsulates the user-set control parameter targets into declarative instructions conforming to pre-defined specifications.

[0179] • The declarative instructions are securely delivered to the edge-side transceiver guard service on the target edge node through the declarative interface provided by the data link and management subsystem.

[0180] C) Edge-side reception, parsing, and addressing:

[0181] • The edge node's edge-side transceiver module (or through the service bus) receives the declarative control instructions from the cloud.

[0182] • The edge-side transceiver guard service parses the instructions, identifying the target device identifier and specific control parameters and target values.

[0183] D) Execution of control through the device manager module:

[0184] • The edge node's device manager module locates the specific edge device based on the parsed device identifier.

[0185] • The device manager module issues the parsed control parameters (converted into device-recognizable low-level instructions) to the edge device, driving the device to perform parameter adjustments.

[0186] E) Local persistence of control parameters:

[0187] • After successful issuance of control instructions or device confirmation of parameter updates, the new control parameter values are recorded and updated in the edge node's persistent configuration management module. This ensures that even in the case of disconnection with the cloud, the edge node retains the latest configuration state of the device and can restore the configuration after device restart or reconnection.

[0188] (2) Edge-side device state monitoring and data backflow process:

[0189] A) Generation of edge device state and data:

[0190] • After applying new control parameters, the edge device's working state or generated data characteristics may change (e.g., adjustment of sampling rate directly affects data generation rate and density).

[0191] • The device continuously collects or generates perception data.

[0192] B) Perception data injection and edge real-time processing:

[0193] • The raw perception data (perception data records) generated by the end-side device is injected into the "event bus" of the edge node (or the edge-side real-time message engine concept, which is a high-efficiency data distribution mechanism within the edge) through the underlying data collector.

[0194] • Specific business applications deployed on the edge node can subscribe to and obtain these raw data from the "event bus" in real time, for immediate analysis, processing, or local decision-making.

[0195] C) Data semantic conversion:

[0196] • The edge node has the ability to process and convert raw perception data. Through pre-set or dynamically loaded "vector extraction methods" (or other data processing algorithms), raw and possibly redundant perception data records are converted into more structured and information-dense semantic data records. This process helps reduce the amount of data transmitted and improves data value.

[0197] D) On-demand intelligent backhaul:

[0198] • The converted semantic data records are not all unconditionally backhauled to the cloud.

[0199] • The backhaul strategy is based on the configuration of the "cloud-edge data flow conversion subsystem" stored in the "persistent configuration management" module (e.g., by time window, by event trigger, by data change amplitude, by network condition, etc.).

[0200] • The "edge-side transceiver module" of the edge node selectively and on-demand backhauls important semantic data records to the "cloud-side transceiver guard service" of the cloud through the channels provided by the "data link and management subsystem" according to these strategies.

[0201] E) Cloud-side data reception and application:

[0202] • After receiving the backhauled semantic data records, the cloud-side can perform further storage, analysis, display, or use for higher-level decision support and model training.

[0203] Through the collaborative work of the above control parameter management process and state monitoring process, this subsystem realizes closed-loop control and intelligent management of the end-side load device, effectively improving the system response speed, data processing efficiency, and network resource utilization in complex satellite communication environments.

[0204] In summary, the application constructs a highly synergistic, efficient and reliable cloud edge integrated data transmission and management system through the organic combination of a series of key technical features. The system not only effectively solves the bottleneck problem of data transmission in the satellite communication environment, improves the depth and efficiency of data processing, enhances the stability and edge autonomy of the system, but more importantly, it lays a solid technical foundation for realizing complex spatial information services and intelligent applications, with significant technical progress and wide application prospects.

[0205] The present application proposes a unified data expression method for various types of satellite-oriented heterogeneous data (perception data, semantic data, geospatial data), and designs a data generation and update method combining current activity vector comparison and threshold triggering. These methods solve the problem of heterogeneous data being difficult to utilize uniformly, significantly reduce the data update frequency and data transmission overhead from the end device to the edge storage to the ground cloud, and at the same time take into account the semantic description of the data, which is particularly suitable for the characteristics of limited bandwidth of satellite-ground transmission link.

[0206] The application proposes a cloud edge collaborative data transmission and management subsystem that adapts to the characteristics of satellite data link and network. The method and system, including message bus mechanism and declarative interface service, are designed specifically for the characteristics of satellite data link, significantly improving the practicality of the method and system.

[0207] The application emphasizes the edge autonomy mechanism of data and application. Through edge metadata persistence (data catalog management, persistent configuration management), edge and side spatio-temporal underlying storage, local model and pattern storage, and task management, the application ensures offline load operation and fault recovery. Thus, the edge node can still operate efficiently and autonomously and has fast fault recovery capability when disconnected for a long time or the network deteriorates. Furthermore, the "task manager" and "data processor" on the edge side can support more complex autonomous decision-making based on local data and models.

[0208] The application is dedicated to solving how to realize remote precise closed-loop control of end-side load equipment and intelligently acquire its state, optimize response speed and data efficiency. In the "edge-end collaborative end-side load equipment control parameter management and state monitoring subsystem", the process of issuing control parameters through a declarative interface, executing them by the "device manager" on the edge side, and recording them in the "persistent configuration management" is elaborated, which guarantees the accuracy and recoverability of control. The state monitoring data is semantically converted on the edge side and intelligently returned on demand based on the strategy (stored in the "persistent configuration management"), forming an efficient closed loop. The concept of "device twin management" on the cloud side is introduced to cooperate with the "device manager" on the edge side to perform more detailed cloud mapping and management of the end-side load equipment.

[0209] The application discloses a satellite data abstraction and data collaboration mechanism based on three-level cooperation of cloud, edge and end:

[0210] A processing procedure of data abstraction in the direction of "end→edge→cloud" (from perception data record to semantic data record, and to global view of cloud) is proposed. The hierarchical abstraction mechanism, combined with the vectorized semantic data extraction method of the edge side, significantly reduces data redundancy, optimizes transmission efficiency on the bandwidth-limited link, and improves the application value of data.

[0211] A unified data expression method for heterogeneous data, namely, a perception data record unified expression method based on Schema+Value, is proposed, and corresponding distributed storage and management strategies (such as distribution based on end device identification and distribution based on space-time attribute) are designed for different types of data (perception data, semantic data and geographic spatial data).

[0212] The storage, transmission and processing pressure brought by massive raw perception data, and the problem that heterogeneous data is difficult to be uniformly utilized are solved.

[0213] The application discloses a semantic data record generation and update method based on vectorized expression:

[0214] A vector-based semantic data record expression method (vector=(f(t),t start ,t end )) is creatively designed, and an update mechanism combining current activity vector comparison and threshold triggering is adopted. This makes it possible to generate a new vector and upload it only when the data actually deviates from the model prediction, instead of uploading raw data in full or periodically.

[0215] The data update frequency and data volume overhead of the cloud platform are greatly reduced, which is particularly suitable for compression and efficient expression of numerical time series data, and also takes into account the semantic description of multimedia data.

[0216] The application discloses a cloud-edge collaborative data transmission and management subsystem architecture suitable for satellite data link and network characteristics:

[0217] A collaborative subsystem including a message bus mechanism, a declarative interface service and an edge autonomous mechanism is designed for the characteristics of satellite link, such as time continuity, time discontinuity, bandwidth limitation and high delay. In the architecture (such as Figure 3 ), the "real-time message engine" on the cloud side and the "message bus" on the edge side work collaboratively, guaranteeing the reliability and sequence of asynchronous communication; the standardized "interface service" and "declarative interface" simplify the interaction and tolerate delay; and the "metadata manager", "persistent configuration management" and "local storage" of the edge endow the edge node with high autonomous operation capability.

[0218] The vulnerability of traditional tight coupling or simple master-slave architecture under poor network conditions is overcome, and the robustness of the data link and the continuity of the edge service are ensured.

[0219] The present application realizes edge-end collaborative control parameter management and state monitoring:

[0220] A control system is formed, which is initiated by a user on the cloud side, accurately reaches the edge device through a declarative interface and an edge-side ''device manager'', and then forms a closed-loop control and monitoring system through state monitoring, data semantic conversion and on-demand intelligent feedback. The innovation points include local persistence of control parameters and on-demand feedback based on strategies.

[0221] Remote accurate control and intelligent state feedback of the edge device are realized, the efficiency of control instruction delivery and state data feedback is optimized, and an effective service closed loop is formed.

[0222] Although the specific implementation methods of the present application are described above, those skilled in the art should understand that these are only illustrative, and various changes or modifications can be made to these embodiments without departing from the principles and implementation of the present application, therefore, the protection scope of the present application is defined by the appended claims.

Claims

1. A cloud-edge-device collaborative satellite data management method, characterized in that, The method includes: The device acquires physical parameter signals at the edge and converts these signals into sensing data. The edge side processes the perceived data in conjunction with auxiliary data to obtain semantic data, and then stores the perceived data, the auxiliary data, and the semantic data in a distributed manner; wherein, the auxiliary data includes: geospatial data, global ionospheric model data, and target recognition model library data; The cloud-side management includes spatiotemporal big data, including the aforementioned semantic data, and performs unified fusion, expression, querying, and analysis processing on this spatiotemporal big data. The cloud side includes: interface service, device twin management module and real-time message engine; the edge side includes: data service, task manager, data catalog manager, device manager, data processor, message bus, model and pattern module, second cloud-edge-device collaborative management module and edge-side spatiotemporal underlying storage; and the device side includes: devices and applications. The process by which the endpoint, the edge, and the cloud work together includes: The cloud side uses interface services and a real-time message engine to send relevant instructions to the edge side through the first cloud-edge collaborative management module; wherein, the relevant instructions include: control instructions, task configuration, and model updates; After receiving the instruction, the second cloud-edge collaborative management module on the edge side schedules local resources to execute tasks through the task manager, and controls the devices on the edge side through the device manager and device exporter. After the device on the edge collects the data, it uploads the data to the message bus on the edge through the device exporter; After the edge-side data processor processes the data, it is uploaded to the cloud side via the second cloud-edge-device collaborative management module and data service, and then via the cloud-side real-time message engine. The cloud-side device twin management module, the edge-side device manager, and related storage modules work together to maintain the consistency of the device and application status on the edge side, so that when the network is interrupted, the edge side can operate autonomously based on the model and pattern module, the edge-side spatiotemporal underlying storage, and the data catalog manager. The control of the cloud-side load device to the end-side is achieved through the following steps: The user selects the target end-side load device and sets the control parameters to be adjusted and the target values ​​of those control parameters; The cloud side encapsulates the target-side load device, the control parameters that need to be adjusted, and the target value of the control parameters into a declarative instruction that conforms to a predefined specification, and then transmits the declarative instruction to the target edge side. The target edge parses the declarative instruction, identifies the target end-side load device, the control parameter that needs to be adjusted, and the target value of the control parameter, and encapsulates the control parameter that needs to be adjusted and the target value of the control parameter into a low-level instruction that can be recognized by the target end-side load device; After addressing and locating the target-side load device using the device manager, the target edge side sends the underlying instruction to the target-side load device and persists the parsing result of the declarative instruction locally on the target edge side.

2. The method according to claim 1, characterized in that, The sensing data includes: numerical sensing data and multimedia sensing data; The edge side combines auxiliary data to process the perceived data to obtain semantic data, including: Numerical sensing data within a time period is extracted in units of components to obtain a vector representation of the numerical sensing data; wherein, the components include: the pos component or the speed component of the spatial position sensor sample value; Intelligent processing is performed on multimedia-type perceptual data to obtain a semantic description of the multimedia-type perceptual data; wherein, the intelligent processing includes: orthorectification of image data, georegistration, and target recognition; Based on the sampling time and location of the sensing data, the vector representation of the numerical sensing data, the semantic description of the multimedia sensing data, and the identifier of the edge side, semantic data corresponding to the sensing data is generated.

3. The method according to claim 1, characterized in that, Distributed storage of the sensed data includes: The entire set of sensing data records is distributed and stored on various storage nodes according to a Hash, Round Robin, or regional distribution strategy; where sensing data containing file pointers is stored on the same storage node as the corresponding data file.

4. The method according to claim 1, characterized in that, Distributed storage of the geospatial data and the semantic data includes: For the geospatial and semantic data required for spatiotemporal computation, spatiotemporal query and spatiotemporal data access operations on the edge side, filter out geospatial and semantic data whose usage frequency is greater than the set value. A spatial or spatiotemporal data distribution algorithm is used to distribute the geospatial data and semantic data, and the geospatial data and semantic data whose usage frequency is greater than a set value are concentrated on a specific storage node.

5. The method according to claim 1, characterized in that, The cloud side also includes: a data cataloging management module, an edge management module, a command and pattern persistence module, a cloud-side spatiotemporal big data storage module, a cloud-side computing resource pool, and a first cloud-edge-device collaborative management module; The interface service is used to provide standardized API interfaces for edge-side or other authorized users to access cloud-side data resources and computing power, and to issue control commands. The data cataloging management module is used for unified metadata management, cataloging and indexing of spatiotemporal big data stored in the cloud, as well as metadata management and cataloging of data accessible at the edge. The edge-side management module is used for registering, monitoring, distributing configurations, and managing the accessed edge devices. The device twin management module is used to maintain a digital twin on the cloud side for each edge-side load device, and the digital twin synchronizes the status, attributes and configuration of the edge-side load device. The command and pattern persistence module is used to store commands, task configurations, data patterns, and models issued to the edge and endpoint sides. The cloud-side spatiotemporal big data storage module is used to store spatiotemporal big data and other business data from the edge and terminal sides. A cloud-side computing resource pool is used to provide computing resources to support model training and data analysis. The real-time messaging engine is used to process message queues for command issuance and data upload between the cloud and edge sides based on a message bus mechanism. The first cloud-edge-device collaborative management module is used to coordinate task allocation, data flow, and status synchronization among the cloud side, edge side, and device side.

6. The method according to claim 1, characterized in that, The edge side also includes: an edge side computing resource pool; The data service is used for data query processing, data subscription, and data transmission; The task manager is used to receive, schedule, and execute computing tasks and business logic defined from the cloud or local side. The data catalog manager is used to manage and catalog metadata for data that can be accessed by the edge side local storage, cloud side, and edge side. The device manager is used to manage the end-side connections via the device exporter; The data processor is used to perform preprocessing operations on raw data collected from the edge or data received from the cloud. The preprocessing operations include cleaning, transformation, aggregation, and feature extraction. The message bus is used for asynchronous communication between the internal modules on the edge side and with the end side. The model and pattern module is used to store model files, data processing patterns, and business rules locally on the edge side. The second cloud-edge-device collaborative management module is used for the collaborative management unit on the edge side, and cooperates with the corresponding module on the cloud side; The edge-side spatiotemporal underlying storage is used to persistently store raw data, processed data, application status, and necessary configuration information locally on the edge side. The edge computing resource pool is used to provide local computing resources.

7. The method according to claim 1, characterized in that, The device is used for data acquisition or receiving instructions to execute actions, and the device communicates with the device manager and message bus on the edge side through a device exporter; The application refers to logic that runs on smart devices or is directly controlled from the edge.

8. A cloud-edge-device collaborative satellite data management system, characterized in that, The system includes: On the end side, it is used to collect physical parameter signals and convert the physical parameter signals into sensing data; On the edge side, the system is used to process the perceived data in conjunction with auxiliary data to obtain semantic data, and to perform distributed storage of the perceived data, the auxiliary data, and the semantic data; wherein, the auxiliary data includes: geospatial data, global ionospheric model data, and target recognition model library data; On the cloud side, it is used to manage spatiotemporal big data, including the semantic data, and to perform unified fusion, expression, query and analysis processing of the spatiotemporal big data; The cloud side includes: interface service, device twin management module and real-time message engine; the edge side includes: data service, task manager, data catalog manager, device manager, data processor, message bus, model and pattern module, second cloud-edge-device collaborative management module and edge-side spatiotemporal underlying storage; and the device side includes: devices and applications. The process by which the endpoint, the edge, and the cloud work together includes: The cloud side uses interface services and a real-time message engine to send relevant instructions to the edge side through the first cloud-edge collaborative management module; wherein, the relevant instructions include: control instructions, task configuration, and model updates; After receiving the instruction, the second cloud-edge collaborative management module on the edge side schedules local resources to execute tasks through the task manager, and controls the devices on the edge side through the device manager and device exporter. After the device on the edge collects the data, it uploads the data to the message bus on the edge through the device exporter; After the edge-side data processor processes the data, it is uploaded to the cloud side via the second cloud-edge-device collaborative management module and data service, and then via the cloud-side real-time message engine. The cloud-side device twin management module, the edge-side device manager, and related storage modules work together to maintain the consistency of the device and application status on the edge side, so that when the network is interrupted, the edge side can operate autonomously based on the model and pattern module, the edge-side spatiotemporal underlying storage, and the data catalog manager. The control of the cloud-side load device to the end-side is achieved through the following steps: The user selects the target end-side load device and sets the control parameters to be adjusted and the target values ​​of those control parameters; The cloud side encapsulates the target-side load device, the control parameters that need to be adjusted, and the target value of the control parameters into a declarative instruction that conforms to a predefined specification, and then transmits the declarative instruction to the target edge side. The target edge parses the declarative instruction, identifies the target end-side load device, the control parameter that needs to be adjusted, and the target value of the control parameter, and encapsulates the control parameter that needs to be adjusted and the target value of the control parameter into a low-level instruction that can be recognized by the target end-side load device; After addressing and locating the target-side load device using the device manager, the target edge side sends the underlying instruction to the target-side load device and persists the parsing result of the declarative instruction locally on the target edge side.

Images