A lightweight federated learning method for low earth orbit satellite network and related device

Abstract

The application discloses a lightweight federated learning method for a low-orbit satellite network and related devices, comprising the following steps: a ground station trains a teacher model and distills a lightweight student model; the ground station selects a satellite node and distributes the lightweight student model to the selected satellite node; the selected satellite node receives and trains the lightweight student model; the selected satellite node uploads the trained lightweight student model to the ground station in a star-ground visible window; the ground station aggregates the received lightweight student model and calibrates the aggregated lightweight student model by batch normalization statistics to obtain a next round of global student model; when the configuration is turned on, the ground station updates the teacher model by reverse distillation of the next round of global student model. The method and related devices solve the problem of performance decline of federated learning training caused by limited star-ground communication, heterogeneous satellite data and dynamic access in the low-orbit satellite network.

Description

Technical Field

[0001] This invention belongs to the fields of communication technology and federated learning technology, and relates to a lightweight federated learning method and related devices for low-Earth orbit satellite networks. Background Technology

[0002] Once low-Earth orbit satellite constellations are equipped with remote sensing imaging and onboard sensors, they will continuously generate a large amount of Earth observation data. Downloading this raw data to ground stations for unified model training is limited by factors such as satellite-to-ground link bandwidth, visibility windows, and data privacy.

[0003] Federated learning enables multiple satellite nodes to collaboratively train models without uploading raw data, making it suitable for low-Earth orbit (LEO) satellite scenarios. However, directly adopting traditional federated learning in LEO satellite networks still presents three main problems: First, the intermittent visibility of the satellite-to-ground link leads to high communication overhead within the limited visibility window if a large number of model parameters are transmitted in each round. Second, the significant differences in data distribution across different orbits and regions can easily cause client drift under non-independent and identically distributed conditions. Third, the dynamic access and departure of satellite nodes from ground station coverage areas due to orbital movement results in a constantly changing training node set, leading to instability among clients participating in aggregation in each round and thus affecting global model convergence and training stability.

[0004] Existing methods primarily address the federated learning problem for low-Earth orbit (LEO) satellites from the perspectives of server-satellite switching, intra-orbit aggregation, inter-satellite topology reconstruction, shared data allocation, dynamic aggregation weights, or secure aggregation through secret sharing. The focus is typically on communication paths, aggregation locations, or aggregation security. However, for scenarios with limited satellite-to-ground links and weak satellite computing power, there is still a lack of an integrated approach that combines ground-based teacher model preservation, lightweight student model transmission, rapid satellite adaptation, and stable local self-distillation updates. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a lightweight federated learning method and related apparatus for low-Earth orbit satellite networks. This method and apparatus can combine ground teacher model retention, lightweight student model transmission, rapid satellite adaptation, and stable local self-distillation updates, thereby solving the problem of decreased federated learning training performance caused by limited satellite-to-ground communication, heterogeneous satellite data, and dynamic access in low-Earth orbit satellite networks.

[0006] To achieve the above objectives, this invention discloses a lightweight federated learning method for low-Earth orbit satellite networks, comprising: Initialize the low-Earth orbit satellite federated learning scenario; The ground station trains the teacher model, and distills it to obtain a lightweight student model; The ground station selects a satellite node and sends the lightweight student model to the selected satellite node; The selected satellite node receives and trains the lightweight student model; The selected satellite node uploads the trained lightweight student model to the ground station within the satellite-to-ground visibility window; The ground station aggregates the received lightweight student models and performs batch normalization statistics calibration on the aggregated lightweight student models to obtain the next round of global student models. When the configuration is enabled, the ground station uses the next round of global student models to back-distill and update the teacher models.

[0007] Furthermore, the process of initializing the low-Earth orbit satellite federated learning scenario is as follows: The ground station acts as the federated learning server, and multiple satellite nodes act as clients. Each satellite node stores its local observation data. The ground station maintains the global student model and teacher model and records link metrics, including satellite node distance, uplink and downlink rates, lightweight student model size, visibility window time, and communication feasibility.

[0008] Furthermore, the process of the ground station training the teacher model and distilling it to obtain the lightweight student model is as follows: the ground station uses representative data to train the teacher model, and then uses the teacher model output and real labels to jointly guide the training of the student model to obtain the lightweight student model.

[0009] Furthermore, the visible window constraint is expressed as:

[0010] in, The total transmission time required to complete one model distribution and upload, To keep the student model size lightweight, and These are the downlink speed and the uplink speed, respectively. The duration of the visible window.

[0011] Furthermore, the loss function used by the selected satellite node in receiving and training the lightweight student model is:

[0012] in, Used to constrain the prediction of the true label by the lightweight student model. Used to constrain the consistency between lightweight student models and snapshot soft objectives or teacher knowledge. The weights are adjustable.

[0013] Furthermore, the process by which the ground station aggregates the received lightweight student models is as follows: the ground station performs sample-weighted aggregation on the received lightweight student models, wherein,

[0014] in, These are the global student model parameters after aggregation in the (r+1)th round. For the lightweight student model parameters uploaded by the k-th participating satellite node, This represents the number of local samples corresponding to that satellite node. This refers to the set of satellite nodes participating in the aggregation in the current round. For set The total number of local samples across all satellite nodes.

[0015] Furthermore, the teacher model employs an image classification network deployed at a ground station, while the lightweight student model employs a lightweight convolutional neural network deployed on satellite nodes with fewer parameters than the teacher model. Only the parameters of the lightweight student model are transmitted in the satellite-to-ground link.

[0016] This invention discloses a lightweight federated learning system for low-Earth orbit satellite networks, comprising: The initialization module is used to initialize the federated learning scenario for low-Earth orbit satellites. The first training module is used to train the teacher model at the ground station and distill it to obtain a lightweight student model; The selection module is used by the ground station to select satellite nodes and send the lightweight student model to the selected satellite nodes. The second training module is used for the selected satellite nodes to receive and train the lightweight student model; The upload module is used by the selected satellite node to upload the trained lightweight student model to the ground station within the satellite-to-ground visibility window; The aggregation module is used by the ground station to aggregate the received lightweight student models and perform batch normalization statistics calibration on the aggregated lightweight student models to obtain the next round of global student models. When the configuration is enabled, the ground station uses the next round of global student models to back-distill and update the teacher models.

[0017] The present invention discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the lightweight federated learning method for low-Earth orbit satellite networks.

[0018] The present invention discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the lightweight federated learning method for low-Earth orbit satellite networks.

[0019] The present invention has the following beneficial effects: In practical operation, the lightweight federated learning method and related apparatus for low-Earth orbit (LEO) satellite networks described in this invention transmits only the parameters of the lightweight student model via the satellite-to-ground link, while the large-capacity teacher model remains at the ground station. This reduces the communication overhead for model distribution and uploading, and differs from federated learning schemes that directly transmit the complete global model. Furthermore, this invention uses adaptive progressive self-distillation, with student model snapshots during the training phase as a soft objective, to reduce output fluctuations and improve training stability during local training of the lightweight model on the satellite. Finally, this invention, through the combination of satellite node selection and batch normalization statistic calibration after aggregation, can adapt to the dynamic access and exit of LEO satellites and reduce the statistical bias caused by differences in the distribution of data from different satellites. This solves the problem of decreased federated learning training performance caused by limited satellite-to-ground communication, heterogeneous satellite data, and dynamic access in LEO satellite networks. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention; Figure 3 This is a performance comparison chart from the simulation experiment. Detailed Implementation

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

[0023] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0025] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0026] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0027] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0029] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0030] Example 1 refer to Figure 1 and Figure 2 The lightweight federated learning method for low-Earth orbit satellite networks described in this invention includes the following steps: 1) Initialize the low-Earth orbit satellite federated learning scenario; The ground station serves as the federated learning server, while multiple low-Earth orbit satellite nodes act as clients. Each satellite node stores its local observation data, and the ground station maintains the global student and teacher models, recording link metrics such as satellite node distance, uplink and downlink rates, lightweight student model size, visible window time, and communication feasibility.

[0031] 2) The ground station trains the teacher model and distills it to obtain a lightweight student model; In this embodiment, the teacher model uses ResNet18, and the student model uses MobileNetV2. The ground station trains the teacher model using representative data, and then uses the teacher model output and the real labels to jointly guide the training of the student model, resulting in a lightweight student model. This lightweight student model serves as the subsequent satellite-to-ground transmission model, while the teacher model remains at the ground station and is not transmitted in the satellite-to-ground link.

[0032] 3) The ground station selects satellite nodes based on the size of the lightweight student model, the uplink and downlink rates of each satellite node, and the visible window, and then sends the lightweight student model to the selected satellite nodes; The ground station reads the link metrics reported by each satellite node, determines whether the visibility window constraint is met based on the sum of the download time and upload time of the lightweight student model, and prioritizes satellite nodes that meet the visibility window to participate in training. For satellite nodes that do not meet the communication conditions temporarily, the ground station marks them as blocking or probe nodes and re-evaluates them in subsequent rounds.

[0033] In one embodiment, the communication feasibility of the k-th satellite node is determined using the following visibility window constraint:

[0034] in, The total transmission time required to complete one model distribution and upload, To keep the student model size lightweight, and These are the downlink speed and the uplink speed, respectively. The duration of the visible window.

[0035] 4) Local training of lightweight student models on satellite nodes; Satellite nodes train a lightweight student model using local observation data. For the meta-learning part, the satellite node divides the local observation data into adaptation support data and query data, performs a few steps of adaptation first, and then uses the query data to form the update direction. For the adaptive progressive self-distillation part, the satellite node retains a snapshot of the lightweight student model during the training phase, uses the snapshot output as a soft objective, and updates the lightweight student model in combination with the current task loss to obtain the trained lightweight student model, thereby reducing local training oscillations.

[0036] The task loss is represented as follows:

[0037] in, Used to constrain the prediction of the true label by the lightweight student model. Used to constrain the consistency between lightweight student models and snapshot soft objectives or teacher knowledge. The weights are adjustable.

[0038] 5) Satellite nodes upload locally trained lightweight student models to the ground station within the satellite-to-ground visibility window; Since the uploaded object is a lightweight student model, the satellite-to-ground link does not need to transmit ground teacher model parameters or upload raw observation data, thereby reducing the amount of communication per training round and protecting local data.

[0039] 6) The ground station aggregates the received lightweight student models and performs batch normalization statistics calibration on the aggregated lightweight student models to obtain the next round of global student models. When the configuration is enabled, the ground station uses the next round of global student models to back-distill and update the teacher model, so that the teacher model can absorb the lightweight student model information after multiple rounds of satellite node training.

[0040] When aggregating lightweight student models at ground stations, a sample size-weighted average method can be used for aggregation, where:

[0041] in, These are the global student model parameters after aggregation in the (r+1)th round. For the lightweight student model parameters uploaded by the k-th participating satellite node, This represents the number of local samples corresponding to that satellite node. This refers to the set of satellite nodes participating in the aggregation in the current round. For set The total number of local samples across all satellite nodes.

[0042] Simulation Experiment refer to Figure 3 This simulation experiment uses the Flower federated learning framework, with 5 satellite clients. Four clients are selected for each training round, and five clients are selected for each evaluation round, with 50 training rounds. The CIFAR-10 dataset is used, and a Dirichlet Non-IID partitioning setting is employed to compare different levels of heterogeneity. Ablation experiments include FedAvg, GroundKD, FedMeta, APSKD, and combinations thereof. The complete method A7 incorporates ground knowledge distillation, meta-learning rapid adaptation, and adaptive progressive self-distillation. In a strong Non-IID scenario, the complete method A7 achieves a local accuracy of 0.8683 and a peak accuracy of 0.8695 in round 50, while maintaining a lightweight student model-level communication throughput.

[0043] Example 2 The lightweight federated learning system for low-Earth orbit satellite networks described in this invention includes: The initialization module is used to initialize the federated learning scenario for low-Earth orbit satellites. The first training module is used to train the teacher model at the ground station and distill it to obtain a lightweight student model; The selection module is used by the ground station to select satellite nodes and send the lightweight student model to the selected satellite nodes. The second training module is used for the selected satellite nodes to receive and train the lightweight student model; The upload module is used by the selected satellite node to upload the trained lightweight student model to the ground station within the satellite-to-ground visibility window; The aggregation module is used by the ground station to aggregate the received lightweight student models and perform batch normalization statistics calibration on the aggregated lightweight student models to obtain the next round of global student models. When the configuration is enabled, the ground station uses the next round of global student models to back-distill and update the teacher models.

[0044] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0045] Example 3 A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a lightweight federated learning method for low-Earth orbit (LEO) satellite networks. For example, the steps include: initializing a LEO satellite federated learning scenario; a ground station training a teacher model and distilling it to obtain a lightweight student model; the ground station selecting satellite nodes and distributing the lightweight student model to the selected satellite nodes; the selected satellite nodes receiving and training the lightweight student model; the selected satellite nodes uploading the trained lightweight student model to the ground station within a satellite-to-ground visibility window; the ground station aggregating the received lightweight student models and performing batch normalization statistics calibration on the aggregated lightweight student models to obtain the next-round global student model; and, when configuration is enabled, the ground station using the next-round global student model to back-distill and update the teacher model. The memory may include main memory, such as high-speed random access memory (RAM), or non-volatile memory, such as at least one disk storage device. The processor, network interface, and memory are interconnected via an internal bus, which may be an industry-standard architecture bus, a peripheral component interconnection standard bus, or an extended industry-standard architecture bus. The bus can be categorized as an address bus, data bus, or control bus. The memory stores programs; specifically, the program may include program code, which includes computer operation instructions. The memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0046] Example 4 A computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of a lightweight federated learning method for low-Earth orbit (LEO) satellite networks. For example, the program includes: initializing a LEO satellite federated learning scenario; a ground station training a teacher model and distilling it to obtain a lightweight student model; the ground station selecting satellite nodes and distributing the lightweight student model to the selected satellite nodes; the selected satellite nodes receiving and training the lightweight student model; the selected satellite nodes uploading the trained lightweight student model to the ground station within a satellite-to-ground visibility window; the ground station aggregating the received lightweight student models and performing batch normalization statistics calibration on the aggregated lightweight student models to obtain the next-round global student model; and, when configuration is enabled, the ground station using the next-round global student model to back-distill and update the teacher model. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0047] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0049] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0050] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0051] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0052] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0053] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A lightweight federated learning method for low-Earth orbit satellite networks, characterized in that, include: Initialize the low-Earth orbit satellite federated learning scenario; The ground station trains the teacher model, and distills it to obtain a lightweight student model; The ground station selects a satellite node and sends the lightweight student model to the selected satellite node; The selected satellite node receives and trains the lightweight student model; The selected satellite node uploads the trained lightweight student model to the ground station within the satellite-to-ground visibility window; The ground station aggregates the received lightweight student models and performs batch normalization statistics calibration on the aggregated lightweight student models to obtain the next round of global student models. When the configuration is enabled, the ground station uses the next round of global student models to back-distill and update the teacher models.

2. The lightweight federated learning method for low-Earth orbit satellite networks according to claim 1, characterized in that, The process of initializing the low-Earth orbit satellite federated learning scenario is as follows: The ground station acts as the federated learning server, and multiple satellite nodes act as clients. Each satellite node stores its local observation data. The ground station maintains the global student model and teacher model and records link metrics, including satellite node distance, uplink and downlink rates, lightweight student model size, visibility window time, and communication feasibility.

3. The lightweight federated learning method for low-Earth orbit satellite networks according to claim 1, characterized in that, The process of training the teacher model at the ground station and distilling it to obtain the lightweight student model is as follows: the ground station trains the teacher model using representative data, and then uses the teacher model output and real labels to jointly guide the training of the student model, thus obtaining the lightweight student model.

4. The lightweight federated learning method for low-Earth orbit satellite networks according to claim 1, characterized in that, The visible window constraint is represented as follows: in, The total transmission time required to complete one model distribution and upload, To keep the student model size lightweight, and These are the downlink speed and the uplink speed, respectively. The duration of the visible window.

5. The lightweight federated learning method for low-Earth orbit satellite networks according to claim 1, characterized in that, The loss function for the selected satellite node in the process of receiving and training the lightweight student model is: in, Used to constrain the prediction of the true label by the lightweight student model. Used to constrain the consistency between lightweight student models and snapshot soft objectives or teacher knowledge. The weights are adjustable.

6. The lightweight federated learning method for low-Earth orbit satellite networks according to claim 1, characterized in that, The process by which the ground station aggregates the received lightweight student models is as follows: the ground station performs sample-weighted aggregation on the received lightweight student models, wherein... in, These are the global student model parameters after aggregation in the (r+1)th round. For the lightweight student model parameters uploaded by the k-th participating satellite node, This represents the number of local samples corresponding to that satellite node. This refers to the set of satellite nodes participating in the aggregation in the current round. For set The total number of local samples across all satellite nodes.

7. The lightweight federated learning method for low-Earth orbit satellite networks according to claim 1, characterized in that, The teacher model uses an image classification network deployed on a ground station, while the lightweight student model uses a lightweight convolutional neural network deployed on a satellite node with fewer parameters than the teacher model. Only the parameters of the lightweight student model are transmitted in the satellite-to-ground link.

8. A lightweight federated learning system for low-Earth orbit satellite networks, characterized in that, include: The initialization module is used to initialize the federated learning scenario for low-Earth orbit satellites. The first training module is used to train the teacher model at the ground station and distill it to obtain a lightweight student model; The selection module is used by the ground station to select satellite nodes and send the lightweight student model to the selected satellite nodes. The second training module is used for the selected satellite nodes to receive and train the lightweight student model; The upload module is used by the selected satellite node to upload the trained lightweight student model to the ground station within the satellite-to-ground visibility window; The aggregation module is used by the ground station to aggregate the received lightweight student models and perform batch normalization statistics calibration on the aggregated lightweight student models to obtain the next round of global student models. When the configuration is enabled, the ground station uses the next round of global student models to back-distill and update the teacher models.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the lightweight federated learning method for low-Earth orbit satellite networks as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the lightweight federated learning method for low-Earth orbit satellite networks as described in any one of claims 1-7.

Images