Low-orbit aircraft instruction cluster generation method, system and equipment based on semantic mapping and storage medium

Abstract

The invention provides a semantic mapping-based low-orbit aircraft instruction cluster generation method, which comprises the following steps of: analyzing a satellite task demand and a satellite real-time state through a semantic mapping module, calculating a matching degree between a task demand vector T and a satellite state vector S, generating a preliminary task instruction set, and performing space-time dependent modeling; the method comprises the steps of optimizing a task execution sequence and resource allocation, adjusting a task scheduling strategy in real time based on a reinforcement learning algorithm, verifying the integrity and security of an instruction through a smart contract, and the like. The core problems of insufficient intelligence, lack of space-time dependent modeling, poor dynamic adaptability, low safety and the like in satellite task scheduling are systematically solved, and the task execution efficiency, the resource utilization rate and the system reliability are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of low-orbit aircraft control, and in particular to a low-orbit aircraft instruction cluster generation method, system, device and storage medium based on semantic mapping. BACKGROUND

[0002] With the rapid development of low-orbit aircraft technology, satellite task complexity has been significantly improved, involving remote sensing observation, communication relay, environmental monitoring and other application scenarios. Efficient scheduling and instruction generation of satellite tasks have become the core link to ensure task success rate and resource utilization. However, the existing satellite task instruction generation and scheduling technology still has significant shortcomings, which is difficult to meet the modern demand of dynamic, intelligent and high security, which is specifically shown as follows: 1. Insufficient automation and intelligence. Traditional satellite task instruction generation highly depends on manual experience or preset strategies based on static rules, which cannot respond to satellite state (such as orbit deviation, energy fluctuation, load failure) and sudden task demand in real time. Manual intervention is not only inefficient (according to statistics, complex task scheduling can take several hours), but also prone to errors due to human error. Although some researches try to introduce rule engine or expert system, its flexibility is limited, which is difficult to adapt to the changing space environment and task scenario; 2. Weak ability to model space-time dependence. There is a complex space-time correlation between satellite task instructions, for example: observation tasks need to avoid overlapping of orbit coverage areas, communication tasks need to meet the time window matching of relay satellites, etc. Although the existing scheduling methods (such as static modeling based on directed acyclic graph DAG) can handle simple dependence, they lack dynamic modeling ability for space-time coupling characteristics, resulting in frequent resource allocation conflicts (experimental results show that the conflict rate can be more than 30%), and significant increase in task execution delay (average delay more than 120 seconds); 3. Insufficient resource scheduling accuracy and adaptability. Satellite task execution needs to coordinate energy, computing power, communication bandwidth and other multi-dimensional resources. Traditional methods use fixed threshold allocation strategy, which cannot be dynamically adjusted according to task priority or real-time resource margin, resulting in resource waste (such as energy redundancy rate up to 40%) or failure of key tasks due to insufficient resources. In addition, the application of existing reinforcement learning technology in satellite scheduling is mostly limited to single task scenario, which is difficult to solve the resource competition and constraint optimization problem in multi-task parallel environment; 4. Instruction generation and execution security deficiency. The generation, transmission and execution process of satellite mission instructions faces the risk of data tampering, forgery or leakage. Existing technologies rely on traditional encryption methods and lack full-link trusted verification mechanisms. Studies have shown that unprotected instruction links can be attacked successfully at a rate of 15%, seriously affecting mission reliability. Although blockchain technology has been initially introduced to improve data security, its application in real-time scheduling scenarios still has efficiency bottlenecks (single instruction verification takes more than 1 second), making it difficult to meet high-concurrency task requirements; 5. Insufficient scheduling adaptability in dynamic environments. The satellite operating environment is highly dynamic, such as sudden disaster monitoring tasks that need to be inserted immediately, and satellite payload failure that needs to be quickly switched to a backup solution. Existing scheduling systems rely on offline planning and lack real-time feedback and adaptive adjustment capabilities.

[0003] In recent years, academia and industry have tried to optimize satellite scheduling through various technologies, but all have obvious defects: existing satellite task instruction generation and scheduling technologies have the following problems: 1) Rule-based and manual scheduling instruction generation: Instruction generation relies on static rules or manual coding, which is suitable for scheduling simple tasks, but is difficult to adapt to complex task requirements and changing satellite states, and cannot handle dynamic task environments; 2) Scheduling method based on graph model: Task scheduling is performed by constructing a task dependency graph (DAG), which can handle task dependencies, but lacks the ability to model temporal and spatial dependencies, and cannot fully optimize the allocation of spatial and temporal resources for satellite tasks; 3) Application of reinforcement learning in task scheduling: Although reinforcement learning can be used to solve scheduling problems, existing applications are mainly limited to simple scenarios and cannot solve the problems of temporal and spatial constraints and task priority scheduling in complex multi-task environments; 4) Application of blockchain in task instruction security: Blockchain technology has been used to ensure the security of task instructions, but its application in task scheduling is still preliminary and cannot achieve real-time and efficiency in task scheduling and resource management;

[0004] In summary, existing technologies cannot balance the intelligence, accuracy, security and dynamic adaptability of satellite task scheduling, and an innovative solution that combines the advantages of multiple technologies is urgently needed. The present invention addresses the above core problems by proposing a low-orbit aircraft instruction cluster generation method based on semantic mapping, which systematically improves the overall performance of satellite task scheduling through multi-technology collaboration and algorithm optimization. SUMMARY

[0005] The present invention proposes a method, method, and device for generating low-orbit aircraft instruction clusters based on semantic mapping, which solves the problems of low control accuracy and insufficient comprehensive scheduling performance of low-orbit aircraft in the prior art. The technical solution of the present invention is achieved as follows: A method for generating a low-orbit aircraft instruction cluster based on semantic mapping comprises the following steps: Step S1: Analyze the satellite mission requirements and satellite real-time status through the semantic mapping module to generate the mission requirement vector T and satellite state vector S, where: The task demand vector T includes the timeliness, spatial coverage and energy consumption characteristics of the task; The satellite state vector S includes the satellite's orbital altitude, payload status, and energy consumption characteristics; Step S2: Calculate the matching degree between the mission requirement vector T and the satellite state vector S, and generate a preliminary mission instruction set; Step S3: Using a spatiotemporal graph convolutional network (ST-GCN) to model the spatiotemporal dependencies of the preliminary task instruction set, and optimize the task execution sequence and resource allocation; Step S4: Adjust the task scheduling strategy in real time based on the reinforcement learning algorithm, and update the instruction execution order according to the satellite status changes and task priorities; Step S5: Simulate the task execution process through digital twin technology to generate an optimized task instruction cluster; Step S6: Use blockchain technology to hash and encrypt the task instruction cluster, and verify the integrity and security of the instruction through smart contracts.

[0006] As a preferred technical solution, the task requirement vector T in step S1 is expressed as:

[0007] Among them, t i represents the i-th feature of the task, including task timeliness, spatial coverage, and energy consumption; The state vector S of the satellite is:

[0008] Among them, S i represents the i-th state characteristic of the satellite, including orbital altitude, payload status, and energy consumption; By calculating the similarity between the mission requirements and the satellite status, the matching degree is obtained by the following formula:

[0009] The matching degree between mission requirements and satellite status is calculated by cosine similarity to generate appropriate mission instructions.

[0010] As a preferred technical solution, the semantic mapping module in step S1 adopts a natural language processing (NLP) model to convert the unstructured task requirement text into a structured feature vector, thereby more efficiently abstracting and optimizing the task requirements.

[0011] As a preferred technical solution, the spatio-temporal graph convolution network (ST-GCN) in step S3 extracts spatio-temporal dependencies through the following graph convolution formula:

[0012] wherein, represents the representation of the jth task instruction at the kth layer, is the dependency relationship between instructions, is the convolution kernel weight, is the adjacency node set of instruction i, is the activation function.

[0013] As a preferred technical solution, the reinforcement learning algorithm in step S4 adopts a Q-learning algorithm, and the Q value update formula is:

[0014] wherein, Q(s,a) is the expected reward of performing action a in state s, R(s,a) is the immediate reward, γ is the discount factor, s' is the new state after execution, and a' is the subsequent action.

[0015] As a preferred technical solution, the digital twin technology in step S5 optimizes the task execution cost through the following objective function:

[0016] wherein, is the cost of task execution, is the decision variable of whether to execute the task instruction, A is the constraint condition matrix, and b is the constraint vector.

[0017] As a preferred technical solution, the blockchain technology in step S6 adopts the SHA-256 algorithm to generate a task instruction hash value, and verifies the consistency of the hash value and the execution result through a smart contract.

[0018] A satellite task instruction generation and scheduling system, comprising: a semantic mapping module, a spatio-temporal graph convolution network module, a reinforcement learning optimization module, a digital twin simulation module, and a blockchain security module.

[0019] A computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the steps of a low-orbit aircraft instruction cluster generation method based on semantic mapping.

[0020] An electronic device comprises a memory, a processor and a computer program stored on the memory, and the processor implements the steps of a low-orbit aircraft instruction cluster generation method based on semantic mapping when executing the program.

[0021] Compared with the prior art, the present scheme has the following beneficial effects: (1) High automation and intelligence: through semantic mapping and reinforcement learning, the present application can automatically generate and schedule task instructions, reduce manual intervention and improve task execution efficiency.

[0022] Space-time task optimization: ST-GCN is used to model the space-time dependence between task instructions, optimize the task execution sequence and ensure efficient and coordinated execution.

[0023] Safety guarantee for task execution: blockchain technology ensures the tamper-proof nature of task instructions during generation, transmission and execution, ensuring data security.

[0024] Dynamic task scheduling optimization: digital twin and simulation optimization technology provides a real-time feedback mechanism that can fine-tune scheduling and optimization for complex tasks.

[0025] Strong scalability and applicability: this technology is not only suitable for satellite task scheduling, but also widely applicable to task scheduling optimization in many fields such as unmanned driving and intelligent manufacturing. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0027] Figure 1 The flowchart of the present application, a low-orbit aircraft instruction cluster generation method based on semantic mapping; Figure 2 The method flowchart of the example of the low-orbit satellite instruction cluster generation method based on semantic mapping in the embodiment. DETAILED DESCRIPTION

[0028] The technical solutions of the present application will be described in detail below with reference to the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0029] REFERENCE Figure 1In this paper, a low-orbit aircraft instruction cluster generation method based on semantic mapping is invented. By combining semantic mapping, spatio-temporal graph convolution network (ST-GCN), reinforcement learning (RL), digital twin and blockchain technology, a comprehensive low-orbit aircraft task instruction generation and scheduling method is proposed. The method steps are as follows: 1. Semantic mapping and task demand analysis Satellite tasks are usually composed of multiple requirements, including timeliness, resource demand, task priority, etc. Through semantic mapping, we can convert task requirements and satellite state into vector representation, and use these vectors to generate optimized task instructions.

[0030] The representation of task demand vector T is:

[0031] Where ti represents the i-th feature of the task, such as task timeliness, spatial coverage, energy consumption, etc.

[0032] The state vector S of the satellite is:

[0033] Where, represents the i-th state feature of the satellite, such as orbit height, payload state, energy consumption, etc.

[0034] By calculating the similarity between task demand and satellite state, the matching degree can be obtained by the following formula:

[0035] This formula calculates the matching degree of task demand and satellite state through cosine similarity, and generates appropriate task instructions. On this basis, deep learning algorithms (such as BERT, GPT) can be introduced to abstract and optimize task requirements more efficiently.

[0036] 2. Spatio-temporal graph convolution network (ST-GCN) optimization of task scheduling

[0037] There is a complex spatio-temporal dependency between satellite task instructions, and ST-GCN can effectively capture these dependencies to optimize task scheduling. Each task instruction can be regarded as a node in the graph, and the dependency relationship between instructions is represented by the edge in the graph.

[0038] The operation formula of graph convolution is as follows:

[0039] Where, represents the representation of the j-th task instruction at the k-th layer, is the dependency relationship between instructions, is the convolution kernel weight, ) is the set of adjacent nodes for instruction i, and σ is the activation function.

[0040] Through spatio-temporal graph convolution, the system can optimize the order of task scheduling and resource usage, so that tasks are executed along the optimal path, improving execution efficiency and reducing resource conflicts.

[0041] 3. Reinforcement Learning (RL) Adaptive Optimization

[0042] The application of reinforcement learning in the present invention can help the system automatically adjust the scheduling strategy according to the task demand and the change of satellite state. Specifically, through the optimization of Q-learning algorithm, the Q value update formula is as follows:

[0043] Where Q(s,a) is the expected reward of performing action a in state s, R(s,a) is the immediate reward, γ is the discount factor, s' is the new state after execution, and a' is the subsequent action.

[0044] The reinforcement learning algorithm can adaptively optimize according to system feedback, and can adjust the generation strategy of task instructions in real time to ensure optimal task scheduling and resource allocation.

[0045] 4. Digital Twin and Simulation Optimization

[0046] Through digital twin technology, the system can create a virtual satellite task model and optimize the task execution process through real-time simulation. During task execution, the digital twin model can be updated in real time and feedback the optimization results. Task optimization is achieved through the following objective function:

[0047] Where, is the cost of task execution, is the decision variable of whether the task instruction is executed, A is the constraint condition matrix, and b is the constraint vector.

[0048] Digital twin technology can simulate potential problems in task execution in advance, and adjust the scheduling strategy through real-time feedback, thereby improving the accuracy and efficiency of task execution.

[0049] 5. Blockchain technology ensures the security of task instructions

[0050] In order to ensure the security of task instructions, the present invention encrypts and verifies task instructions through blockchain technology. The hash value generation formula of task instructions is: H(I)=SHA256(I) Wherein, H(I) is the hash value of the task instruction I. The execution process of each instruction is verified by the smart contract to ensure the transparency, tamper resistance and security of the instruction in the execution process.

[0051] Compared with the prior art, the scheme has the following substantial improvements and significant technical progress: (1) Efficiency and accuracy improvement: through the cooperation of semantic mapping and ST-GCN, the task instruction generation speed and scheduling accuracy are optimized simultaneously; (2) High robustness in dynamic environment: the closed-loop feedback mechanism of reinforcement learning and digital twinning ensures stable scheduling in complex scenarios; (3) Security and trust protection: blockchain technology provides full-link protection from generation to execution, solving the security pain points of traditional satellite tasks; (4) Technological combination innovation: deep integration of multiple technologies (NLP+ST-GCN+blockchain), forming a "1+1>2" synergistic effect, breaking through the bottleneck of single technology.

[0052] Through the deep integration of multiple technologies and algorithm customization improvement, the scheme systematically solves the core problems of insufficient intelligence, lack of spatiotemporal dependency modeling, poor dynamic adaptability, and low security in satellite task scheduling, significantly improving task execution efficiency, resource utilization, and system reliability.

[0053] Please refer to Figure 2 , in order to verify the above conclusion, the embodiment gives a detailed example of low-orbit satellite instruction cluster generation method based on semantic mapping as further support: (I) Task background and input information Task source: Ecological Environment Department Remote Sensing Monitoring Center Task submission time: May 8, 2025 Task target description (natural language): "Please arrange to use low-orbit remote sensing satellites to conduct multi-spectral imaging on the Nanjing section of the Yangtze River (longitude and latitude range: 118.5°E-119.3°E, 31.8°N-32.3°N) between Beijing time 10:30 and 11:00 on May 10, 2025, for analyzing water pollution. The image resolution should be better than 5 meters, the coverage area should be no less than 30 square kilometers, and the data should be returned to the ground station within 48 hours. The task priority is medium, if it cannot be completed due to orbital restrictions, please automatically postpone to the next available window.". (II) Semantic mapping and feature vector construction (1) Semantic analysis and extraction model: The improved BERT semantic model is adopted, combined with the task ontology word graph, to construct a multi-dimensional task feature extractor.

[0054] (2) Vector standardization encoding: T = [0.72, 0.95, 0.85, 0.90, 2.0, 1.0, 0.65] (Three) Satellite state digital twin modeling and S vector generation Through the digital twin system designed by the application, the simulation state parameters of the current satellite LQ-3 are obtained and converted into a state vector: S = [0.82, 1.0, 0.78, 0.68, 0.87, 0.90] (Four) Similarity calculation and candidate matching selection The cosine similarity is used to calculate the matching degree between the task and the satellite state, and the result is 0.89, which is greater than the threshold value 0.75, and it is judged to be schedulable.

[0055] (Five) Task graph modeling and ST-GCN graph neural network scheduling optimization

[0056] The task graph is constructed, and a multi-layer graph convolutional neural network is used for space-time dependence modeling, and the optimized scheduling result is: Execution time: 10:28:15 - 10:35:45; Resource consumption: power 8.5%, storage 11.2GB, attitude adjustment 36 seconds (Six) Reinforcement learning scheduling policy fine-tuning (DQN) The Deep Q-learning model is used to fine-tune the scheduling policy, and the output optimized Q value is 0.913, which is better than the original strategy.

[0057] (Seven) Digital twin system simulation verification

[0058] The simulation operations such as attitude adjustment, imaging process, data backhaul, etc. are executed, and the task execution success probability reaches 92.8%.

[0059] (Eight) Instruction cluster generation and blockchain encryption publication

[0060] The instruction cluster is generated and encrypted by SHA-256, and all instruction records are recorded in the blockchain: For example: CMD-20250510-003 → imaging task → imaging area coordinates → timestamp information (Nine) Execution and task feedback closed loop The instruction cluster is loaded and executed by the control center, and the task feedback information is used to optimize the model and archive the blockchain.

[0061] The above only describes the preferred embodiments of the application and does not limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. A method for generating low-orbit aircraft instruction clusters based on semantic mapping, characterized in that: The following steps are involved: Step S1: Analyze satellite mission requirements and satellite real-time status through a semantic mapping module to generate a mission requirement vector T and a satellite state vector S, wherein: the mission requirement vector T includes the timeliness, spatial coverage, and energy consumption characteristics of the mission; the satellite state vector S includes the satellite's orbital altitude, payload status, and energy consumption characteristics; Step S2: Calculate the matching degree between the mission requirement vector T and the satellite state vector S, and generate a preliminary mission instruction set; Step S3: Using a spatiotemporal graph convolutional network (ST-GCN) to model the spatiotemporal dependencies of the preliminary task instruction set, and optimize the task execution sequence and resource allocation; Step S4: Adjust the task scheduling strategy in real time based on the reinforcement learning algorithm, and update the instruction execution order according to the satellite status changes and task priorities; Step S5: Simulate the task execution process through digital twin technology to generate an optimized task instruction cluster; Step S6: Use blockchain technology to hash and encrypt the task instruction cluster, and verify the integrity and security of the instruction through smart contracts.

2. The method for generating low-orbit aircraft instruction clusters based on semantic mapping according to claim 1, characterized in that: The task requirement vector T in step S1 is expressed as: Among them, t i represents the i-th feature of the task, including task timeliness, spatial coverage, and energy consumption; The state vector S of the satellite is: Among them, S i represents the i-th state characteristic of the satellite, including orbital altitude, payload status, and energy consumption; By calculating the similarity between the mission requirements and the satellite status, the matching degree is obtained by the following formula: The matching degree between mission requirements and satellite status is calculated by cosine similarity to generate appropriate mission instructions.

3. The method for generating low-orbit aircraft instruction clusters based on semantic mapping according to claim 1, characterized in that: The semantic mapping module in step S1 uses a natural language processing (NLP) model to convert unstructured task requirement text into structured feature vectors, thereby more efficiently abstracting and optimizing the task requirements.

4. The method for generating low-orbit aircraft instruction clusters based on semantic mapping according to claim 1, wherein: The spatiotemporal graph convolutional network (ST-GCN) in step S3 extracts spatiotemporal dependencies through the following graph convolution formula: in, represents the representation of the j-th task instruction at the k-th layer, is the dependency between instructions, is the convolution kernel weight, is the set of adjacent nodes of instruction i, and σ is the activation function.

5. The method for generating low-orbit aircraft instruction clusters based on semantic mapping according to claim 1, wherein: The reinforcement learning algorithm in step S4 adopts the Q-learning algorithm, and its Q value update formula is: Here, Q(s,a) is the expected reward for performing action a in state s, R(s,a) is the immediate reward, γ is the discount factor, s′ is the new state after execution, and a′ is the subsequent action.

6. The method for generating low-orbit aircraft instruction clusters based on semantic mapping according to claim 1, characterized in that: The digital twin technology in step S5 optimizes the task execution cost through the following objective function: in, is the cost of task execution, is the decision variable for whether the task instruction is executed, A is the constraint matrix, and b is the constraint vector.

7. The method for generating low-orbit aircraft instruction clusters based on semantic mapping according to claim 1, wherein: The blockchain technology in step S6 uses the SHA-256 algorithm to generate a task instruction hash value, and verifies the consistency of the hash value and the execution result through a smart contract.

8. A satellite mission instruction generation and scheduling system based on the method according to any one of claims 1 to 7, characterized in that: include: Semantic mapping module, spatiotemporal graph convolutional network module, reinforcement learning optimization module, digital twin simulation module and blockchain security module.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images