Satellite on-orbit sensing and processing system

Abstract

The invention provides a satellite on-orbit sensing and processing system, and belongs to the technical field of satellite on-orbit calculation. The system comprises an SDN controller, an SDN switch, a data system and a satellite platform management and control system. The SDN controller is configured to enable the response time of an on-orbit sensing and processing task to be the shortest by utilizing a hybrid processing distribution model according to the on-orbit sensing and processing task, meet memory and storage capacity constraint optimization and generate a sensing and task processing flow; obtaining the source and the destination of the sensing data and the processing process data corresponding to each time point and each step, forming a mesh topology relation, and generating network control parameters; and completing an on-orbit sensing and processing task based on the task processing flow and the network control parameters. According to the invention, through an on-satellite load, storage, computing power and network integrated architecture, the total time delay of data from acquisition and processing to transmission to the ground is reduced; and the SDN is applied to on-satellite data flow scheduling, so that dynamic optimization of a calculation path can be realized.

Description

Technical Field

[0001] This application belongs to the field of satellite on-orbit computing technology, and specifically relates to a satellite on-orbit sensing and processing system. Background Technology

[0002] In relevant satellite system architectures, the data processing mode can be referred to as the "store-and-forward" mode. The core process of this mode includes: after the satellite's sensors (e.g., optical cameras, SAR radar, hyperspectral imagers, etc.) collect raw data, the massive amounts of data are temporarily stored in the satellite's onboard memory with little or no preprocessing (e.g., compression, formatting). When the satellite passes over the ground receiving station, all the data is then downlinked to the ground data center. Finally, high-performance computing clusters and professionals on the ground perform refined data processing, analysis, and information extraction.

[0003] In related technologies, satellite data processing faces the following bottlenecks: 1) High latency, relying on ground stations to transmit data to ground data centers for processing, which cannot meet the real-time service needs of disaster monitoring, emergency response, etc.; 2) Poor flexibility, as related satellite system architectures are CPU / GPU-based, making it difficult to dynamically adapt to the complex spatiotemporal correlations of multi-source (e.g., remote sensing, IoT, positioning data) and multi-mode (e.g., visible light, infrared, hyperspectral, microwave) data; 3) Low resource utilization, with fixed hardware functions, making it impossible to dynamically reconstruct computing processes and network paths according to mission requirements, resulting in wasted computing power and bandwidth resources. Summary of the Invention

[0004] The purpose of this application is to provide a satellite on-orbit sensing and processing system, which aims to solve the problems existing in satellite data processing and satellite on-orbit computing.

[0005] This application provides a satellite on-orbit sensing and processing system, comprising an SDN controller, an SDN switch electrically connected to the SDN controller, a data system electrically connected to the SDN switch, and a satellite platform management and control system. The SDN controller is configured to: acquire on-orbit sensing and processing tasks received by the data system and / or the satellite platform management and control system, as well as task requirements and parameters sent by the satellite platform management and control system; based on the acquired on-orbit sensing and processing tasks, task requirements, and task parameters, utilize the hybrid processing allocation model in the SDN controller to minimize the response time of the on-orbit sensing and processing tasks while satisfying memory and storage capacity constraints, and generate a sensing and task processing flow; form a mesh topology based on the source and destination of sensing data and processing data corresponding to each time point and step of the sensing and task processing flow, and generate network control parameters based on the mesh topology; and control the SDN switch to implement a dynamic network topology based on the generated sensing and task processing flow and network control parameters, driving the data system to complete the on-orbit sensing and processing tasks.

[0006] In an optional implementation, the SDN controller includes: an onboard SDN control plane module for managing the data flow path of the sensing and task processing process; an onboard SDN application plane module for managing the sensing data corresponding to each time point and each step of the sensing and task processing process, and for optimizing the sensing and task processing process; and a network engine for providing timing, synchronization, and triggering functions for each module of the satellite on-orbit sensing and processing system.

[0007] In an optional implementation, the SDN switch includes an SDN data plane layer module implemented using a programmable switch. The SDN switch is used to: provide corresponding data interfaces according to the type of the data system; and dynamically divide virtual network slices to respectively carry the sensing data and processing data corresponding to each time point and each step of the sensing and task processing flow.

[0008] In an optional implementation, the data system includes a satellite remote sensing payload, a navigation payload, a computing and processing module, a storage device, and an inter-satellite and inter-ground signal processing and conversion module.

[0009] In an optional implementation, the satellite platform management system includes: a satellite mission computer for managing the satellite's on-orbit sensing and processing system and satellite application mission management; and telemetry and control equipment for serving as a carrier for satellite remote control and telemetry functions.

[0010] In an optional implementation, the hybrid processing allocation model includes the correlation between each time point and step in the on-orbit sensing and processing task and the required resource consumption data. The hybrid processing allocation model in the SDN controller is used to minimize the response time of the on-orbit sensing and processing task and meet memory and storage capacity constraints to generate a sensing and task processing flow. This includes: taking the computation time overhead and / or the time consumed by data supplemented from ground equipment, sensing data, data transmission and transfer between equipment and modules, and the corresponding constraints as optimization objectives; dynamically allocating the computing resources of the satellite on-orbit sensing and processing system based on the spatiotemporal correlation of data supplemented from ground equipment and sensing data; and using a preset task scheduling algorithm to minimize the response time of the on-orbit sensing and processing task.

[0011] In an optional implementation, the process of forming a mesh topology and generating network control parameters based on the mesh topology includes: forming a network topology for on-orbit sensing and processing tasks; and switching the network topology based on the timing of the sensing data and processing data corresponding to each time point and step of the sensing and task processing flow, either on a time basis or based on network events.

[0012] In an optional implementation, the SDN controller is further configured to: unify the spatiotemporal reference of the sensing data and processing data corresponding to each step through spatiotemporal tags.

[0013] In an optional implementation, completing the on-orbit sensing and processing task includes: calling the SDN switch and the data system to perform the on-orbit sensing and processing task; and recording relevant data such as the time consumption, memory usage, intermediate storage, and inter-module communication overhead of each step of the on-orbit sensing and processing task.

[0014] In an optional implementation, after the on-orbit sensing and processing task is completed, the SDN controller is further configured to: receive the calculation results from the SDN switch and the data system, and the task requirements for the next stage determined based on the calculation results; and optimize the hybrid processing allocation model based on the recorded data related to the time consumption, memory usage, intermediate storage, and inter-module communication overhead of each step of the on-orbit sensing and processing task.

[0015] Through the above technical solutions, the SDN-based satellite on-orbit sensing and processing system provided in this application embodiment, through the integrated architecture of on-board payload, storage, computing power and network, can reduce the total time delay of data from acquisition, processing to transmission to the ground, and improve the real-time performance of satellite on-orbit sensing and processing; applying SDN to on-board data stream scheduling can realize dynamic optimization of the sensing and task processing process path; supporting multi-level dynamic configuration from the underlying hardware to the upper-level algorithms, from data acquisition, data transmission, distribution path to the reconstruction of processing algorithms at each step, meeting the diverse needs of tasks, and enabling agile algorithm reconstruction.

[0016] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures and processes shown in the description and the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of the satellite on-orbit sensing and processing system provided in the exemplary embodiment of this application. Figure 1 .

[0019] Figure 2 This is a schematic diagram of the workflow of the SDN controller provided in an exemplary embodiment of this application.

[0020] Figure 3 This is a schematic diagram of the structure of the satellite on-orbit sensing and processing system provided in the exemplary embodiment of this application. Figure 2 .

[0021] Figure 4 This is a schematic diagram of the network topology provided in an exemplary embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In related technologies, satellite computing architectures (e.g., using CPU+GPU heterogeneous computing units), while possessing 40 TOPS of computing power, rely on fixed hardware architectures, making it difficult to support dynamic algorithm reconstruction. SDN-based terrestrial network load balancing technology has achieved dynamic traffic allocation, but it only addresses optimization issues in the transmission phase, failing to optimize for multi-source data acquisition and spatiotemporal alignment characteristics, nor achieving balanced optimization of computing load. While the Tiansuan satellite achieves onboard AI processing, it has not solved problems related to sensing payloads, onboard storage, computing power, network, and inter-satellite / ground-satellite collaborative scheduling; furthermore, data flow still relies on preset paths, unable to dynamically adapt to mission changes. Therefore, this application provides an on-orbit sensing and processing system for this satellite, a satellite intelligent computing system based on Software Defined Network (SDN).

[0024] Figure 1 This is a schematic diagram of the structure of a satellite on-orbit sensing and processing system provided in an exemplary embodiment of this application. The satellite on-orbit sensing and processing system may include an SDN controller 10, an SDN switch 20 electrically connected to the SDN controller 10, a data system 30 electrically connected to the SDN switch 20, and a satellite platform management and control system 40. Please refer to... Figure 2 The SDN controller can be configured as follows: Step S210: Obtain the on-orbit sensing and processing tasks received by the data system and / or satellite platform control system, as well as the task requirements and task parameters sent by the satellite platform control system.

[0025] Step S220: Based on the acquired on-orbit sensing and processing tasks, task requirements, and task parameters, the hybrid processing allocation model in the SDN controller is used to minimize the response time of the on-orbit sensing and processing tasks and meet memory and storage capacity constraints to generate the sensing and task processing flow.

[0026] Step S230: Based on the source and destination of the sensing data and processing data corresponding to each time point and step in the sensing and task processing flow, a mesh topology is formed, and network control parameters are generated based on the mesh topology.

[0027] Step S240: Based on the generated sensing and task processing flow and network control parameters, control the SDN switch to realize dynamic network topology and drive the data system to complete on-orbit sensing and processing tasks.

[0028] The SDN-based satellite on-orbit sensing and processing system provided in this application, through its integrated architecture of on-board payload, storage, computing power, and network, can reduce the total time delay from data acquisition and processing to transmission to the ground, thereby improving the real-time performance of satellite on-orbit sensing and processing. By applying SDN to on-board data stream scheduling, it can achieve dynamic optimization of the sensing and task processing workflow. It supports multi-level dynamic configuration from the underlying hardware to the upper-level algorithms, and the reconstruction of data acquisition, data transmission, distribution paths, and processing algorithms at each step, meeting the diverse needs of tasks and enabling agile algorithm reconstruction.

[0029] Please refer to Figure 3 The preferred SDN controller in this application embodiment may include: an onboard SDN control plane module for managing the data flow path of the sensing and task processing process; an onboard SDN application plane module for managing the sensing data and processing data corresponding to each time point and each step of the sensing and task processing process, and also for optimizing the sensing and task processing process; a network engine (not shown in the figure) for providing timing, synchronization and triggering functions for each module of the satellite on-orbit sensing and processing system; and a spatiotemporal tagging engine (not shown in the figure) for setting a spatiotemporal reference for the sensing data corresponding to each time point and each step.

[0030] For example, the onboard SDN control plane deployed by the SDN controller can centrally manage data flow paths; the deployed onboard SDN application plane can manage the acquisition and processing of sensing data and processing data, manage transmission processes and path optimization, etc.; the integrated real-time network engine can provide time synchronization, triggering and other functional support for payloads and various modules; the integrated spatiotemporal tagging engine can add high-precision spatiotemporal references to multi-source data (e.g., remote sensing images, IoT sensor data, etc.).

[0031] Please refer to Figure 3 The preferred SDN switch in this application embodiment may include an SDN data plane layer module implemented using a programmable switch. The SDN switch can be used to: provide corresponding data interfaces according to the type of data system; and dynamically divide virtual network slices to carry the sensing data and processing data corresponding to each time point and each step of the sensing and task processing process.

[0032] For example, SDN switches are configured to provide corresponding data interfaces based on the type and implementation of the data system (which may include data devices or modules) (e.g., standalone unit, single board shared with the switch, module on the same circuit board, etc.). Data interfaces may include Ethernet interfaces, inter-board bus interfaces, or intra-board bus interfaces between the SDN switch and the data device. SDN switches are also configured to dynamically divide virtual network slices to separately carry the sensing data and processing data of the data system, such as remote sensing data streams, IoT data streams, and navigation data streams.

[0033] Please refer to Figure 3 The preferred data system in this application embodiment may include a satellite remote sensing payload, a navigation payload, a computing and processing module, a storage device, and an inter-satellite and inter-ground signal processing and conversion module, etc.

[0034] In the payloads and processing modules included in the data system, satellite remote sensing payloads can acquire ground feature information according to instructions using technologies such as optics and electromagnetic waves. Satellite remote sensing payloads may include data source devices such as visible light cameras, infrared cameras, hyperspectral imagers, and synthetic aperture radar (SAR). For navigation payloads, signals and data from navigation satellites (e.g., BeiDou, GPS) are acquired in orbit and sent to the navigation augmentation center as observation data for the navigation augmentation system (for calculation). Simultaneously, the navigation augmentation correction data uploaded by the navigation augmentation center is used for in-orbit calculation to determine the satellite's position, velocity, and time synchronization. The computational processing module may include data and signal processing modules such as CPUs, FPGAs, GPUs, and NPUs. It processes the incoming data according to instructions and outputs the processing results through an interface, while also generating processing log information. Storage devices are used to store incoming sensing data, processing data, and processing results data, and to read and output corresponding data according to instructions, including sensing data, processing data, processing results data, model data, and algorithm libraries. The inter-satellite-to-ground signal processing and conversion module can perform analog-to-digital or digital-to-analog conversion, modulation and demodulation, encoding and decoding of inter-satellite or satellite-to-ground communication signals. As a radio frequency or laser front end, it works with computing processing modules and SDN equipment to form software-defined radio or laser communication capabilities. The processing algorithm can also be reconfigured on orbit.

[0035] In a preferred embodiment of this application, the payload and processing modules included in the data system can also be configured to support dynamic loading and unloading of algorithms; and to automatically call pre-trained models or user-defined algorithms based on task requirements.

[0036] Please refer to Figure 3 The preferred satellite platform management system in this application embodiment may include: a satellite mission computer for managing the satellite's on-orbit sensing and processing system and satellite application mission management; and telemetry and control equipment for serving as a carrier for satellite remote control and telemetry functions.

[0037] In this embodiment, after the satellite's on-orbit sensing and processing system is built, algorithm uploading is performed. For example, the ground system adds or updates intelligent processing software such as algorithms, logic, and models to each processing module of the satellite's on-orbit sensing and processing system via the uploading link. The intelligent processing software may include software modules configured in the SDN controller, firmware in the SDN switch, processor software and FPGA logic in each payload, software in the CPU processing module, software and models in the GPU and Neural-network Processing Unit (NPU), logic in the FPGA processing module, firmware and logic in the communication signal transceiver or conversion module, etc. Different versions of the intelligent processing software can also be stored in a storage device and loaded or reconstructed into the corresponding payload or module as needed.

[0038] In step S210, the on-orbit sensing and processing tasks are received through the telemetry and control equipment and the inter-satellite-to-ground signal processing and conversion module. The satellite mission computer of the satellite platform management system sends the task requirements and parameters to the SDN controller. The on-orbit sensing and processing tasks may include sensing data collected or queried from ground equipment using the inter-satellite-to-ground signal processing and conversion module; autonomously performing remote sensing payload observations, on-orbit calculations, and intelligent processing according to rules or intermediate results of on-orbit processing; and sending the processing results to the ground control center, ground terminals, ground actuators (e.g., electrically controlled valves), ground autonomous intelligent devices (e.g., robots, drones), and other networked satellites to execute disposal tasks or subsequent tasks. On-orbit sensing and processing tasks can be tasks to be executed immediately, delayed tasks to be executed in the future, or periodically executed tasks.

[0039] In a preferred embodiment of this application, the hybrid processing allocation model may include the correlation between each time point and step in the on-orbit sensing and processing task and the required resource consumption data. Utilizing the hybrid processing allocation model on the SDN switch, the response time of the on-orbit sensing and processing task is minimized while satisfying memory and storage capacity constraints. The resulting sensing and task processing flow may include: using computational time overhead and / or the time consumed by data supplemented from ground equipment, sensing data, and the time consumed by data transmission and transfer between devices and modules, along with corresponding constraints, as optimization objectives; dynamically allocating the computing resources of the satellite's on-orbit sensing and processing system based on the spatiotemporal correlation of data supplemented from ground equipment and sensing data; and employing a preset task scheduling algorithm to minimize the response time of the on-orbit sensing and processing task.

[0040] For example, the SDN controller is configured to generate a sensing and task processing flow based on a hybrid processing allocation model, according to the task requirements, parameters, and types of on-orbit sensing and processing tasks, with the shortest processing response time (e.g., including data acquisition and transmission time) and while satisfying memory and storage capacity constraints. In this embodiment, the hybrid processing allocation model may include resource usage data such as the time, transmission bandwidth, and storage overhead required for each sensing, transmission, and processing step in the on-orbit sensing and processing task, as well as resource relationships (e.g., the relationship between computing power and computation time, which can be approximated using a polynomial). Resource usage data can be obtained from theoretical calculations (e.g., remote sensing push-broom time, algorithm complexity, etc.) or experimental measurements, and corrected using measured data during actual operation.

[0041] In this embodiment, the optimization objective of minimizing response time while satisfying memory and storage capacity constraints may include computation time overhead and / or the time consumed by acquiring data from ground equipment, acquiring data through remote sensing equipment, and the time and corresponding constraints for data transmission and transfer between devices and modules. The optimization process can dynamically allocate computing resources based on the spatiotemporal correlation of data (e.g., the correspondence between geographic coordinates of ground data and remote sensing data, temporal alignment characteristics of multi-source data, etc.). This dynamic allocation of computing resources may include: based on the continuous changes in data volume, data correlation, computing power requirements, and sensing action time at each time point and step during the execution of the on-orbit sensing and processing task, the SDN controller dynamically adjusts resource allocation according to the changes in required computing resources at each time point. For example, based on the relationship between computing power and computation time of parallel processing steps at a certain time point, a resource allocation scheme that allows each parallel processing step to be completed simultaneously or approximately simultaneously can be calculated. The calculation algorithm is to simultaneously solve the aforementioned resource correlation polynomials.

[0042] In this embodiment, a preset task scheduling algorithm (e.g., a task scheduling algorithm based on a genetic algorithm, a heuristic task scheduling algorithm, or a task scheduling algorithm based on an annealing algorithm) can be used to minimize the response time of on-orbit sensing and processing tasks. The boundary conditions and optimization objectives differ from parallel computing task scheduling strategies. Factors such as the phased imaging time of the remote sensing payload and the time series of ground sensor data collected by the IoT payload are also used as optimization calculation variables to achieve real-time sensing and rapid response from the satellite.

[0043] Taking the optimization of a task scheduling algorithm based on genetic algorithms as an example, the training process of the hybrid processing allocation model is as follows: 1) Generate the input dataset, which may include: a list of task steps and the dependencies between steps; resource requirements for each step (time, bandwidth, storage, etc.); the relationship between computing power and computing time (e.g., the aforementioned polynomial coefficients); resource boundary conditions (e.g., total computing power, total bandwidth, total storage, etc.); genetic algorithm parameters (e.g., population size, crossover rate, mutation rate, termination conditions, etc.).

[0044] 2) Initializing the population may include generating random genes to ensure that step dependencies are satisfied and resource allocation is within a reasonable range. The genes need to represent the task processing flow, which may include the step sequence, resource allocation, and triggering decisions. In this embodiment, a hybrid encoding based on sequences and parameters can be used. For example, the gene structure: each gene corresponds to a step and may include the following information: step type (e.g., sensing, transmission, processing, etc.); device identifier used (e.g., corresponding sensing device, corresponding transmission path, corresponding processing device, etc.); resource allocation amount (e.g., computing power allocation value, bandwidth allocation value, etc.); for processing steps, it also includes a decision flag (Boolean value) for whether to trigger supplementary sensing.

[0045] Example encoding: Gene = [Gene1, Gene2, ...], where Gene1 = {Type: Sensing, Device: Remote Sensing Camera 1, Resource: None}, Gene2 = {Type: Transmission, Path: Link1, Bandwidth: 3Gbps}, Gene3 = {Type: Processing, Device: GPU1, Computing Power: 5U, Trigger Flag: True}.

[0046] 3) Evaluate fitness: For each gene, run the simulator to calculate the total task time, considering resource constraints and triggering events; record the fitness value. The fitness value can be the total time (the shortest response time for on-orbit task processing). If resource constraints are violated, a penalty can be added, for example: Fitness = Total Time + α × Resource Excess (α is a large constant).

[0047] 4) Evolutionary cycles. For example: While (termination condition not met): Selection: Select parent genes from the current population; Crossover: Performs a crossover operation on the selected parent generation to generate offspring; Mutation: Performing mutation operations on offspring; Evaluate offspring fitness: Run the simulator to calculate the fitness of the offspring; Forming a new population: Merge offspring and parents (or use an elite preservation strategy) to generate a new population.

[0048] End While 5) Obtain the planning results: Return the optimal gene (i.e., the perception and task processing flow, which may include serial and parallel perception and data computation steps) and its fitness value (the shortest response time for on-orbit perception and processing tasks).

[0049] In this embodiment, the aforementioned optimization calculations can be rapidly completed by the SDN controller using the computing power of the processing module in the data system connected to the SDN switch. Depending on the characteristics of the algorithm, GPUs, NPUs, FPGAs, etc., can be used, or a combination thereof can be employed. For example, the time-parallel computing portion of the planning algorithm can be processed using FPGA logic.

[0050] The SDN-based satellite on-orbit sensing and processing system provided in this application applies SDN to on-board data stream scheduling to achieve dynamic optimization of the sensing and task processing flow. Optimization objectives can include not only computation time overhead, but also the time and corresponding constraints consumed in acquiring data from ground equipment, acquiring data through remote sensing equipment, and the transmission and transfer of data between devices and modules, thereby achieving SDN-driven sensor-memory-computing-network collaboration.

[0051] In a preferred embodiment of this application, forming a mesh topology and generating network control parameters based on the topology may include: forming a network topology for on-orbit sensing and processing tasks; and switching the timing of the network topology based on the source and destination of the sensing data and processing data corresponding to each time point and step in the sensing and task processing flow, either on a time basis or based on network events, thereby driving the switching of the network topology.

[0052] In this embodiment, the SDN controller can also be configured to extract the data sources and destinations at each time point and step according to the sensing and task processing flow, forming a network topology relationship by connecting each pair of sources and destinations; and generating network control parameters based on the network topology relationship. For example, at a certain time point in the on-orbit sensing and processing task, the remote sensing imaging data source is the camera output interface; and at the same time point, the destinations are storage devices, GPU processing units, and CPU processing units, respectively. Therefore, network control parameters that satisfy the above one-to-three network topology relationship can be generated; and processing messages can be sent to the payload and each processing module to enable corresponding operations and algorithms.

[0053] In this embodiment, network control parameters can be configured to minimize transmission time by comprehensively employing a combination of point-to-point, broadcast, and multicast transmission methods based on the many-to-many relationships between the sources and destinations of data in each step of the on-orbit sensing and processing task. This allows SDN to be flexibly integrated into intelligent processing.

[0054] In this application embodiment, the network topology may include the network of the satellite on-orbit sensing and processing system, as well as the inter-satellite network with other networked satellites and the satellite-to-ground network with ground terminals. An example network topology is shown below. Figure 4 As shown in the diagram. 1.1, 1.2, and 1.3 represent the paths from which the sensing data of remote sensing payload 1 is simultaneously sent to the CPU processing module, GPU processing module, and storage device; 2 represents the path from which intermediate results from GPU processing are sent to the CPU; 3.1 and 3.2 represent the paths from which supplementary sensing information and inter-satellite coordination information generated by the CPU are sent to the satellite's onboard computer and FPGA processing module, respectively; 4.1 and 4.2 represent the paths from which communication data generated by the FPGA processing module is sent to the satellite-to-ground communication link and the inter-satellite communication link, respectively; 5 represents the path from which supplementary sensing mission information generated by the onboard computer is sent to remote sensing payload 2; 6 represents the path from which ground data signals retrieved via satellite-to-ground communication are sent to the FPGA processing module; and 7.1, 7.2, and 7.3 represent the paths from which the sensing data of remote sensing payload 2 is simultaneously sent to the CPU processing module, GPU processing module, and storage device.

[0055] In this embodiment, the SDN switch uses the SDN data plane layer module to form the (initial) network topology relationship for the on-orbit sensing and processing task; the SDN controller switches the network topology relationship according to the network topology relationship switching sequence of this task stage, either at a predetermined time point or based on network event-driven switching. For example, in the network topology relationship shown in Figure (4), at the time point when the sensing action is initiated, the SDN controller configures the remote sensing payload 1 to simultaneously access the multicast network channels of the CPU processing module, GPU processing module, and storage device. The data transmission completion event in path 5 triggers the SDN controller to configure the remote sensing payload 2 to simultaneously access the multicast network channels of the CPU processing module, GPU processing module, and storage device.

[0056] For example, sensing data can be acquired from various devices in the data system according to the initial network topology. Processing and computation are performed by calling the corresponding software in each processing module in a hybrid serial-parallel manner, recording the computation time, memory usage, intermediate storage, and inter-module communication overhead for each processing step. The sensing data can include data collected or queried from ground devices using the satellite-to-ground communication link, data acquired through autonomous remote sensing payload observations following rules or intermediate results processed in orbit, and data actively transmitted from the ground or other networked satellites. The hybrid serial-parallel processing and computation by calling the corresponding software in each processing module can include: executing the processor software and FPGA logic in each payload, the software in the CPU processing module, the software and models in the GPU and NPU processing modules, and the logic in the FPGA processing module. Based on the generated sensing and task processing flow, each device can execute the data serially and in parallel. The synchronization and triggering operations required in the sensing and task processing flow are planned using the network's Time-Related Processing (TSN) capabilities.

[0057] The preferred SDN controller in this application embodiment can also be configured to: unify the spatiotemporal reference of the sensing data corresponding to the steps through spatiotemporal tags.

[0058] Following the example above, Time-Sensitive Networking (TSN) technology can be used to synchronize the internal clocks of various modules, payloads, and devices within the system. For instance, TSN technology can trigger corresponding actions at precise times, such as initiating remote sensing imaging or acquiring ground sensor data. Satellites utilize precise orbit and attitude determination technology to determine spatial parameters such as orbital position, satellite attitude, and the Earth's surface they are pointing to. Ground terminals use navigation enhancement technology to obtain their location coordinates. The data acquired or processed by each module, payload, and device is simultaneously labeled with its acquisition timescale, the extrapolated or interpolated timescale of the processing, and the spatial coordinates of the source and the extrapolated or interpolated spatial coordinates of the processing, forming spatiotemporal labels. During data processing, the data is selected according to its spatiotemporal labels and aligned using extrapolation or interpolation methods.

[0059] In this embodiment, the SDN controller is used to dynamically plan the data flow path, which can adapt to complex spatiotemporal relationships (e.g., the spatiotemporal relationship and processing sequence relationship between ground sensor locations, values ​​and remote sensing images), and can achieve efficient organization of multi-source data.

[0060] In a preferred embodiment of this application, completing the on-orbit sensing and processing task may include: calling the SDN switch and data system to perform on-orbit processing task calculations; and recording data related to the computation time, memory, intermediate storage, and inter-module communication overhead of each calculation step of the on-orbit processing task calculation.

[0061] In a preferred embodiment of this application, after completing the on-orbit sensing and processing task, the SDN controller can also be configured to: receive the calculation results of the SDN switch and the data system and the task requirements for the next stage determined based on the calculation results; and optimize the hybrid processing allocation model based on the recorded data related to the computation time, memory, intermediate storage and inter-module communication overhead of the steps of the on-orbit sensing and processing task.

[0062] For example, the system receives the calculation results from the SDN switch and data system, and based on these results, determines the task requirements for the next stage, including the on-orbit sensing and processing tasks for that stage. These on-orbit sensing and processing tasks can include tasks completed or collaboratively completed by the satellite's on-orbit sensing and processing system, other networked satellites, ground equipment, or personnel. When invoking the SDN switch and data system for on-orbit sensing and processing tasks, the system records data related to the computation time, memory usage, intermediate storage, and inter-module communication overhead for each calculation step of the on-orbit sensing and processing task to optimize the hybrid processing allocation model.

[0063] The SDN-based satellite on-orbit sensing and processing system provided in this application, through its integrated architecture of onboard payload, storage, computing power, and network, can reduce the total time delay from data acquisition and processing to transmission to the ground, thereby improving the real-time performance of satellite on-orbit sensing and processing. Applying SDN to onboard data stream scheduling enables dynamic optimization of the sensing and task processing workflow. It supports multi-layered dynamic configuration from underlying hardware to upper-layer algorithms, and the reconfiguration of data acquisition, data transmission, distribution paths, and processing algorithms at each step, meeting diverse task requirements and enabling agile algorithm reconfiguration. The technical advantages achieved by this application embodiment may also include: 1) Apply SDN to on-board data stream scheduling to achieve dynamic optimization of sensing and task processing workflows. Optimization objectives can include not only computation time overhead, but also the time and corresponding constraints consumed in acquiring data from ground equipment, acquiring data through remote sensing equipment, and the transmission and storage of data between equipment and modules, thereby achieving SDN-driven sensor-memory-computing-network collaboration.

[0064] 2) Based on the characteristics of satellite sensing and intelligent processing, a combination of point-to-point, broadcast, and multicast transmission methods (i.e., combined topology mode) is adopted in the computing network topology to achieve the shortest response time for on-orbit sensing and processing tasks.

[0065] 3) The scope of on-orbit sensing and processing task optimization includes data collection or querying from ground equipment using the satellite-to-ground communication link, autonomous execution of remote sensing payload observations according to rules or intermediate results processed in orbit, computation and intelligent processing by multiple on-orbit processing modules, and transmission of processing results to ground operation and control centers, ground terminals, ground actuators (e.g., electrically controlled valves), ground autonomous intelligent devices (robots, drones, etc.), and collaborative processing with other networked satellites. This greatly expands the scope of on-orbit intelligent computing for satellites.

[0066] 4) Apply the synchronization and triggering capabilities of Time-Sensitive Networking (TSN) to the scheduling of intelligent computing processes.

[0067] 5) Communication methods in SDN can include Ethernet, high-speed serial bus between boards, inter-satellite laser communication, and satellite-to-ground wireless communication, supporting intelligent computing in collaboration between constellations and large-scale space-ground systems.

[0068] 6) Supports full-stack programmability from the physical layer to the application layer, enabling multi-level algorithm reconfiguration and significantly improving flexibility.

[0069] Taking forest fire monitoring as an example, the workflow of the SDN-based satellite on-orbit sensing and processing system provided in this application embodiment is explained as follows: S11, Event Triggered. Ground sensors or satellite wide-field cameras detect suspected forest fire spots on the ground, and onboard early warning software triggers the forest fire perception and intelligent processing task (i.e., on-orbit perception and processing task).

[0070] S12, Sensing and Processing Mission Initiated. The satellite service will send the forest fire sensing and intelligent processing mission information (i.e., on-orbit sensing and processing mission, mission requirements, and mission parameters) to the SDN controller.

[0071] S13, Task Planning (i.e., generating the perception and task processing flow). The SDN controller forms the task plan, which may include supplementing the collection of surrounding sensor data, activating the infrared camera to image the target area, loading the APP in the GPU and NPU to preprocess the image, fire identification and location, and forming the network topology and switching sequence in this task process (i.e., generating network control parameters).

[0072] S14, on-orbit sensing and processing mission execution. The SDN switch dynamically constructs the network topology; the satellite-to-ground signal transceiver processing module, FPGA processing module, and channel data processing software in the GPU constitute the satellite-based IoT to collect sensor data around the target area, and the infrared camera remote sensing payload images the target area; the GPU and NPU analyze the sensed data to identify and locate the fire, generate alarm information and send it to the ground terminal and forward it to the ground center through inter-satellite communication, generate follow-up tasks for subsequent satellites and send them to the corresponding network satellites.

[0073] S15, Algorithm Reconstruction: Based on the mission execution process records, the on-orbit hybrid processing allocation model is updated, the sensing data is transmitted to the ground center to further train the fire situation processing and analysis model, and the training results are updated on the satellite.

[0074] Accordingly, the SDN-based satellite on-orbit sensing and processing system provided in this application can significantly improve response time and reduce latency through on-board real-time processing, serial-parallel hybrid processing, and space-ground collaborative processing; resource utilization will be significantly improved by dynamically allocating transmission bandwidth, computing power, and storage access capabilities based on SDN; and multi-source data fusion will aggregate and collaboratively process multi-source and multi-mode data from multiple satellites and ground sensors according to spatiotemporal relationships.

[0075] It is understood that the circuit structures, names, and parameters described in the above embodiments are merely examples. Those skilled in the art can also make readily conceived combinations and adjustments to the structural features of the above embodiments according to their needs, and the concept of this application should not be limited to the specific details of the above examples.

[0076] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A satellite on-orbit sensing and processing system, characterized by, The satellite on-orbit sensing and processing system comprises an SDN controller, an SDN switch electrically connected with the SDN controller, a data system electrically connected with the SDN switch, and a satellite platform management system, the SDN controller is configured to: acquire on-orbit sensing and processing tasks received by the data system and / or the satellite platform management system, task requirements and task parameters sent by the satellite platform management system; according to the acquired on-orbit sensing and processing tasks, task requirements and task parameters, utilize a hybrid processing allocation model in the SDN controller to make the response time of the on-orbit sensing and processing tasks shortest and meet the memory and storage capacity constraint optimization, and generate a sensing and task processing flow; according to the source and destination of the sensing data and processing process data corresponding to each time point and each step of the sensing and task processing flow, form a mesh topology relationship, and based on the mesh topology relationship, generate network control parameters; and based on the generated sensing and task processing flow and network control parameters, control the SDN switch to realize a dynamic network topology, and drive the data system to complete the on-orbit sensing and processing tasks.

2. The satellite on-board sensing and processing system of claim 1, wherein, The SDN controller comprises: an on-board SDN control plane module for managing a data flow path of the sensing and task processing flow; an on-board SDN application plane module for managing sensing data corresponding to each time point and each step of the sensing and task processing flow, and for optimizing the sensing and task processing flow; and a network engine for providing time system, synchronization and triggering functions for each module of the satellite on-orbit sensing and processing system.

3. The satellite on-board sensing and processing system of claim 1, wherein, The SDN switch comprises an SDN data plane layer module realized by a programmable switch, and the SDN switch is configured to: provide corresponding data interfaces according to the type of the data system; and dynamically divide virtual network slices to respectively carry the sensing data and processing process data corresponding to each time point and each step of the sensing and task processing flow.

4. The satellite on-board sensing and processing system of claim 1, wherein, The data system comprises a satellite remote sensing payload, a navigation payload, a computing processing module, a storage device, and an inter-satellite and satellite-signal processing and conversion module.

5. The satellite on-board sensing and processing system of claim 1, wherein, The satellite platform management system comprises: a satellite service computer for managing the satellite on-orbit sensing and processing system and satellite application task management; and a TT&C device for serving as a carrier of satellite remote control and telemetry functions.

6. The satellite on-board sensing and processing system of claim 1, wherein, The hybrid processing allocation model comprises an association relationship between each time point and each step in the on-orbit sensing and processing tasks and required resource occupation data, and the utilization of the hybrid processing allocation model in the SDN controller to make the response time of the on-orbit sensing and processing tasks shortest and meet the memory and storage capacity constraint optimization to generate the sensing and task processing flow comprises: taking the computing time overhead and / or data supplemented from ground equipment, sensing data, time and corresponding constraints consumed by data transmission and storage between devices and modules as optimization objectives; dynamically allocating computing resources of the satellite on-orbit sensing and processing system based on the spatio-temporal correlation of the data supplemented from the ground equipment and the sensing data; and A preset task scheduling algorithm is adopted to shorten the response time of the in-orbit sensing and processing task.

7. The satellite on-board sensing and processing system of claim 1, wherein, The network topology relationship is formed, and network control parameters are generated based on the network topology relationship, including: The network topology relationship of the in-orbit sensing and processing task is formed; and According to the sensing data and processing procedure data corresponding to each time point and each step of the sensing and processing procedure, the timing of the switching of the network topology relationship is switched according to the network event, and the switching of the network topology relationship is driven.

8. The satellite on-board sensing and processing system of claim 1, wherein, The SDN controller is further configured to: Through the space-time label, the space-time reference of the sensing data and processing procedure data corresponding to each step is unified.

9. The satellite on-board sensing and processing system of claim 1, wherein, The in-orbit sensing and processing task includes: The SDN switch and the data system are called to perform the in-orbit sensing and processing task; and The time consumption, memory occupation, intermediate storage, and inter-module communication overhead related data of each step of the in-orbit sensing and processing task are recorded.

10. The satellite on-board sensing and processing system of claim 9, wherein, After the in-orbit sensing and processing task is completed, the SDN controller is further configured to: Receive the calculation results of the SDN switch and the data system, and the determined task requirements of the next stage according to the calculation results; And According to the recorded time consumption, memory occupation, intermediate storage, and inter-module communication overhead related data of each step of the in-orbit sensing and processing task, the hybrid processing allocation model is optimized.

Images