Medium-orbit space-based management and control system based on edge side large model

Abstract

The invention relates to a medium orbit space-based management and control system based on an edge side large model, and belongs to the technical field of aerospace, and the system comprises a physical bearing layer, an intelligent service layer, a cooperative control layer and a user service layer. The invention provides application of an edge-side large model in a medium-orbit satellite: aiming at limited computing resources of the satellite, a lightweight large model is deployed, local task evaluation and decision are realized, and high time delay caused by ground transfer is avoided; according to the multi-satellite cooperation mechanism based on edge large model state awareness, satellite capabilities (batteries, loads and bandwidths) are evaluated through an edge side large model, dynamic allocation and cooperative processing of tasks are achieved, and the overall constellation efficiency is improved; according to the intelligent service unit design oriented to the middle track application, an integrated management unit not less than a sensing module, an edge large model module, a decision execution module, a communication management module, a storage module and the like is designed, and the end-to-end large model service capability oriented to the space-based application is constructed.

Description

Technical Field

[0001] This disclosure belongs to the field of aerospace technology, and in particular relates to a medium-orbit space-based control system based on a large edge-side model. Background Technology

[0002] To manage and control the tens of thousands of communication, remote sensing, navigation, and other types of satellites in the future, the following challenges need to be addressed: 1) The necessity of overcoming the inherent defects of ground-based telemetry and control networks and achieving seamless global coverage. (1) Coverage and Real-time Bottlenecks: Traditional ground-based telemetry and control stations are limited by the curvature of the Earth and national borders, resulting in a large number of coverage blind spots (such as oceans, polar regions, and overseas areas). Low-Earth orbit satellites can only obtain 10 to 15 minutes of ground station visibility time per orbit, resulting in long periods of "communication blackout". For tasks such as remote sensing disaster monitoring and emergency response that require instantaneous feedback or real-time command injection, the delay and interruption of the ground-based system are unacceptable.

[0003] (2) Increased risks to the sustainable use of overseas ground stations: Establishing ground stations overseas faces sovereignty issues, long-term operational risks, high operating costs, and potential data security threats. As a globally covered asset that is not limited by geography, space-based systems can fundamentally eliminate dependence on overseas ground stations and ensure the autonomy, controllability, and security of space mission data.

[0004] 2) The necessity of addressing the operational efficiency of large-scale development of low-Earth orbit constellations (1) The need for efficient “one-to-many” control: The current telemetry and control mode in which a single control center manages a small number of satellites or a single constellation cannot adapt to constellations with tens of thousands of satellites. Space-based control systems (such as relay satellites) have multiple access capabilities, and a single relay satellite can provide services to multiple low-orbit satellites at the same time, which improves the efficiency and capacity of the control channel to a certain extent.

[0005] (2) Reduce the overall system operating cost: From the perspective of the entire life cycle, its automated and centralized management and control mode can significantly reduce the dependence on the extremely large operation and maintenance manpower of the global ground station network, and has significant economies of scale in the long run.

[0006] 3) The strategic necessity of enhancing the survivability and resilience of space systems (1) System resilience and durability: Ground-based stations are fixed and vulnerable targets, facing potential threats under extreme conditions. Space-based control nodes have stronger survivability and durability, ensuring the availability of links to critical space assets under extreme conditions, and are a core component of building a resilient, reliable and sustainable national space infrastructure.

[0007] (2) The pivotal role of the space-based information superhighway: The space-based control system is essentially a “space information superhighway”, which realizes the transformation of the observation data from the traditional linear process of “satellite-ground-user” to a flat and rapid process of “satellite-space-based node-user”, which can greatly improve the efficiency of information transmission and survivability.

[0008] Existing space-based control system solutions include space-based telemetry and control solutions based on relay satellite systems, navigation enhancement and telemetry and control fusion solutions based on the BeiDou satellite system, telemetry and control solutions based on high-throughput communication satellites (such as Tiantong-1), and integrated space-ground precision orbit determination solutions that combine medium-high orbit and low-orbit constellations. 1) Space-based telemetry, tracking, and command scheme based on relay satellite system This scheme utilizes the Tianlian relay satellite system (such as Tianlian-1 and Tianlian-2) deployed in geostationary orbit (GEO) to construct an integrated space-ground telemetry and control network. Low-Earth orbit satellites carry relay terminals, establishing two-way links with ground control centers through the relay satellites to achieve real-time command uploading, telemetry downloading, and data transmission over overseas arc segments or globally.

[0009] The system's working principle includes: (1) Uplink command injection: The ground station sends commands to the relay satellite, which then forwards them to the low-Earth orbit satellite via the inter-satellite link (S-band or Ka-band). (2) Downlink telemetry transmission: The low-Earth orbit satellite sends status data to the relay satellite, which then forwards it to the ground station and finally delivers it to the user.

[0010] 2) Navigation enhancement and telemetry integration scheme based on BeiDou satellite system This scheme utilizes the short message communication (RDSS) and navigation signal broadcasting capabilities of BeiDou satellites to provide space-based telemetry, tracking, and command (TT&C) and navigation enhancement services for low-Earth orbit constellations.

[0011] Specific functions include: (1) Short message telemetry and control: The low-orbit satellite carries a Beidou RDSS terminal to forward telemetry data to the ground operation and control center through the Beidou satellite, or to receive remote control commands.

[0012] (2) Navigation enhancement: BeiDou satellites broadcast precise orbit clock corrections, integrity information and enhancement signals to improve the autonomous orbit determination accuracy and real-time positioning capability of low-orbit satellites.

[0013] 3) A precise orbit determination scheme integrating medium-high orbit and low orbit constellations. This scheme improves the orbit determination accuracy of the entire system by combining BeiDou navigation satellites with low-Earth orbit (LEO) constellations. LEO satellites, equipped with high-precision GNSS receivers, act as "space-based monitoring stations" to continuously monitor medium- and high-Earth orbit navigation satellites globally. Combined with data from ground regional stations, precise orbit determination is achieved through Kalman filtering or dynamic orbit optimization algorithms.

[0014] Existing solutions still have shortcomings in terms of service agility, latency, service types, and intelligence. 1) Lack of service agility and convenience: unable to achieve "on-demand access" and "on-demand service". (1) "Siloed" architecture with poor interoperability: Most existing systems are vertically closed architectures (such as dedicated remote sensing satellite networks, communication constellations, and navigation systems), with each system using different protocol stacks, interface standards, and authentication mechanisms. This makes cross-constellation collaborative management, data sharing, and resource interoperability extremely difficult, and users find it hard to obtain integrated and converged services.

[0015] (2) The access process is lengthy and inconvenient: When a new satellite accesses the network or a user applies for services, it often requires a complex process of ground approval, satellite-to-ground authentication, and manual configuration, with weak automation and plug-and-play capabilities. This cannot meet the urgent need for "minute-level" access in scenarios such as emergency response and rapid launch to replenish the network.

[0016] (3) Long service activation cycle: Activating a new service for a user, such as temporarily applying for regional remote sensing data, involves multi-satellite resource coordination, link calculation and instruction uploading. Currently, it relies heavily on manual planning and cannot realize online real-time application and automation on the user side, resulting in poor service agility.

[0017] 2) Real-time performance and low latency challenges: Difficulty in meeting the demands of highly dynamic business operations. (1) Limited on-board processing capacity and extended transmission time: Most low-Earth orbit satellites still use the "transparent relay" mode, where raw data needs to be transmitted back to the ground processing center for processing and distribution. For services such as global remote sensing surveillance and high-speed relay, the end-to-end latency of data transmission to users can be as high as several hours or even several days, which cannot meet the second / minute level requirements of applications such as situational awareness and real-time target tracking.

[0018] (2) Multi-hop transmission across layers with uncertain delay: Data in the space-based network needs to go through multiple hops, possibly more than 10 hops, and each hop introduces processing and queuing delays. Existing routing algorithms are mostly based on static or periodically updated network states, which are difficult to adapt to rapid dynamic changes in the topology, resulting in non-optimal transmission paths, large delay jitter, and poor determinism.

[0019] (3) Control command injection delay: Although the space-based link can reduce the command injection delay from hours to minutes, the response speed of the existing system is still insufficient for real-time closed-loop control of highly dynamic tasks, and has not yet reached the target of "second-level closed loop".

[0020] 3) Limited service types and difficulties in integration: Failure to achieve integrated communication, navigation, and remote sensing services. (1) Functional isolation and resource rigidity: Communication, navigation and remote sensing satellites belong to different systems, and their computing, storage and spectrum resources are physically isolated, making it impossible to dynamically share and adjust them according to global needs. For example, the idle bandwidth of communication satellites cannot be temporarily allocated to remote sensing constellations for emergency data downlink.

[0021] (2) Lack of unified service abstraction and interface: There is no space-based version of Infrastructure as a Service or Platform as a Service capability. Developers find it difficult to quickly build cross-constellation integrated applications by calling remote sensing, navigation, and communication resources from different constellations through standardized APIs, just like using cloud computing.

[0022] (3) Data and application layer separation: The management and control system usually only manages the network and resource layers, and is not sufficiently decoupled from applications and services. The system does not understand the specific needs of upper-layer applications and cannot achieve fine-grained resource allocation and guarantee for application QoS.

[0023] 4) Insufficient level of intelligence: High reliance on ground-based and human decision-making (1) Lack of onboard intelligence: Most decision-making services, such as mission planning, fault diagnosis, and resource allocation, rely on ground computing and human intervention. Onboard computing resources are limited, and there is a lack of the ability to carry high-performance AI processing units, making it impossible to achieve real-time intelligent processing and autonomous collaborative decision-making in orbit.

[0024] (2) Lack of prediction and optimization capabilities: The existing system mainly uses reactive control based on the current state and lacks the ability to use big data and AI for predictive maintenance, dynamic traffic prediction and intelligent resource reservation.

[0025] (3) Unable to handle complex anomalies: For complex anomaly scenarios such as network congestion, malicious interference, and multi-satellite failure, the system mainly relies on preset rules and the experience of ground personnel to handle them, lacking self-learning and adaptive recovery capabilities.

[0026] Therefore, it is necessary to provide a new medium-orbit space-based control system based on a large edge-side model to solve the above-mentioned technical problems. Summary of the Invention

[0027] The purpose of this disclosure is to provide a medium-orbit space-based control system based on a large edge-side model in order to solve the above-mentioned problems.

[0028] This disclosure achieves the above objectives through the following technical solutions: A medium-orbit space-based control system based on an edge-side large model includes a physical support layer, an intelligent service layer, a collaborative control layer, and a user service layer. The physical support layer includes nine medium-Earth orbit satellites, with three satellites forming a group. The three satellites in each group are interconnected via laser links, and satellites in different orbits have the ability to communicate on demand. Each satellite is equipped with an intelligent service unit for local task processing, collaborative decision-making, and distributed computing. The intelligent service unit is the hardware and software carrier entity of the intelligent service layer. The intelligent service layer integrates edge-side large models with perception, decision-making, and communication functions, and is adapted to the low power consumption and limited computing resources of satellites. The collaborative control layer enables inter-satellite task scheduling, data sharing, and resource coordination through inter-satellite links; The user service layer provides an interface for interaction between user terminals and satellites, allowing users to directly send service requests and receive results.

[0029] As a further optimization of this disclosure, the intelligent service unit includes a perception service module, an edge big model module, a decision execution module, a communication management module, and a storage module; The perception service module is used to collect satellite body status data, preprocess the collected satellite body status data, and collect and parse user request data. The edge-side large model module is used for task evaluation, collaborative decision-making, and resource optimization; The decision execution module is used to generate specific control commands based on the output of the edge-side large model; The communication management module is used to provide link services between satellites, between satellites and ground stations, and between satellites and users based on the link resources of medium Earth orbit satellites. The storage module is used to store model parameters, business data, and task logs.

[0030] As a further optimization of this disclosure, the satellite body status data includes attitude, power supply, payload operating status, and orbital parameters; The perception service module performs correction and preprocessing on the collected satellite body state data, including noise reduction, noise removal, and normalization; and inputs the preprocessed data into the edge-side large model. The perception service module collects and parses user request data by receiving service requests from user terminals through the communication module, and parsing the request type, priority, and parameter requirements; the request type includes communication, remote sensing, and navigation; the priority includes urgent and normal.

[0031] As a further optimization of this disclosure, the edge-side large model module adopts a lightweight large model or a domain-specific large model designed based on the characteristics of satellite missions, including a sensing mission model and a communication resource allocation model.

[0032] As a further optimization of this disclosure, the task evaluation capability of the edge-side large model module includes: The inputs include satellite status, preprocessed data, and user-requested sensing data, which are used to assess satellite processing capabilities and mission feasibility. The collaborative decision-making of the edge-side large model module includes: If a satellite is unable to carry out the task due to factors including energy, attitude, and computational load, the edge-side large model module sends a collaboration request to neighboring satellites via inter-satellite links, while simultaneously receiving capability assessment results from other satellites and generating a collaboration scheme. Resource optimization for the edge-side large model module includes: Based on mission requirements, optimize satellite resource allocation, including calculating the required attitude adjustment angle, imaging time, energy consumption, and data processing resource consumption for remote sensing missions; and allocating bandwidth resources, link selection, and processing resources according to mission priorities for communication missions.

[0033] As a further optimization of this disclosure, the edge-side large model module supports on-orbit optimization, including federated learning and model compression; Federated learning: Each satellite trains a model using local data, sends the model parameters to the ground station, and the ground station aggregates the parameters and returns them to the satellite to achieve model iteration; Model compression: The original model is compressed using processing techniques including pruning, quantization, and knowledge distillation.

[0034] As a further optimization of this disclosure, the control commands generated by the decision execution module include attitude adjustment commands, load control commands, communication control commands, and collaborative control commands; after the commands are executed, the execution status is monitored in real time, and if an abnormality occurs, feedback is given to the edge-side large model to regenerate commands.

[0035] As a further optimization of this disclosure, the inter-satellite link service provided by the communication management module includes the use of laser communication to realize inter-satellite task request transmission, data sharing, and model parameter transmission; The communication management module provides satellite and ground station link services, including preferential use of Q and V bands for receiving mission management data uploads, reconstruction data and model update instructions from the ground station, while transmitting business data and control data to the ground station. The communication management module provides satellite and user-side link services, including the use of multi-beam Ka-band link resources to enable direct communication with satellites and user terminals, supporting user request sending and result return.

[0036] As a further optimization of this disclosure, the core mechanism for the collaborative control layer to achieve inter-satellite task scheduling through inter-satellite links is a task allocation mechanism; The task allocation mechanism includes: When a satellite receives a user request, the edge-side large model first evaluates its own capabilities: When its own capability value is not less than a preset threshold, the satellite autonomously processes the task; when its own capability value is less than the preset threshold, the satellite sends a cooperation request to neighboring satellites through an inter-satellite link. Upon receiving the request, the edge-side large model assesses its own capabilities, returns a collaborative response, and requests the satellite to generate a collaborative scheme based on the response.

[0037] As a further optimization of this disclosure, the core mechanism of the collaborative control layer for coordinating resources among satellites through inter-satellite links is a resource coordination mechanism; The resource coordination mechanism includes: Regarding bandwidth resources: When multiple user requests arrive simultaneously, the edge-side large model allocates bandwidth according to request priority using a weighted fair queue algorithm; Regarding energy resources: The satellite optimizes battery charging strategies by planning routine mission requirements and changes in space illumination conditions for a duration no less than a preset time using a large edge-side model.

[0038] As a further optimization of this disclosure, the core mechanism for the collaborative control layer to achieve data sharing between satellites through inter-satellite links is a data sharing mechanism; The data sharing mechanism includes: Satellites share critical data, including user service data, system resource status data, and user request data, through a publish-subscribe protocol. Data sharing adopts incremental updates to reduce bandwidth consumption of inter-satellite links.

[0039] As a further optimization of this disclosure, the user service layer provides an interaction interface between the user terminal and the satellite, allowing the user to directly send service requests and receive results, including: For remote sensing services, it supports on-orbit mission analysis based on user business requests, scheduling low-orbit satellites to provide services including sensing and detection, and transmitting data back to medium-orbit control satellites in real time to provide real-time on-orbit fusion computing and directly provide usable data products. For communication and navigation service needs, the medium orbit control system, as an access and relay service node, supports transparent transmission and regeneration forwarding between low orbit satellites, supports the rapid transfer and landing of low orbit satellite data, and provides low-latency services.

[0040] The beneficial effects of this disclosure are as follows: 1. A large-scale model at the edge drives a closed loop of "perception-decision-execution," enabling satellites to make "autonomous decisions" and reducing reliance on ground stations: Traditional satellites rely on ground stations to transmit command sequences, which cannot cope with dynamic scenarios. In this solution, the satellite collects data through a sensing module, and a large-scale model at the edge evaluates mission feasibility and risks in real time, generates decision commands, and the decision execution module directly controls the satellite to execute them, forming a closed-loop control.

[0041] 2. Based on a multi-satellite collaboration mechanism of "capability assessment," improve the constellation's "overall mission processing efficiency" and "resource utilization": In traditional space-based systems, satellites process tasks independently. If a single satellite's capabilities are insufficient, the task may fail or be delayed. In this scheme, satellites assess their own capabilities using a large edge-side model. If their capabilities are insufficient, they send a coordination request to neighboring satellites via inter-satellite links. Other satellites return their capability assessment results, and the edge-side model generates a coordination plan.

[0042] 3. Edge-side large-scale model-driven "precise resource optimization" solves the problem of "unreasonable resource allocation" in traditional systems: Traditional space-based systems typically allocate resources (bandwidth, energy, attitude) statically, making them unsuitable for dynamic needs. In this solution, the edge-side large model optimizes resource allocation based on remote sensing / communication / navigation mission type, priority (urgent / normal), and satellite status.

[0043] 4. The user "direct interaction" mode solves the problem of "poor service convenience" in traditional systems: In traditional space-based systems, users must send requests through ground station interfaces, a complex process that doesn't provide real-time results. In this solution, users send requests directly to the satellite using ordinary terminals via a unified API interface (supporting RESTful, MQTT, and other protocols), and the satellite processes the requests and returns the results directly. This transforms the "ground station-centric" service model into a "user-centric" one, improving service convenience and accessibility.

[0044] 5. Federated learning optimizes edge-side models, addressing the traditional system's problems of "difficult model iteration" and "privacy protection": Traditional space-based systems often rely on ground-based stations to train models, which cannot adapt to local satellite data. Furthermore, model updates require the transmission of large amounts of data, leading to slow model iteration and privacy risks. This solution employs federated learning, shifting model optimization from a centralized to a distributed approach. Each satellite trains its edge model using local log data; the satellite sends model parameters to the ground station, which aggregates the parameters to generate a global model; the ground station then distributes the global model parameters to the satellites, which then update their local models. Attached Figure Description

[0045] Figure 1 This is a system structure block diagram in an embodiment of this disclosure; Figure 2This is a functional design diagram of the Intelligent Service Unit (ISU) in an embodiment of this disclosure. Detailed Implementation

[0046] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0047] like Figure 1 As shown, the technical solution disclosed herein is based on a "medium-orbit intelligent control satellite system + edge-side intelligent service unit + multi-satellite collaborative mechanism," which is a medium-orbit space-based control system based on an edge-side large model. It comprises four layers: a physical support layer, an intelligent service layer, a collaborative control layer, and a user service layer. The bottom-up architecture design is as follows: 1. Physical load-bearing layer: The basic configuration consists of nine medium Earth orbit (MEO) satellites, arranged in groups of three at an altitude of 11,000 km–12,000 km. Each group of three satellites is interconnected via laser links, enabling on-demand communication between satellites in different orbits. Each satellite is equipped with an Intelligent Service Unit (ISU) responsible for local task processing, collaborative decision-making, and distributed computing.

[0048] 2. Intelligent Service Layer: like Figure 2 As shown, the intelligent service unit is the hardware and software carrier entity of the intelligent service layer and the core of this invention. It integrates the large edge model with perception, decision-making and communication functions, adapts to the low power consumption and limited computing resources of satellites, and realizes the end-to-end closed loop of "perception-processing-decision-execution".

[0049] The intelligent service unit consists of business modules such as a perception service module, an edge large model module, a decision execution module, a communication management module, and a storage module, as well as a power supply module. The functions of each business module are as follows: 1) Perception Service Module: (1) Supports the collection of satellite body status data, including attitude, power supply, payload working status, orbital parameters and other information, providing a data source for health management.

[0050] (2) Data preprocessing: The received raw data is subjected to correction preprocessing such as noise reduction, removal, and normalization, and then input into the edge side large model.

[0051] (3) Supports the collection and parsing of user request data: Receives service requests from user terminals through the communication module, and parses the request type (communication / remote sensing / navigation), priority (urgent / normal), and parameter requirements.

[0052] 2) Edge-side large model module: (1) Model selection: Use lightweight large models or domain-specific large models designed based on the characteristics of satellite missions (such as remote sensing mission models and communication resource allocation models). The model size should be controlled within 100MB and the computational load should be controlled within 50TOPS.

[0053] (2) Model Functions: First, mission assessment: Input sensing data (satellite status, preprocessed data, user requests) to assess satellite processing capabilities (e.g., whether there is enough energy to complete imaging, whether there is bandwidth to transmit results) and mission feasibility (e.g., whether there are cloud cover in the area, whether there is a space debris threat).

[0054] Second, collaborative decision-making: If the satellite itself cannot bear the load due to factors such as energy, attitude, and computational load, the model sends a collaborative request to neighboring satellites through inter-satellite links, while receiving the capability assessment results of other satellites and generating a collaborative solution.

[0055] Third, resource optimization: Optimize satellite resource allocation according to mission requirements. For example, for remote sensing missions, calculate the required attitude adjustment angle, imaging time, energy consumption, and data processing resource consumption; for communication missions, allocate bandwidth resources, link selection, and processing resources according to mission priority.

[0056] (3) The model supports on-orbit optimization: First, federated learning: the ground station acts as the "server" and the satellite acts as the "client". Each satellite trains the model with local data (mission logs, sensing data) and sends the model parameters (rather than the raw data) to the ground station. The ground station aggregates the parameters and returns them to the satellite to realize model iteration.

[0057] Second, model compression: using techniques such as pruning, quantization, and knowledge distillation, the original large model is compressed to 1 / 10 of its size while maintaining more than 90% of the task evaluation accuracy.

[0058] 3) Decision Execution Module: Based on the output of the edge-side large model, specific control commands are generated, such as attitude adjustment commands, load control commands, communication control commands, and collaborative control commands. After the commands are executed, the execution status is monitored in real time (e.g., whether the attitude has been adjusted correctly and whether the load has been activated). If any abnormality occurs, feedback is sent to the edge-side large model to regenerate the commands.

[0059] 4) Communication Management Module: Based on the link resources of medium Earth orbit satellites, it provides various link services between satellites (such as within the medium Earth orbit system, between medium Earth orbit and LEO or GEO, etc.), between satellites and ground stations, and between satellites and users: (1) Inter-satellite link: The maximum laser communication speed is no less than 5Gbps to realize the transmission of mission requests, data sharing and model parameter transmission between satellites.

[0060] (2) Star-ground link: Q and V bands are used first to receive mission management data upload, reconstruction data and model update instructions from the ground station, while transmitting business data and control data to the ground station.

[0061] (3) Satellite-end link: Using multi-beam Ka-band link resources, it realizes direct communication with user terminals such as satellites and air targets, and supports user request sending and result return.

[0062] 5) Storage module: (1) Storage model parameters: The weights, biases and other parameters of the large model on the edge side are stored using non-volatile storage (such as NAND Flash) with a capacity of ≥1TB.

[0063] (2) Stored business data: Preprocessed satellite status, space environment, user request data and sensing business data, etc., with a storage time of ≥7 days.

[0064] (3) Store task logs: Record task processing flow, such as request content, processing time, cooperating satellites, results and satellite status, so as to facilitate backtracking.

[0065] 3. Collaborative Control Layer Mechanism Design Inter-satellite links enable mission scheduling, data sharing, and resource coordination among satellites. The core mechanisms include: (1) Task allocation mechanism: When a satellite receives a user request, the edge-side large model first assesses its own capabilities: ; in, , , , To optimize weights through federated learning, Battery represents battery power, ranging from 0 to 1; Payload represents payload availability, ranging from 0 to 1; Bandwidth represents remaining bandwidth, ranging from 0 to 1; and TaskLoad represents current task load, ranging from 0 to 1.

[0066] like The satellite autonomously processes its tasks; if The satellite sends a coordination request (including the request content and its own capabilities) to neighboring satellites via an inter-satellite link; where Threshold is a preset threshold.

[0067] Upon receiving the request, other satellites use the edge-side large model to assess their own capabilities, return a collaborative response, and request the satellites to generate a collaborative plan based on the response.

[0068] (2) Resource coordination mechanism: Regarding bandwidth resources: When multiple user requests arrive simultaneously, the edge-side large model allocates bandwidth based on request priority (e.g., emergency rescue requests have a priority of 1, while regular user requests have a priority of 0.5) using a weighted fair queue algorithm. ; in, Bandwidth allocation for user i Priority for user i For user j's priority, This represents the remaining bandwidth of the satellite.

[0069] Regarding energy resources: The satellite optimizes its battery charging strategy by planning for routine mission requirements of no less than 24 hours and changes in space lighting conditions through a large-scale edge model.

[0070] (3) Data sharing mechanism: Satellites share critical data, such as user service data, system resource status data, and user request data, via a publish / subscribe (Pub / Sub) protocol. Data sharing employs incremental updates to reduce bandwidth consumption on inter-satellite links.

[0071] 4. User Service Layer: It provides an interface for interaction between user terminals and satellites, allowing users to directly send service requests (communication, remote sensing, navigation) and receive results. For remote sensing services, it supports on-orbit mission analysis based on user service requests, scheduling low-Earth orbit (LEO) satellites to provide sensing and detection services, and transmitting data back to medium-Earth orbit (MEO) control satellites in real time, providing real-time on-orbit fusion computing and directly delivering usable data products. For communication and navigation services, the MEO control system acts as an access and relay service node, supporting transparent transmission and regeneration forwarding between LEO satellites, enabling rapid data transfer and deployment from LEO satellites, and providing low-latency services.

[0072] This disclosure aims to address the core issues faced by existing space-based control systems when dealing with large-scale, multi-type low-Earth orbit constellations, including centralized control bottlenecks, low on-board intelligence levels, prolonged service response times, and insufficient system scalability. Specifically, the purpose of this disclosure is as follows: 1) Breaking the bottleneck of centralized architecture: By distributing intelligence to "intelligent service units" on multiple medium Earth orbit (MEO) satellites, a distributed and decentralized space-based control network is built to avoid the performance and security risks brought about by a single master control node.

[0073] 2) Achieve near real-time intelligent control: Utilize the large edge model deployed on MEO satellites to process and analyze data on orbit and generate decision commands locally, greatly reducing cross-domain transmission delays between "on-board-ground-on-board" and achieving minute-level or even second-level closed-loop response for low-Earth orbit constellations.

[0074] 3) Provide integrated, on-demand management and control services: Through the unified service interface provided by the intelligent service unit, basic capabilities such as communication, telemetry, and navigation enhancement are encapsulated into "management and control as a service" that can be flexibly invoked, supporting users to obtain integrated space-based services on demand and conveniently.

[0075] 4) Enhance the overall resilience and autonomy of the system: Through the collaboration and autonomous decision-making of multiple agents, the system can autonomously reconstruct and maintain critical services when some nodes fail or the network topology changes dynamically, thus possessing a high degree of resilience and survivability.

[0076] Compared with the prior art, this disclosure can achieve the following significant technical effects: 1) Extremely low control and management service latency: Utilizing the wide coverage and lower path loss of MEO orbit compared to LEO, and combined with on-board edge intelligent processing, the telemetry data analysis, anomaly diagnosis, and control command generation of low-orbit satellites are completed in MEO orbit. The end-to-end control and management response time is shortened from hours to minutes or even seconds, meeting the real-time requirements of highly dynamic tasks.

[0077] 2) On-demand intelligent services: The edge-side big model empowers each MEO satellite with powerful on-orbit information extraction, situational awareness and collaborative decision-making capabilities, enabling it to provide advanced intelligent services such as predictive maintenance, intelligent resource scheduling and autonomous collision avoidance planning, far exceeding the level of automation based on rules and ground pre-computation in the traditional sense.

[0078] 3) Convenient service access and integration: Through standardized service interfaces, low-orbit satellites or ground users can transparently call the required communication, navigation and remote sensing integrated services through service requests, shielding the complexity of the underlying heterogeneous network and greatly improving the convenience and efficiency of space-based resource use.

[0079] 4) Enhanced system scalability and resilience: The decentralized MEO intelligent node network largely eliminates single points of failure. When some MEO satellites fail or links are interrupted, the system can autonomously reroute tasks and take over services through intelligent agents, ensuring the continuity of core management and control functions. Simultaneously, the distributed architecture allows for a linear increase in the system's total processing capacity and communication bandwidth by increasing the number of MEO intelligent satellites, resolving the single-point capacity bottleneck problem of centralized GEO relay systems.

[0080] The embodiments described above are merely examples of several implementations of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent disclosure. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these modifications and improvements all fall within the protection scope of this disclosure.

Claims

1. A medium-orbit space-based control system based on an edge-side large model, characterized in that, It includes the physical layer, intelligent service layer, collaborative control layer, and user service layer; The physical support layer includes nine medium-Earth orbit satellites, with three satellites forming a group. The three satellites in each group are interconnected via laser links, and satellites in different orbits have the ability to communicate on demand. Each satellite is equipped with an intelligent service unit for local task processing, collaborative decision-making, and distributed computing. The intelligent service unit is the hardware and software carrier entity of the intelligent service layer. The intelligent service layer integrates edge-side large models with perception, decision-making, and communication functions, and is adapted to the low power consumption and limited computing resources of satellites. The collaborative control layer enables inter-satellite task scheduling, data sharing, and resource coordination through inter-satellite links; The user service layer provides an interface for interaction between user terminals and satellites, allowing users to directly send service requests and receive results.

2. The medium-orbit space-based control system based on a large edge-side model according to claim 1, characterized in that, The intelligent service unit includes a perception service module, an edge big model module, a decision execution module, a communication management module, and a storage module; The perception service module is used to collect satellite body status data, preprocess the collected satellite body status data, and collect and parse user request data. The edge-side large model module is used for task evaluation, collaborative decision-making, and resource optimization; The decision execution module is used to generate specific control commands based on the output of the edge-side large model; The communication management module is used to provide link services between satellites, between satellites and ground stations, and between satellites and users based on the link resources of medium Earth orbit satellites. The storage module is used to store model parameters, business data, and task logs.

3. A medium-orbit space-based control system based on a large edge-side model according to claim 2, characterized in that, The satellite's status data includes attitude, power supply, payload operating status, and orbital parameters. The perception service module performs correction and preprocessing on the collected satellite body state data, including noise reduction, noise removal, and normalization. The preprocessed data is input into the edge-side large model; The perception service module collects and parses user request data by receiving service requests from user terminals through the communication module and parsing the request type, priority, and parameter requirements. The request types include communication, remote sensing, and navigation; the priorities include urgent and normal.

4. A medium-orbit space-based control system based on a large edge-side model according to claim 2, characterized in that, The edge-side large model module adopts a lightweight large model or a domain-specific large model designed based on the characteristics of satellite missions, including a sensing mission model and a communication resource allocation model.

5. A medium-orbit space-based control system based on a large edge-side model according to claim 2, characterized in that, The task evaluation of the edge-side large model module includes: The inputs include satellite status, preprocessed data, and user-requested sensing data, which are used to assess satellite processing capabilities and mission feasibility. The collaborative decision-making of the edge-side large model module includes: If a satellite is unable to carry out the task due to factors including energy, attitude, and computational load, the edge-side large model module sends a collaboration request to neighboring satellites via inter-satellite links, while simultaneously receiving capability assessment results from other satellites and generating a collaboration scheme. Resource optimization for the edge-side large model module includes: Based on mission requirements, optimize satellite resource allocation, including calculating the required attitude adjustment angle, imaging time, energy consumption, and data processing resource consumption for remote sensing missions; and allocating bandwidth resources, link selection, and processing resources according to mission priorities for communication missions.

6. A medium-orbit space-based control system based on a large edge-side model according to claim 2, characterized in that, The edge-side large model module supports on-orbit optimization, including federated learning and model compression; Federated learning: Each satellite trains a model using local data, sends the model parameters to the ground station, and the ground station aggregates the parameters and returns them to the satellite to achieve model iteration; Model compression: The original model is compressed using processing techniques including pruning, quantization, and knowledge distillation.

7. A medium-orbit space-based control system based on a large edge-side model according to claim 2, characterized in that, The control commands generated by the decision execution module include attitude adjustment commands, load control commands, communication control commands, and collaborative control commands. After the commands are executed, the execution status is monitored in real time. If an abnormality occurs, the command is fed back to the edge-side large model to regenerate the commands.

8. A medium-orbit space-based control system based on a large edge-side model according to claim 2, characterized in that, The communication management module provides inter-satellite link services, including laser communication, for transmitting mission requests, sharing data, and transmitting model parameters between satellites. The communication management module provides satellite and ground station link services, including preferential use of Q and V bands for receiving mission management data uploads, reconstruction data and model update instructions from the ground station, while transmitting business data and control data to the ground station. The communication management module provides satellite and user-side link services, including the use of multi-beam Ka-band link resources to enable direct communication with satellites and user terminals, supporting user request sending and result return.

9. A medium-orbit space-based control system based on an edge-side large model according to claim 1, characterized in that, The core mechanism for inter-satellite task scheduling in the collaborative control layer through inter-satellite links is the task allocation mechanism. The task allocation mechanism includes: When a satellite receives a user request, the edge-side large model first evaluates its own capabilities: When its own capability value is not less than a preset threshold, the satellite autonomously processes the task; when its own capability value is less than the preset threshold, the satellite sends a cooperation request to neighboring satellites through an inter-satellite link. Upon receiving the request, the edge-side large model assesses its own capabilities, returns a collaborative response, and requests the satellite to generate a collaborative scheme based on the response.

10. A medium-orbit space-based control system based on an edge-side large model according to claim 1, characterized in that, The core mechanism of the collaborative control layer for coordinating resources among satellites through inter-satellite links is the resource coordination mechanism. The resource coordination mechanism includes: Regarding bandwidth resources: When multiple user requests arrive simultaneously, the edge-side large model allocates bandwidth according to request priority using a weighted fair queue algorithm; Regarding energy resources: The satellite optimizes battery charging strategies by planning routine mission requirements and changes in space illumination conditions for a duration no less than a preset time using a large edge-side model.

11. A medium-orbit space-based control system based on an edge-side large model according to claim 1, characterized in that, The core mechanism by which the collaborative control layer achieves data sharing between satellites through inter-satellite links is a data sharing mechanism. The data sharing mechanism includes: Satellites share critical data, including user business data, system resource status data, and user request data, through a publish-subscribe protocol. Data sharing uses incremental updates to reduce bandwidth consumption of inter-satellite links.

12. A medium-orbit space-based control system based on an edge-side large model according to claim 1, characterized in that, The user service layer provides an interface for interaction between the user terminal and the satellite, allowing users to directly send service requests and receive results, including: For remote sensing services, it supports on-orbit mission analysis based on user business requests, scheduling low-orbit satellites to provide services including sensing and detection, and transmitting data back to medium-orbit control satellites in real time to provide real-time on-orbit fusion computing and directly provide usable data products. For communication and navigation service needs, the medium orbit control system, as an access and relay service node, supports transparent transmission and regeneration forwarding between low orbit satellites, supports the rapid transfer and landing of low orbit satellite data, and provides low-latency services.

Images