Satellite remote control instruction automatic generation system based on artificial intelligence

Abstract

The invention discloses a satellite remote control instruction automatic generation system based on artificial intelligence, and relates to the technical field of satellite remote control instruction generation. A dynamic value evaluation module of the satellite remote control instruction automatic generation system fuses multi-source heterogeneous data, breaks through the rigidity of a traditional fixed priority strategy, and realizes real-time dynamic balance of task values; the conflict resolution module converts inter-satellite resource competition into a graph structure optimization problem, and effectively solves a hidden instruction mutual phagy phenomenon in multi-satellite collaborative observation; constructing a cross-model influence relation chain by the knowledge graph database, and eliminating the hidden conflict risk of the heterogeneous satellite instruction template; and the physical rule engine embeds dynamic constraints into a decision process to prevent generation of dangerous instructions violating the physical rules of the spacecraft.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite remote control instruction generation, in particular to an artificial intelligence-based satellite remote control instruction automatic generation system. Background Art

[0002] With the rapid expansion of low-orbit satellite constellations, multi-satellite collaborative observation has become the core mission model for scenarios such as ocean monitoring and disaster warning. The current mainstream system adopts a distributed architecture, with each satellite independently generating a sequence of instructions and achieving basic collaboration through inter-satellite links. The new generation of remote sensing constellations generally introduces dynamic priority scheduling algorithms to allocate observation tasks based on the real-time position and energy status of satellites. However, this type of system is difficult to cope with the global conflicts that emerge in complex scenarios at the constellation level.

[0003] To mitigate multi-satellite command conflicts, recent research has focused on strategies that combine online optimization with rule engines. Some systems use graph neural networks to model intersatellite influences and generate conflict prediction models through offline training. Others incorporate lightweight game theory frameworks to precompute resource usage matrices during the command generation phase. While these methods can effectively identify explicit conflicts, they are less effective at identifying implicit problems such as command cannibalization. When multiple satellites respond to the same emergency, the system may generate a large number of homogeneous commands simultaneously, resulting in wasted observation resources and energy overload.

[0004] Some systems have introduced digital twin technology for command rehearsal, but the simulation granularity mostly remains at the physical device layer and lacks in-depth modeling of the task logic layer; existing conflict resolution algorithms mostly rely on manually defined rule bases, and when faced with sudden multi-satellite coordination needs, ground stations still need to perform manual strategy selection and parameter adjustment. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides an artificial intelligence-based satellite remote control command automatic generation system to solve the problem that the existing system relies on static rules and manual intervention, is difficult to deal with dynamic command conflicts and heterogeneous equipment adaptation in multi-satellite collaborative observation, and restricts the expansion of constellation scale.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] The embodiment of the present invention provides a satellite remote control command automatic generation system based on artificial intelligence, which includes:

[0009] Dynamic value assessment module, which receives satellite real-time status data and mission requirement parameters and generates a multi-satellite mission priority sequence;

[0010] The conflict resolution module is connected to the dynamic value assessment module to eliminate command conflicts based on the inter-satellite resource occupation relationship map;

[0011] The spatiotemporal prediction module analyzes spatial environment monitoring data streams in real time and predicts abnormal events;

[0012] Knowledge graph library, which stores the impact relationship chains between heterogeneous satellite subsystems;

[0013] Physical rule engine, embedded with satellite dynamic constraints;

[0014] Federated learning architecture enables secure updates of multi-node model parameters.

[0015] As a preferred solution of the satellite remote control command automatic generation system based on artificial intelligence of the present invention, the dynamic value evaluation module includes:

[0016] Satellite remaining power sensor interface, receiving battery voltage and temperature data;

[0017] The task timeliness calculation unit calculates the observation urgency coefficient based on the time and space coordinates of the target event;

[0018] The multi-head attention computing layer performs feature fusion on the energy status, orbital phase, and historical imaging quality of at least three satellites and outputs dynamic priority weights.

[0019] As a preferred solution of the artificial intelligence-based satellite remote control command automatic generation system of the present invention, the operation of the conflict resolution module includes:

[0020] Construct an inter-satellite influence graph, where nodes represent satellite observation tasks and edge weights represent the communication frequency band overlap and attitude adjustment range conflict values;

[0021] Use graph attention network to iteratively update node feature vectors to generate the optimal task allocation solution;

[0022] When an energy overload risk is detected, a power management sub-command sequence is automatically inserted.

[0023] As a preferred solution of the artificial intelligence-based satellite remote control command automatic generation system of the present invention, the step of constructing the inter-satellite influence diagram in the conflict resolution module includes:

[0024] Map each satellite observation task to a node set N and construct an undirected weighted graph:

[0025] G=(N,E,W),

[0026] Where N represents the set of all observed task nodes, i, j∈N represents the i-th and j-th tasks respectively, E represents the set of edges between nodes, when there is a conflict between tasks i and j, (i, j)∈E, W is the edge weight function, defined as W(i, j)=C ij , where C ij represents the comprehensive conflict value between the i-th and j-th tasks;

[0027] The conflict value is linearly weighted, and the formula is:

[0028] C ij =αB ij +βA ij ,

[0029] Among them, C ij represents the comprehensive conflict value, α, β are normalized weight coefficients, satisfying α+β=1, B ij Indicates the degree of overlap of communication frequency bands, A ij Indicates the attitude adjustment range conflict value;

[0030] The communication band overlap is quantified as:

[0031] in, are the lower and upper limits of the i-th satellite communication frequency band, are the lower limit and upper limit of the j-th satellite communication frequency band respectively;

[0032] The attitude adjustment range conflict value is quantified as:

[0033]

[0034] Among them, c i ,c j are the attitude adjustment angles required for the i-th and j-th satellites to perform observation missions, respectively. Z is the maximum attitude adjustment angle range preset by the system, which is used for normalization processing.

[0035] As a preferred solution of the satellite remote control command automatic generation system based on artificial intelligence of the present invention, the spatiotemporal prediction module includes:

[0036] The causal convolutional neural network layer takes as input the spatial particle flux and geomagnetic index time series data stream;

[0037] Sliding time window processor, outputs the probability of abnormal events occurring in the next 5-10 minutes;

[0038] Pre-compiled instruction cache, dynamically loading emergency instruction templates based on predicted probabilities.

[0039] As a preferred solution of the satellite remote control command automatic generation system based on artificial intelligence described in the present invention, the method for constructing the knowledge graph library includes:

[0040] Parsing command-response pairs in satellite interface control documents;

[0041] The random walk algorithm is used to mine the hidden impact paths of different satellite models;

[0042] The confidence level and minimum safety time interval parameters of the storage instruction conflict chain are stored.

[0043] As a preferred solution of the satellite remote control command automatic generation system based on artificial intelligence described in the present invention, wherein: in the knowledge graph library, the heterogeneous satellite subsystems and their influence chains are represented as a weighted graph G K =(V,E K ,W K ), where W K is the edge weight function, and w ij =W K (i, j), and based on the random walk restart algorithm RWR, the implicit impact path between different satellite models is mined. The specific process includes: constructing a node transfer probability matrix P on the weighted graph, where each element is obtained by normalizing the edge weight, and let P be the node transfer probability matrix on the graph, whose elements are P ij :

[0044]

[0045] Among them, P ij represents the probability of transferring from node i to node j, w ij represents the edge weight between nodes i and j, represents the set of all nodes adjacent to node i, and k is the index of the adjacent node;

[0046] A random walk is initiated from the source node s, and the node influence probability vector is updated in each iteration as follows:

[0047]

[0048] Among them, p (t) represents the node impact probability distribution vector of the t-th iteration, γ represents the restart probability, satisfying 0<γ<1, P T represents the transpose of the transfer matrix, e s represents the standard basis vector with the corresponding component of the source node s being 1 and the other components being 0, t represents the number of iterations, and s represents the starting node of the walk;

[0049] When the probability vector converges, it satisfies:

[0050] |p(t+1) -p (t) |1<∈,

[0051] Among them, |·|1 represents the L1 norm, ∈ represents the preset convergence threshold;

[0052] After convergence, nodes with a probability exceeding the threshold τ are selected according to the steady-state distribution to obtain the set of key influencing nodes:

[0053]

[0054] in, Represents the filtered key node set, p i represents the component of the i-th node in the steady-state probability vector, and τ represents the preset probability threshold.

[0055] As a preferred solution of the satellite remote control command automatic generation system based on artificial intelligence described in the present invention, the physical rule engine includes:

[0056] Orbital mechanics constraint unit, verifies whether the command parameters are consistent with the current orbital phase;

[0057] Thermal control boundary check unit, comparing instruction triggering actions with device temperature thresholds;

[0058] A differentiable validation layer transforms the satellite dynamics equation into a neural network regularization term.

[0059] As a preferred solution of the artificial intelligence-based satellite remote control command automatic generation system of the present invention, the operations of the federated learning architecture include:

[0060] When training the local model, a differential privacy algorithm is used to add Gaussian noise;

[0061] During the parameter aggregation phase, the model hash value is verified through blockchain smart contracts;

[0062] Before model deployment, perform adversarial sample stress testing in the digital twin environment.

[0063] As a preferred solution of the satellite remote control command automatic generation system based on artificial intelligence described in the present invention, wherein: in the federated learning architecture, a Gaussian mechanism is introduced to update the gradient vector g of the local model of each client. i The steps for cropping and injecting noise include:

[0064] The norm of the local gradient is clipped to limit the sensitivity. The formula is:

[0065]

[0066] Among them, g irepresents the local model update gradient vector of the i-th client, C represents the gradient clipping threshold norm, which is used to limit the gradient sensitivity, and |·|2 represents the vector bi-norm operation;

[0067] Adding Gaussian noise to the clipped gradients yields a differentially private update:

[0068]

[0069] in, represents the gradient vector after adding noise, σ represents the noise scaling factor, I represents the unit matrix of the same dimension as the gradient vector, (σC) 2 is the noise variance;

[0070] According to Gaussian mechanism theory, the noise scaling factor satisfies:

[0071]

[0072] Where ∈ represents the privacy budget, which measures the strength of privacy protection, δ represents the upper limit of the failure probability allowed by differential privacy, and ln(·) represents the natural logarithm operation.

[0073] The beneficial effects of the present invention are as follows: the present invention significantly improves the autonomous control capability of satellite clusters through multi-dimensional technological innovation; the dynamic value assessment module integrates multi-source heterogeneous data, breaks through the rigidity of traditional fixed priority strategies, and realizes real-time dynamic weighing of mission values; the conflict resolution module transforms the competition for inter-satellite resources into a graph structure optimization problem, effectively solving the implicit instruction cannibalization phenomenon in multi-satellite collaborative observation; the knowledge graph library constructs a cross-model influence relationship chain to eliminate the implicit conflict risk of heterogeneous satellite instruction templates; the physical rule engine embeds dynamic constraints into the decision-making process to prevent the generation of dangerous instructions that violate the physical laws of spacecraft; the federated learning architecture realizes the collaborative evolution of multi-node models while ensuring data privacy, so that the system has the ability to continuously adapt to the entry of new satellites into the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0075] Figure 1 This is a schematic diagram of the framework of the satellite remote control command automatic generation system based on artificial intelligence in Example 1. DETAILED DESCRIPTION

[0076] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0077] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0078] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0079] Example 1, with reference to Figure 1 This embodiment provides a satellite remote control command automatic generation system based on artificial intelligence, including:

[0080] Dynamic value assessment module, which receives satellite real-time status data and mission requirement parameters and generates a multi-satellite mission priority sequence;

[0081] The dynamic value assessment module includes:

[0082] Satellite remaining power sensor interface, receiving battery voltage and temperature data;

[0083] The task timeliness calculation unit calculates the observation urgency coefficient based on the time and space coordinates of the target event;

[0084] The multi-head attention calculation layer performs feature fusion on the energy status, orbit phase, and historical imaging quality of at least three satellites and outputs dynamic priority weights;

[0085] The conflict resolution module is connected to the dynamic value assessment module to eliminate command conflicts based on the inter-satellite resource occupation relationship map;

[0086] The operations of the conflict resolution module include:

[0087] Construct an inter-satellite influence graph, where nodes represent satellite observation tasks and edge weights represent the communication frequency band overlap and attitude adjustment range conflict values;

[0088] Use graph attention network to iteratively update node feature vectors to generate the optimal task allocation solution;

[0089] Automatically insert power management sub-command sequences when energy overload risk is detected;

[0090] The steps for constructing the intersatellite influence diagram in the conflict resolution module include:

[0091] Map each satellite observation task to a node set N and construct an undirected weighted graph:

[0092] G=(N,E,W),

[0093] Where N represents the set of all observed task nodes, i, j∈N represents the i-th and j-th tasks respectively, E represents the set of edges between nodes, when there is a conflict between tasks i and j, (i, j)∈E, W is the edge weight function, defined as W(i, j)=C ij , where C ij represents the comprehensive conflict value between the i-th and j-th tasks;

[0094] The conflict value is linearly weighted, and the formula is:

[0095] C ij =αB ij +βA ij ,

[0096] Among them, C ij represents the comprehensive conflict value, α, β are normalized weight coefficients, satisfying α+β=1, B ij Indicates the degree of overlap of communication frequency bands, A ij Indicates the attitude adjustment range conflict value;

[0097] The communication band overlap is quantified as:

[0098] in, are the lower and upper limits of the i-th satellite communication frequency band, are the lower limit and upper limit of the j-th satellite communication frequency band respectively;

[0099] The attitude adjustment range conflict value is quantified as:

[0100]

[0101] Among them, c i ,c j are the attitude adjustment angles required for the i-th and j-th satellites to perform observation missions, respectively. Z is the maximum attitude adjustment angle range preset by the system, which is used for normalization processing.

[0102] Specifically, the communication band overlap and attitude difference are converted into comparable conflict indicators to achieve a unified measurement of resource conflict. The frequency band overlap is based on the ratio of spectrum intersection to minimum bandwidth, which can reflect the intensity of inter-satellite communication interference. The attitude conflict value is normalized by the angle difference to characterize the attitude adjustment risk. The linear weighting introduces a weight coefficient, allowing the system to flexibly adjust the contribution ratio of the two parts according to mission requirements.

[0103] The spatiotemporal prediction module analyzes spatial environment monitoring data streams in real time and predicts abnormal events;

[0104] The spatiotemporal prediction module includes:

[0105] The causal convolutional neural network layer takes as input the spatial particle flux and geomagnetic index time series data stream;

[0106] Sliding time window processor, outputs the probability of abnormal events occurring in the next 5-10 minutes;

[0107] Pre-compiled instruction cache, dynamically loading emergency instruction templates based on predicted probabilities;

[0108] Knowledge graph library, which stores the impact relationship chains between heterogeneous satellite subsystems;

[0109] The construction methods of the knowledge graph library include:

[0110] Parsing command-response pairs in satellite interface control documents;

[0111] The random walk algorithm is used to mine the hidden impact paths of different satellite models;

[0112] Confidence level and minimum safety time interval parameters of storage instruction conflict chain;

[0113] In the knowledge graph database, heterogeneous satellite subsystems and their impact chains are represented as weighted graphs G K =(V,E K ,W K ), where W K is the edge weight function, and w ij =W K (i, j), and based on the random walk restart algorithm RWR, the implicit impact path between different satellite models is mined. The specific process includes: constructing a node transfer probability matrix P on the weighted graph, where each element is obtained by normalizing the edge weight, and let P be the node transfer probability matrix on the graph, whose elements are P ij :

[0114]

[0115] Among them, P ij represents the probability of transferring from node i to node j, w ij represents the edge weight between nodes i and j, represents the set of all nodes adjacent to node i, and k is the index of the adjacent node;

[0116] A random walk is initiated from the source node s, and the node influence probability vector is updated in each iteration as follows:

[0117]

[0118] Among them, p (t) represents the node impact probability distribution vector of the t-th iteration, γ represents the restart probability, satisfying 0<γ<1, P T represents the transpose of the transfer matrix, e s represents the standard basis vector with the corresponding component of the source node s being 1 and the other components being 0, t represents the number of iterations, and s represents the starting node of the walk;

[0119] When the probability vector converges, it satisfies:

[0120] |p (t+1) -p (t) |1<∈,

[0121] Among them, |·|1 represents the L1 norm, ∈ represents the preset convergence threshold;

[0122] After convergence, nodes with a probability exceeding the threshold τ are selected according to the steady-state distribution to obtain the set of key influencing nodes:

[0123]

[0124] in, Represents the filtered key node set, p i represents the component of the i-th node in the steady-state probability vector, and τ represents the preset probability threshold;

[0125] Specifically, the random walk restart algorithm can measure the influence of each node on the source node on a global scale. The restart probability takes into account both local and global information and mines non-obvious cross-model influence links. The transfer probability matrix normalizes the original edge weights to ensure the comparability of influence weights between different models. The convergence criterion and threshold screening provide stability and controllability for path extraction. It does not rely on preset path lengths and can automatically discover influence paths of varying lengths.

[0126] Physical rule engine, embedded with satellite dynamic constraints;

[0127] The physics rules engine includes:

[0128] Orbital mechanics constraint unit, verifies whether the command parameters are consistent with the current orbital phase;

[0129] Thermal control boundary check unit, comparing instruction triggering actions with device temperature thresholds;

[0130] A differentiable validation layer that transforms the satellite dynamics equation into a neural network regularization term;

[0131] Federated learning architecture to achieve secure updates of multi-node model parameters;

[0132] The operations of the federated learning architecture include:

[0133] When training the local model, a differential privacy algorithm is used to add Gaussian noise;

[0134] During the parameter aggregation phase, the model hash value is verified through blockchain smart contracts;

[0135] Before model deployment, perform adversarial example stress testing in the digital twin environment;

[0136] In the federated learning architecture, the Gaussian mechanism is introduced to update the gradient vector g for each client's local model. i The steps for cropping and injecting noise include:

[0137] The norm of the local gradient is clipped to limit the sensitivity. The formula is:

[0138]

[0139] Among them, g i represents the local model update gradient vector of the i-th client, C represents the gradient clipping threshold norm, which is used to limit the gradient sensitivity, and |·|2 represents the vector bi-norm operation;

[0140] Adding Gaussian noise to the clipped gradients yields a differentially private update:

[0141]

[0142] in, represents the gradient vector after adding noise, σ represents the noise scaling factor, I represents the unit matrix of the same dimension as the gradient vector, (σC) 2 is the noise variance;

[0143] According to Gaussian mechanism theory, to satisfy (∈,δ) differential privacy, the noise scaling factor satisfies:

[0144]

[0145] Where ∈ represents the privacy budget, which measures the strength of privacy protection, δ represents the upper limit of the failure probability allowed by differential privacy, and ln(·) represents the natural logarithm operation;

[0146] Specifically, the Gaussian mechanism first limits the sensitivity of each client's update to a constant C by clipping, ensuring that different clients contribute the same amount to privacy leakage, and then injects a variance of (σC) 2 Gaussian noise is used to ensure that each update satisfies the (∈, δ) differential privacy guarantee. The relationship between the noise factor σ, the privacy budget ∈, and the failure probability δ is the theoretical minimum lower bound, which can minimize the noise intensity under the established privacy requirements, thereby taking into account the model performance.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based satellite remote control command automatic generation system, characterized in that: include, Dynamic value assessment module, which receives satellite real-time status data and mission requirement parameters and generates a multi-satellite mission priority sequence; The conflict resolution module is connected to the dynamic value assessment module to eliminate command conflicts based on the inter-satellite resource occupation relationship map; The spatiotemporal prediction module analyzes spatial environment monitoring data streams in real time and predicts abnormal events; Knowledge graph library, which stores the impact relationship chains between heterogeneous satellite subsystems; Physical rule engine, embedded with satellite dynamic constraints; Federated learning architecture enables secure updates of multi-node model parameters.

2. The artificial intelligence-based satellite remote control command automatic generation system according to claim 1, characterized in that: The dynamic value assessment module includes: Satellite remaining power sensor interface, receiving battery voltage and temperature data; The task timeliness calculation unit calculates the observation urgency coefficient based on the time and space coordinates of the target event; The multi-head attention computing layer performs feature fusion on the energy status, orbital phase, and historical imaging quality of at least three satellites and outputs dynamic priority weights.

3. The artificial intelligence-based satellite remote control command automatic generation system according to claim 1, characterized in that: The operations of the conflict resolution module include: Construct an inter-satellite influence graph, where nodes represent satellite observation tasks and edge weights represent the communication frequency band overlap and attitude adjustment range conflict values; Use graph attention network to iteratively update node feature vectors to generate the optimal task allocation solution; When an energy overload risk is detected, a power management sub-command sequence is automatically inserted.

4. The artificial intelligence-based satellite remote control command automatic generation system according to claim 3, characterized in that: The step of constructing the intersatellite influence diagram in the conflict resolution module includes: Map each satellite observation task to a node set N and construct an undirected weighted graph: G=(N,E,W), Where N represents the set of all observed task nodes, i, j∈N represents the i-th and j-th tasks respectively, E represents the set of edges between nodes, when there is a conflict between tasks i and j, (i, j)∈E, W is the edge weight function, defined as W(i, j)=C ij , where C ij represents the comprehensive conflict value between the i-th and j-th tasks; The conflict value is linearly weighted, and the formula is: C ij =αB ij +βA ij , Among them, C ij represents the comprehensive conflict value, α, β are normalized weight coefficients, satisfying α+β=1, B ij Indicates the degree of overlap of communication frequency bands, A ij Indicates the attitude adjustment range conflict value; The communication band overlap is quantified as: in, are the lower and upper limits of the i-th satellite communication frequency band, are the lower limit and upper limit of the j-th satellite communication frequency band respectively; The attitude adjustment range conflict value is quantified as: Among them, c i ,c j are the attitude adjustment angles required for the i-th and j-th satellites to perform observation missions, respectively. Z is the maximum attitude adjustment angle range preset by the system, which is used for normalization processing.

5. The artificial intelligence-based satellite remote control command automatic generation system according to claim 1, characterized in that: The spatiotemporal prediction module comprises: The causal convolutional neural network layer takes as input the spatial particle flux and geomagnetic index time series data stream; Sliding time window processor, outputs the probability of abnormal events occurring in the next 5-10 minutes; Pre-compiled instruction cache, dynamically loading emergency instruction templates based on predicted probabilities.

6. The artificial intelligence-based satellite remote control command automatic generation system according to claim 1, characterized in that: The method for constructing the knowledge graph library includes: Parsing command-response pairs in satellite interface control documents; The random walk algorithm is used to mine the hidden impact paths of different satellite models; The confidence level and minimum safety time interval parameters of the storage instruction conflict chain are stored.

7. The artificial intelligence-based satellite remote control command automatic generation system according to claim 6, characterized in that: In the knowledge graph library, heterogeneous satellite subsystems and their influence chains are represented as weighted graphs G K =(V,E K ,W K ), where W K is the edge weight function, and w ij =W K (i, j), and based on the random walk restart algorithm RWR, the implicit impact path between different satellite models is mined. The specific process includes: constructing a node transfer probability matrix P on the weighted graph, where each element is obtained by normalizing the edge weight, and let P be the node transfer probability matrix on the graph, whose elements are P ij : Among them, P ij represents the probability of transferring from node i to node j, w ij represents the edge weight between nodes i and j, represents the set of all nodes adjacent to node i, and k is the index of the adjacent node; A random walk is initiated from the source node s, and the node influence probability vector is updated in each iteration as follows: Among them, p (t) represents the node impact probability distribution vector of the t-th iteration, γ represents the restart probability, satisfying 0<γ<1, P T represents the transpose of the transfer matrix, e s represents the standard basis vector with the corresponding component of the source node s being 1 and the other components being 0, t represents the number of iterations, and s represents the starting node of the walk; When the probability vector converges, it satisfies: |p (t+1) -p (t) |1<∈, Among them, |·|1 represents the L1 norm, ∈ represents the preset convergence threshold; After convergence, nodes with a probability exceeding the threshold τ are selected according to the steady-state distribution to obtain the set of key influencing nodes: in, Represents the filtered key node set, p i represents the component of the i-th node in the steady-state probability vector, and τ represents the preset probability threshold.

8. The artificial intelligence-based satellite remote control command automatic generation system according to claim 1, characterized in that: The physical rule engine includes: Orbital mechanics constraint unit, verifies whether the command parameters are consistent with the current orbital phase; Thermal control boundary check unit, comparing instruction triggering actions with device temperature thresholds; A differentiable validation layer transforms the satellite dynamics equation into a neural network regularization term.

9. The artificial intelligence-based satellite remote control command automatic generation system according to claim 1, characterized in that: The operations of the federated learning architecture include: When training the local model, a differential privacy algorithm is used to add Gaussian noise; During the parameter aggregation phase, the model hash value is verified through blockchain smart contracts; Before model deployment, perform adversarial sample stress testing in the digital twin environment.

10. The satellite remote control command automatic generation system based on artificial intelligence according to claim 9, characterized in that: In the federated learning architecture, a Gaussian mechanism is introduced to update the gradient vector g of each client's local model i The steps for cropping and injecting noise include: The norm of the local gradient is clipped to limit the sensitivity. The formula is: Among them, g i represents the local model update gradient vector of the i-th client, C represents the gradient clipping threshold norm, which is used to limit the gradient sensitivity, and |·|2 represents the vector bi-norm operation; Adding Gaussian noise to the clipped gradients yields a differentially private update: in, represents the gradient vector after adding noise, σ represents the noise scaling factor, I represents the unit matrix of the same dimension as the gradient vector, (σc) 2 is the noise variance; According to Gaussian mechanism theory, the noise scaling factor satisfies: Where ∈ represents the privacy budget, which measures the strength of privacy protection, δ represents the upper limit of the failure probability allowed by differential privacy, and ln(·) represents the natural logarithm operation.

Images