Electromagnetic digital parallel space construction method and system based on ternary space architecture

Abstract

The application belongs to the field of electronic information technology, and particularly relates to an electromagnetic digital parallel space construction method and system based on a ternary space architecture, which comprises the following steps: collecting source data of a physical electromagnetic space, a digital space and a parallel space through a multi-dimensional sensor network, and performing standardized preprocessing; based on the processed data, constructing and synchronizing a high-fidelity mapping model between the physical space and the digital space; migrating the mapping model to the parallel space to form an intelligent agent model with autonomous decision-making capability; driving the agent model to perform multi-branch interactive evolution in the parallel space to generate optimized decision instructions; issuing the instructions to the physical space for execution, and performing closed-loop iterative optimization on the mapping model and the agent model based on execution effect data. The application realizes global dynamic perception, intelligent pre-performance and adaptive optimization of a complex electromagnetic system, and significantly improves the cognitive, decision-making and control capabilities of the system in a dynamic uncertain environment.

Description

Technical Field

[0001] This invention belongs to the field of electronic information technology, specifically relating to a method and system for constructing electromagnetic digital parallel spaces based on a ternary spatial architecture. Background Technology

[0002] With the rapid development of wireless communication, the Internet of Things, and space information technology, accurate understanding, efficient management, and dynamic optimization of complex electromagnetic environments have become key technical requirements in fields such as industrial testing, network planning, and system operation and maintenance. Existing technical solutions mostly focus on the condition monitoring and offline simulation of physical electromagnetic equipment, with a typical process of "data acquisition - model analysis - result output." While these solutions achieve a certain degree of basic digitization of the electromagnetic environment, their core remains a static mapping of a single physical domain. They fail to establish a dynamic cognitive and decision-making architecture that coordinates the "physical domain - digital domain - intelligent pre-simulation domain," resulting in significant deficiencies in the system's overall support for the correlation, dynamic interaction, and optimal resource allocation of multiple elements within complex electromagnetic systems.

[0003] Specifically, existing technologies face the following challenges: First, the interaction between the physical and digital domains is mostly one-way data transmission, lacking a collaborative evolution and bidirectional calibration mechanism based on real-time feedback, leading to a decline in model fidelity over time. Second, system optimization and decision-making are largely based on fixed rules or historical data, failing to integrate real-time states, multi-branch simulations, and adaptive strategy optimization, making it difficult to cope with the rapid time-varying nature and sudden uncertainties of the electromagnetic environment. Finally, the system architecture is rigid, lacking the ability to dynamically reorganize and self-optimize based on environmental changes, equipment status, and task objectives. Therefore, there is an urgent need for a method and application system for constructing electromagnetic digital parallel spaces that can achieve full-domain dynamic perception, high-fidelity modeling, intelligent pre-simulation, and closed-loop optimization. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for constructing electromagnetic digital parallel spaces based on a ternary spatial architecture, so as to solve the problems mentioned in the background art.

[0005] The present invention achieves the above objectives through the following technical solutions: Firstly, this invention proposes a method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture, the method comprising: Source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space are collected, and the source data is preprocessed to generate a standardized ternary space dataset. Based on the ternary space dataset, a mapping model between the electromagnetic physical space and the electromagnetic digital space is constructed and synchronized, and the mapping model is migrated to the electromagnetic parallel space to form a proxy model. Based on the ternary space dataset and the agent model, interactive evolution is performed to generate decision instructions; The decision instruction is sent to the electromagnetic physical space for execution, and the execution effect data is collected. Based on the execution effect data, the mapping model and the proxy model are iteratively optimized.

[0006] As a preferred embodiment, the acquisition of source data from the electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space, and the preprocessing of the source data, include: Electromagnetic physical space data is collected by deploying a multi-domain sensor network in different spatial dimensions at a set sampling frequency. The electromagnetic physical space data includes device status data, signal characteristic data, and environmental parameter data. The different spatial dimensions include one or more of ground, air, space-based, and sea-based. Collect digital model parameter data and simulation process data of the electromagnetic digital space; Collect evolution and deduction data, decision command data, and effect feedback data of the electromagnetic parallel space; The electromagnetic physical space data, the digital model parameter data and simulation process data, as well as the evolution inference data, decision command data and effect feedback data are cleaned, normalized and feature extracted respectively to generate the standardized ternary space dataset.

[0007] As a preferred embodiment, the construction and synchronization of the mapping model between the electromagnetic physical space and the electromagnetic digital space includes: based on the preprocessed electromagnetic physical space data, using a method that combines physical modeling and data-driven fusion, constructing a device digital twin model representing the physical equipment; Based on the preprocessed environmental parameter data, an electromagnetic environment model is constructed using multiphysics coupling modeling technology to simulate the propagation characteristics of electromagnetic signals in real geographic space and atmospheric environment, and is dynamically updated based on the real-time collected environmental parameter data. The digital twin model of the equipment is integrated with the electromagnetic environment model to construct an electromagnetic digital space application scenario model for simulation and deduction. The real-time status data of the electromagnetic physical space is synchronized to the electromagnetic digital space through a high-speed communication network, thereby driving the application scenario model of the electromagnetic digital space to be dynamically updated.

[0008] As a preferred embodiment, the step of migrating the mapping model to the electromagnetic parallel space to form a surrogate model includes: The device digital twin model and electromagnetic environment model in the electromagnetic digital space application scenario model are migrated to the electromagnetic parallel space; To integrate reinforcement learning and game theory algorithms into the digital twin models of devices migrated to electromagnetic parallel space, the agent models are constructed. Each agent model corresponds to one of the digital twin models of the devices. Its decision-making strategy in the inference process is determined based on the real-time simulation state of the electromagnetic digital space application scenario model. The real-time simulation state includes at least the operating parameters of its corresponding digital twin model of the devices and the parameters of the electromagnetic environment model.

[0009] As a preferred embodiment, the derivation of the surrogate model in the electromagnetic parallel space includes the following steps: For any agent model i, the state information it receives at time t during the deduction process. The state information of the corresponding device's digital twin model itself. State information of the digital twin model of the device corresponding to other observable proxy models. and the state information of the electromagnetic environment model. It is calculated through the information fusion function F and is expressed as: ; The agent model i is based on its policy function Based on the state information Determine the action it performs at time t. ,action Obey the strategy In state The probability distribution under the following conditions is expressed as: ; The set of actions of all agent models at time t The electromagnetic parallel space acts together, and through the parallel deduction function P, determines the state information at the next deduction time t+1. , is represented as: The reward obtained by the agent model i at time t By reward function Determined based on the global state and all actions at time t, and expressed as: The reward Used to update its policy function .

[0010] As a preferred embodiment, the step of interactively evolving and generating decision instructions based on the ternary space dataset and the proxy model includes: By fusing the preprocessed ternary spatial dataset, a comprehensive state vector is generated to describe the overall situation. ,in In terms of physical space state, For digital space state, It is a parallel space state; Based on the comprehensive state vector, the electromagnetic digital space application scenario model is driven to perform simulation to obtain at least one set of simulation results; In the electromagnetic parallel space, at least two types of evolution scenarios, namely, normal scenarios and extreme sudden events and multi-domain collaboration, are initialized, and the proxy model is driven to perform multi-threaded parallel inference in the evolution scenarios to generate multi-branch evolution results corresponding to different initial conditions and strategies. Based on the preset optimization objectives, the multi-branch evolution results are evaluated and filtered to generate decision instructions for controlling physical devices.

[0011] As a preferred embodiment, the evaluation and screening of the multi-branch evolution results based on a preset optimization objective includes: Construct a reward function R, which is expressed as R = α(1 - Pd) + βCr + γUr, where Pd is the probability of the other party's detection, Cr is the communication success rate of our side, Ur is the resource utilization rate, and α, β, and γ are preset weight coefficients. Based on the reward function R, calculate the cumulative reward for each branch evolution result throughout the entire deduction process; The agent model strategy sequence corresponding to the branch evolution result with the highest cumulative reward is selected as the basis for generating the decision instruction.

[0012] As a preferred embodiment, after generating the decision instructions for controlling the physical device and before issuing them to the electromagnetic physical space for execution, the method further includes: The decision-making instructions are converted into control parameters that can be recognized by the electromagnetic digital space application scenario model; In the electromagnetic digital space, the application scenario model is rapidly simulated and verified based on the control parameters; If the simulation verification results meet the preset performance threshold, the decision instruction is converted into a standardized control instruction; otherwise, a new interactive evolution process is triggered.

[0013] As a preferred embodiment, the iterative optimization of the mapping model and the proxy model based on the execution effect data includes: The execution effect data is compared with the inference and prediction effect data corresponding to the decision instruction to generate model deviation data; The parameters of the digital twin model of the equipment and the parameters of the electromagnetic environment model are corrected using the model deviation data. The agent model's policy function is retrained using the model bias data and the reward function to optimize its subsequent decision-making strategy.

[0014] Secondly, this invention proposes an electromagnetic digital parallel space system based on a ternary spatial architecture, used to implement the electromagnetic digital parallel space construction method described above. The system includes: The multi-domain data acquisition module is configured to acquire source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space through a sensor network deployed in different spatial dimensions. The data processing and synchronization module is configured to clean, normalize, and extract features from the source data to generate a standardized ternary spatial dataset and to achieve real-time data synchronization between the electromagnetic physical space and the electromagnetic digital space. The digital modeling and simulation module is configured to construct and dynamically update the device digital twin model, the electromagnetic environment model, and the electromagnetic digital space application scenario model that integrates the two, based on the standardized ternary spatial dataset. The parallel evolution and decision-making module is configured to migrate the electromagnetic digital space application scenario model to the electromagnetic parallel space to form multiple agent models with autonomous decision-making capabilities, drive the agent models to perform multi-branch interactive evolution, and generate decision instructions based on the evolution results. The closed-loop control and optimization module is configured to issue and execute the decision command, collect execution effect data, and iteratively optimize the model in the digital modeling and simulation module and the surrogate model in the parallel evolution and decision module based on the execution effect data.

[0015] The beneficial effects of this invention are as follows: This invention constructs a three-dimensional spatial architecture that coordinates physical electromagnetic space, digital space, and parallel space. First, it collects multi-dimensional data such as device status, signal characteristics, and environmental parameters through a multi-domain sensor network, and establishes a high-fidelity digital twin model to achieve real-time synchronous mapping between the physical and digital spaces. Based on this, the digital twin model is transferred to the parallel space and endowed with autonomous decision-making capabilities based on reinforcement learning and game theory, forming an intelligent agent model capable of multi-threaded parallel inference. These agent models can simulate interactive evolution under various scenarios such as normal, sudden, and collaborative scenarios in the parallel space, continuously optimizing decision-making strategies through a closed-loop mechanism of state perception, strategy decision-making, environmental updates, and reward feedback. Finally, the system issues and executes the optimized decision instructions and uses actual effect data to back-calibrate the digital model and agent strategy; enabling the system to possess feedback-based adaptive optimization and dynamic reconstruction capabilities, effectively responding to environmental changes and task variations. Attached Figure Description

[0016] Figure 1 This is a flowchart of an electromagnetic digital parallel space construction method based on a ternary spatial architecture in this invention; Figure 2This is a system block diagram of an electromagnetic digital parallel space construction system based on a ternary spatial architecture in this invention. Detailed Implementation

[0017] The following description provides specific application scenarios and requirements for this specification, intended to enable those skilled in the art to make and use the contents of this specification. Various partial modifications to the disclosed embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments and applications without departing from the spirit and scope of this specification. Therefore, this specification is not limited to the embodiments shown, but rather to the widest scope consistent with the claims.

[0018] The terminology used herein is for the purpose of describing particular exemplary embodiments only and is not restrictive. For example, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. When used in this specification, the terms “comprising,” “including,” and / or “containing” mean that the associated integers, steps, operations, elements, and / or components are present, but do not exclude the presence of one or more other features, integers, steps, operations, elements, components, and / or groups, or that other features, integers, steps, operations, elements, components, and / or groups may be added to the system / method.

[0019] Considering the following description, these and other features of this specification, as well as the operation and function of the related components of the structure, and the economy of assembly and manufacture of the parts, can be significantly improved. All of these form part of this specification with reference to the accompanying drawings. However, it should be clearly understood that the drawings are for illustrative and descriptive purposes only and are not intended to limit the scope of this specification. It should also be understood that the drawings are not drawn to scale.

[0020] The flowcharts used in this specification illustrate operations implemented according to some embodiments of this specification. It should be clearly understood that the operations in the flowcharts may not be implemented in a sequential order. Instead, the operations may be implemented in reverse order or simultaneously. Furthermore, one or more additional operations may be added to the flowcharts. One or more operations may be removed from the flowcharts.

[0021] First Embodiment like Figure 1 As shown, a preferred embodiment of the present invention proposes a method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture, the method comprising: S1: Collect source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space, and preprocess the source data to generate a standardized ternary space dataset.

[0022] In practice, the data acquisition process employs a multi-dimensional, multi-source fusion approach. This involves the collaborative work of heterogeneous sensor networks deployed across different spatial dimensions, including ground, air, space, and sea, to acquire comprehensive electromagnetic environment and equipment status information. Electromagnetic physical space data primarily includes operating status parameters of various electronic devices (such as transmit power, operating frequency band, and modulation method), electromagnetic signal characteristic data (such as signal strength, spectrum occupancy, and time-frequency characteristics), and spatial environment parameters (such as topography, atmospheric refractive index, and ionospheric state). Electromagnetic digital space data mainly includes the structural parameters and state variables of the constructed digital twin model, as well as intermediate results and output data generated during simulation. Electromagnetic parallel space data includes the situational evolution paths generated during multi-branch inference, agent decision sequences, and comparative feedback information between inference results and actual execution effects.

[0023] In a preferred embodiment, source data from the electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space are acquired, and the source data is preprocessed, including: By deploying a multi-domain sensor network across different spatial dimensions, electromagnetic physical space data is collected at a set sampling frequency. This data includes equipment status data, signal characteristic data, and environmental parameter data. The different spatial dimensions include one or more of ground, air, space-based, and sea-based dimensions. Digital model parameter data and simulation process data of the electromagnetic digital space are also collected. Evolutionary deduction data, decision command data, and effect feedback data of the electromagnetic parallel space are also collected. The electromagnetic physical space data, digital model parameter data and simulation process data, as well as the evolutionary deduction data, decision command data, and effect feedback data, are cleaned, normalized, and feature extracted to generate a standardized ternary spatial dataset.

[0024] Specifically, the preprocessing steps include: cleaning the original data using outlier detection and removal techniques based on statistical methods; converting multi-source heterogeneous data to a unified dimension using maximum-minimum normalization or Z-score standardization methods; and then extracting key features from the original data that can characterize electromagnetic state, equipment health status, and environmental change trends through time-frequency analysis, feature engineering, or deep learning automatic feature extraction methods, forming a standardized feature vector set that can be used for subsequent modeling and inference.

[0025] S2: Based on the ternary space dataset, construct and synchronize the mapping model between the electromagnetic physical space and the electromagnetic digital space, and transfer the mapping model to the electromagnetic parallel space to form a proxy model.

[0026] Understandably, the construction of the mapping model is the core of achieving high-fidelity correlation between virtual and real spaces. Its establishment process follows the principle of mechanism and data fusion to ensure that the digital model can not only reflect the inherent working mechanism of physical entities, but also be continuously corrected through actual observation data to approximate real behavior.

[0027] In a preferred embodiment, a mapping model between the electromagnetic physical space and the electromagnetic digital space is constructed and synchronized, including: based on preprocessed electromagnetic physical space data, a physical modeling and data-driven approach is used to construct a digital twin model representing the physical device; wherein, the physical modeling establishes a parameterized model framework based on prior knowledge such as electromagnetic propagation theory and device working principle; the data-driven approach utilizes machine learning methods such as deep neural networks, using the aforementioned preprocessed real-time state data as training samples, to identify and compensate for unknown parameters or error functions in the mechanism model, thereby improving the model's representation accuracy and generalization ability in complex and variable environments.

[0028] Based on preprocessed environmental parameter data, a multiphysics coupled modeling technique is employed to construct an electromagnetic environment model that simulates the propagation characteristics of electromagnetic signals in real geographic space and atmospheric environments. This model is dynamically updated based on real-time acquired environmental parameter data. The electromagnetic environment model comprehensively considers the influence of various factors on electromagnetic wave propagation, such as terrain shielding, atmospheric refraction, multipath effects, and ionospheric disturbances. By coupling models from multiple disciplines including electromagnetics, atmospheric physics, and geographic information science, it achieves high-precision simulation of signal propagation attenuation, time delay, and disturbance characteristics under large-scale and complex environments. The dynamic update mechanism sets threshold values ​​for key environmental parameters (such as temperature gradient, humidity changes, and ionospheric electron concentration fluctuations). When parameter changes exceed these thresholds, online correction of the model parameters is automatically triggered, ensuring consistency between the simulated environment and the physical environment.

[0029] By integrating the digital twin model of the equipment with the electromagnetic environment model, an electromagnetic digital space application scenario model for simulation and deduction is constructed. The integration process, based on the actual task scenario or system operation logic, deploys multiple equipment twins in the electromagnetic environment model under a unified spatiotemporal reference, and defines the interaction protocols between the equipment, the connection relationships of signal links, and the system-level constraints and performance indicators, thereby forming an operable, observable, and controllable closed-loop simulation system.

[0030] Real-time state data of the electromagnetic physical space is synchronized to the electromagnetic digital space via a high-speed communication network to drive dynamic updates of the electromagnetic digital space application scenario model. This synchronization process relies on low-latency, highly reliable communication links (such as 5G / 6G, fiber optics, and satellite communication) and employs a precise time synchronization protocol to ensure data consistency. When the state of physical space devices changes, the corresponding digital twin model parameters are adjusted, triggering a chain reaction of updates to the associated environmental model and scenario logic, thereby achieving real-time mirroring and advanced simulation of the physical space from the digital space.

[0031] In a preferred embodiment, the mapping model is transferred to an electromagnetic parallel space to form a surrogate model, including: The device digital twin model and electromagnetic environment model in the electromagnetic digital space application scenario model are migrated to the electromagnetic parallel space. The migration process is essentially to use the high-fidelity model in the digital space and its current state as the initial conditions and evolution basis for the parallel space deduction, ensuring that the parallel deduction starts from a starting point that is consistent with or highly similar to the physical world.

[0032] To integrate reinforcement learning and game theory algorithms into the digital twin models of equipment migrated to electromagnetic parallel space, surrogate models are constructed. Each surrogate model corresponds to a digital twin model of the equipment. Its decision-making strategy in the inference process is determined based on the real-time simulation state of the electromagnetic digital space application scenario model. The real-time simulation state includes at least the operating parameters of its corresponding digital twin model of the equipment and the parameters of the electromagnetic environment model.

[0033] Understandably, this disclosure, by introducing reinforcement learning, enables the agent model to autonomously learn optimal behavioral strategies through trial and error and reward mechanisms; by incorporating game theory, it enables the agent model to reason about the possible strategies of other agents and make optimal responses in an environment where multiple agents coexist. This makes the agent model not merely a digital copy of a physical device, but an intelligent entity with autonomous perception, decision-making, and learning capabilities.

[0034] In a preferred embodiment, the derivation of the surrogate model in electromagnetic parallel space includes the following steps: For any agent model i, the state information it receives at time t during the deduction process. The state information of the corresponding device's digital twin model itself. State information of the digital twin model of the device corresponding to other observable proxy models. and the state information of the electromagnetic environment model. It is calculated through the information fusion function F and is expressed as: .

[0035] The information fusion function F can employ methods such as neural networks, Kalman filtering, or attention mechanisms to effectively integrate and reduce the dimensionality of multi-source, heterogeneous state information, thereby forming a unified situational awareness of the surrogate model.

[0036] Agent model i according to its policy function Based on state information Determine the action it performs at time t. ,action Obey the strategy In state The probability distribution under the following conditions is expressed as: Policy function Typically represented by deep neural networks, the output of which is the probability distribution of each possible action in a given state. The agent samples according to this distribution to determine the specific action to be performed. This randomness helps to achieve a balance between exploring unknown strategies and utilizing known experience.

[0037] The set of actions of all agent models at time t Together they act in electromagnetic parallel space, and through the parallel deduction function P, determine the state information at the next deduction time t+1. , is represented as: The parallel inference function P encapsulates the world evolution model of the parallel space. Based on the joint actions of all agents and according to predefined physical laws, interaction rules, or learned environmental dynamics models, it calculates the state transition of the entire parallel system at the next moment.

[0038] The reward obtained by agent model i at time t By reward function Determined based on the global state and all actions at time t, and expressed as: ,award Used to update its policy function The reward function This is crucial for guiding the learning direction of the intelligent agent. Its design must be closely aligned with the task objective, and can comprehensively consider multiple dimensions such as communication quality, resource utilization, task completion, and adversarial effectiveness. The intelligent agent continuously optimizes its policy function by constantly trying and receiving rewards from environmental feedback, utilizing reinforcement learning algorithms such as policy gradients. Ultimately, it tends to converge to the optimal strategy that maximizes long-term cumulative rewards.

[0039] S3: Based on the ternary space dataset and agent model, it performs interactive evolution and generates decision instructions.

[0040] Understandably, this step is a key process for realizing the transition from situational awareness to intelligent decision-making. Its core lies in using the constructed agent model and high-fidelity digital environment to explore a large number of possible future development paths in parallel space, and to select the optimal action plan through a scientific evaluation mechanism, thereby providing precise and forward-looking instructions for the control of equipment in physical space.

[0041] In a preferred embodiment, based on a ternary space dataset and an agent model, interactive evolution is performed to generate decision instructions, including: By fusing the preprocessed ternary spatial dataset, a comprehensive state vector is generated to describe the overall situation. ,in In terms of physical space state, For digital space state, This represents a parallel space state. The fusion process employs methods such as data association, feature concatenation, or neural network-based encoders to integrate multi-source heterogeneous information from physical space sensors (reflecting real-time equipment operating conditions and environmental conditions), digital space simulators (reflecting model confidence and simulation progress), and parallel space predictors (reflecting inference paths and strategy evaluations) into a unified, high-dimensional mathematical representation. This integrated state vector S constitutes the total input and context for all subsequent simulations and inferences.

[0042] Based on the comprehensive state vector, a simulation of an electromagnetic digital space application scenario model is driven to obtain at least one set of simulation results. The simulation is conducted in a digital twin environment, based on the current system state (comprehensive state vector S) and a set of operational parameters to be verified (which may be derived from historical experience or preliminary strategies from parallel spaces). The high-fidelity model is run for a finite period of time or cycle to predict the immediate changing trends of key system performance indicators (such as signal coverage, communication error rate, and equipment load rate) after implementing the set of operations. These predicted results constitute the simulation results, and their purpose is to eliminate obviously infeasible or high-risk solutions from an engineering feasibility perspective before actual implementation.

[0043] In the electromagnetic parallel space, at least two types of evolution scenarios are initialized, including conventional confrontation, extreme emergencies, and multi-domain collaboration. The agent model is driven to perform multi-threaded parallel inference in the evolution scenarios to generate multi-branch evolution results corresponding to different initial conditions and strategies.

[0044] In practice, firstly, based on the task scenario or risk assessment, several typical evolutionary scenarios can be defined: Typical adversarial scenarios: Simulate the standard interaction process between the system and the external environment or other systems within the range of known rules and typical parameters.

[0045] Extreme emergency scenarios: Simulate low-probability, high-impact emergencies, such as sudden strong signal occurrences, sudden failures of critical equipment, or severe environmental deterioration, to test the system's robustness and emergency response capabilities.

[0046] Multi-domain collaboration scenarios: Simulate complex collaborative tasks involving multiple different types of devices, multiple subsystems, or across physical domains (such as air, space, and ground) to test the effectiveness of the overall collaboration strategy.

[0047] Secondly, leveraging the powerful computing capabilities of the cloud or high-performance computing clusters, independent computing threads are created for each scenario or different initial strategies within the same scenario, simultaneously driving hundreds or even thousands of intelligent agent models to "live" and "play" in parallel within these virtual scenarios. Each thread starts from the current comprehensive state vector S, but may be assigned different initial strategies, explore random seeds, or encounter different virtual random events. After a certain period of deduction, each thread generates a unique future state evolution trajectory, recording the action sequences, state transitions, and final outcomes of all agents. The collection of trajectories generated by all threads constitutes a multi-branch evolutionary result library covering a wide range of possibilities.

[0048] Based on a preset optimization objective, the multi-branch evolution results are evaluated and filtered to generate decision instructions for controlling physical devices. This is the step of converging a massive number of possibilities into a single optimal decision. The evaluation process scores each evolution branch according to a predefined, quantified optimization objective (usually expressed as a multi-objective optimization function).

[0049] In a preferred approach, the multi-branch evolution results are evaluated and screened based on a preset optimization objective, including: A reward function R is constructed, which is expressed as R = α(1 - Pd) + βCr + γUr, where Pd is the enemy detection probability, Cr is the friendly communication success rate, Ur is the resource utilization rate, and α, β, and γ are preset weight coefficients. This reward function R is a specific mathematical manifestation of the optimization objective. It combines the three objectives of reducing the detection probability, maintaining reliable communication, and efficiently utilizing resources, which are usually in trade-off relationships, into a single optimization index through linear weighting, thereby guiding the agent to learn a balanced strategy.

[0050] The surrogate model policy sequence corresponding to the branch evolution result with the highest cumulative reward is selected as the basis for generating decision instructions. By comparing the total rewards of all parallel inference branches, the best-performing future trajectory is selected. From this best trajectory, the surrogate model action sequence in the short term (beginning in the next decision cycle) is extracted. This action sequence represents a set of synthetic operations that have been validated in the parallel space and are expected to lead to the optimal future.

[0051] In a preferred embodiment, after generating decision instructions for controlling physical devices and before issuing them to the electromagnetic physical space for execution, the method further includes: The decision-making instructions are converted into control parameters that can be recognized by the electromagnetic digital space application scenario model; that is, the abstract action sequence (such as "adjust the transmission power of node A" or "switch the operating frequency band of node B") is concretized into precise parameter values ​​that can be accepted by the simulation interface of the digital twin model of that node (such as "set the power to 20dBm" or "switch the center frequency to 2.4GHz").

[0052] In the electromagnetic digital space, rapid simulation verification of the application scenario model is performed based on control parameters; this step can be regarded as an engineering-level reproduction and verification of the parallel space derivation results. Using the transformed specific control parameters, one or more short-term, deterministic simulations are performed in a digital twin environment with higher fidelity or more detailed modeling. The purpose is not to explore possibilities, but to rigorously verify whether the key performance indicators of the system under this set of specific parameters do indeed improve as predicted in the parallel space, and whether all hard safety and engineering constraints (such as power limits, intermodulation interference avoidance, etc.) are met.

[0053] If the simulation verification results meet the preset performance threshold, the decision command is converted into a standardized control command; otherwise, a new interactive evolution process is triggered. The performance threshold is the final checkpoint to ensure the safety and effectiveness of the decision. If the digital space verification confirms performance improvement and no violations, the control parameters are encapsulated into a standard command format (such as a specific control message) that conforms to the physical device communication protocol. If the verification fails (e.g., performance does not meet expectations, or a constraint alarm is triggered), it indicates that the parallel space deduction results may have model errors or unconsidered constraints. In this case, the failure case is used as feedback to trigger a new round of interactive evolution from step S3, using the new feedback information to optimize the next round of deduction.

[0054] S4: Issue decision instructions to the electromagnetic physical space for execution, collect execution effect data, and iteratively optimize the mapping model and proxy model based on the execution effect data.

[0055] In a preferred embodiment, the mapping model and the proxy model are iteratively optimized based on execution performance data, including: The execution performance data is compared with the predicted performance data corresponding to the decision instruction to generate model bias data. After the instruction is executed in physical space, actual performance data (such as actual communication quality measurements, actual spectrum occupancy, and actual equipment energy consumption) is collected through a sensor network. This batch of real data is then compared item by item with the predicted performance data of the optimal evolutionary branch on which the decision instruction was based in step S3 during the same time period. The differences between the two (such as the difference between the predicted communication rate and the actual rate) constitute the model bias data, which quantifies the error of the current digital model and the parallel inference model in predicting the real world.

[0056] Model bias data is used to correct the parameters of the equipment digital twin model and the electromagnetic environment model. The model bias data is used as a monitoring signal to drive the parameter updates of the digital space model. For example, system identification, parameter estimation, or gradient-based optimization methods are used to adjust the internal parameters in the digital twin model that cause prediction bias (such as equipment efficiency coefficient and antenna gain model parameters), or to adjust parameters such as propagation loss factor and noise floor in the electromagnetic environment model, so that the model output is closer to the newly observed real data.

[0057] By utilizing model bias data and the reward function, the policy function of the surrogate model is retrained to optimize its subsequent decision-making strategies. Simultaneously, model bias reveals the degree of distortion in reward calculation during parallel space simulations. Actual performance data can be substituted into the reward function R to calculate the "true reward," which is then compared with the predicted reward in the simulation. This comparison, along with the model bias data itself, can serve as new training samples input into the reinforcement learning training loop of the surrogate model. Through algorithms such as policy gradient analysis, the neural network weights of the surrogate model's policy function are updated, enabling it to better predict the true consequences of actions in future simulations, thereby making better decisions that align with the laws of the physical world.

[0058] Through the aforementioned corrections and retraining, the entire system completes a full closed loop of perception, decision-making, action, and learning. Each closed loop makes the digital model more accurate, parallel inferences more reliable, and intelligent decisions more effective, thereby endowing the system with the ability to continuously evolve in the face of dynamic and complex environments.

[0059] Second Embodiment like Figure 2 As shown, another preferred embodiment of the present invention proposes an electromagnetic digital parallel space system based on a ternary spatial architecture, used to implement the steps of the construction method as described in the first embodiment. The system includes: The multi-domain data acquisition module is configured to acquire source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space through a sensor network deployed in different spatial dimensions. The data processing and synchronization module is configured to clean, normalize, and extract features from the source data to generate a standardized ternary spatial dataset, and to achieve real-time data synchronization between the electromagnetic physical space and the electromagnetic digital space. The digital modeling and simulation module is configured to build and dynamically update the equipment digital twin model, electromagnetic environment model, and electromagnetic digital space application scenario model that integrates the two, based on a standardized ternary spatial dataset. The parallel evolution and decision-making module is configured to migrate the electromagnetic digital space application scenario model to the electromagnetic parallel space to form multiple agent models with autonomous decision-making capabilities, drive the agent models to perform multi-branch interactive evolution, and generate decision instructions based on the evolution results. The closed-loop control and optimization module is configured to issue and execute decision commands, collect execution effect data, and iteratively optimize the model in the digital modeling and simulation module and the surrogate model in the parallel evolution and decision module based on the execution effect data.

[0060] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0061] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0062] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture, characterized in that, The method includes: Source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space are collected, and the source data is preprocessed to generate a standardized ternary space dataset. Based on the ternary space dataset, a mapping model between the electromagnetic physical space and the electromagnetic digital space is constructed and synchronized, and the mapping model is migrated to the electromagnetic parallel space to form a proxy model. Based on the ternary space dataset and the agent model, interactive evolution is performed to generate decision instructions; The decision instruction is sent to the electromagnetic physical space for execution, and the execution effect data is collected. Based on the execution effect data, the mapping model and the proxy model are iteratively optimized.

2. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 1, characterized in that, The acquisition of source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space, and the preprocessing of the source data, include: Electromagnetic physical space data is collected by deploying a multi-domain sensor network in different spatial dimensions at a set sampling frequency. The electromagnetic physical space data includes device status data, signal characteristic data, and environmental parameter data. The different spatial dimensions include one or more of ground, air, space-based, and sea-based. Collect digital model parameter data and simulation process data of the electromagnetic digital space; Collect evolution and deduction data, decision command data, and effect feedback data of the electromagnetic parallel space; The electromagnetic physical space data, the digital model parameter data and simulation process data, as well as the evolution inference data, decision command data and effect feedback data are cleaned, normalized and feature extracted respectively to generate the standardized ternary space dataset.

3. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 1, characterized in that, The construction and synchronization of the mapping model between the electromagnetic physical space and the electromagnetic digital space includes: based on the preprocessed electromagnetic physical space data, using a method that combines physical modeling and data-driven fusion, constructing a device digital twin model representing the physical device; Based on the preprocessed environmental parameter data, an electromagnetic environment model is constructed using multiphysics coupling modeling technology to simulate the propagation characteristics of electromagnetic signals in real geographic space and atmospheric environment, and is dynamically updated based on the real-time collected environmental parameter data. The digital twin model of the equipment is integrated with the electromagnetic environment model to construct an electromagnetic digital space application scenario model for simulation and deduction. The real-time status data of the electromagnetic physical space is synchronized to the electromagnetic digital space through a high-speed communication network, thereby driving the application scenario model of the electromagnetic digital space to be dynamically updated.

4. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 3, characterized in that, The step of transferring the mapping model to the electromagnetic parallel space to form a surrogate model includes: The device digital twin model and electromagnetic environment model in the electromagnetic digital space application scenario model are migrated to the electromagnetic parallel space; To integrate reinforcement learning and game theory algorithms into the digital twin models of devices migrated to electromagnetic parallel space, the agent models are constructed. Each agent model corresponds to one of the digital twin models of the devices. Its decision-making strategy in the inference process is determined based on the real-time simulation state of the electromagnetic digital space application scenario model. The real-time simulation state includes at least the operating parameters of its corresponding digital twin model of the devices and the parameters of the electromagnetic environment model.

5. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 4, characterized in that, The derivation of the surrogate model in the electromagnetic parallel space includes the following steps: For any agent model i, the state information it receives at time t during the deduction process. The state information of the corresponding device's digital twin model itself. State information of the digital twin model of the device corresponding to other observable proxy models. and the state information of the electromagnetic environment model. It is calculated through the information fusion function F and is expressed as: ; The agent model i is based on its policy function Based on the state information Determine the action it performs at time t. ,action Obey the strategy In state The probability distribution under the following conditions is expressed as: ; The set of actions of all agent models at time t The electromagnetic parallel space acts together, and through the parallel deduction function P, determines the state information at the next deduction time t+1. , is represented as: The reward obtained by the agent model i at time t By reward function Determined based on the global state and all actions at time t, and expressed as: The reward Used to update its policy function .

6. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 4, characterized in that, The process of interactive evolution and generation of decision instructions based on the ternary space dataset and the agent model includes: By fusing the preprocessed ternary spatial dataset, a comprehensive state vector is generated to describe the overall situation. ,in In terms of physical space state, For digital space state, It is a parallel space state; Based on the comprehensive state vector, the electromagnetic digital space application scenario model is driven to perform simulation to obtain at least one set of simulation results; In the electromagnetic parallel space, at least two types of evolution scenarios are initialized, including conventional confrontation, extreme emergencies and multi-domain cooperation, and the agent model is driven to perform multi-threaded parallel deduction in the evolution scenarios to generate multi-branch evolution results corresponding to different initial conditions and strategies. Based on the preset optimization objectives, the multi-branch evolution results are evaluated and filtered to generate decision instructions for controlling physical devices.

7. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 6, characterized in that, The evaluation and screening of the multi-branch evolution results based on the preset optimization objective includes: Construct a reward function R, which is expressed as R = α(1 - Pd) + βCr + γUr, where Pd is the probability of the other party's detection, Cr is the communication success rate of our side, Ur is the resource utilization rate, and α, β, and γ are preset weight coefficients. Based on the reward function R, calculate the cumulative reward for each branch evolution result throughout the entire deduction process; The agent model strategy sequence corresponding to the branch evolution result with the highest cumulative reward is selected as the basis for generating the decision instruction.

8. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 6, characterized in that, After generating the decision instructions for controlling the physical devices and before issuing them to the electromagnetic physical space for execution, the method further includes: The decision-making instructions are converted into control parameters that can be recognized by the electromagnetic digital space application scenario model; In the electromagnetic digital space, the application scenario model is rapidly simulated and verified based on the control parameters; If the simulation verification results meet the preset performance threshold, the decision instruction is converted into a standardized control instruction; otherwise, a new interactive evolution process is triggered.

9. The method for constructing an electromagnetic digital parallel space based on a ternary spatial architecture according to claim 1, characterized in that, The iterative optimization of the mapping model and the proxy model based on the execution effect data includes: The execution effect data is compared with the inference and prediction effect data corresponding to the decision instruction to generate model deviation data; The parameters of the digital twin model of the equipment and the parameters of the electromagnetic environment model are corrected using the model deviation data. The agent model's policy function is retrained using the model bias data and the reward function to optimize its subsequent decision-making strategy.

10. An electromagnetic digital parallel space system based on a ternary spatial architecture, used to implement the electromagnetic digital parallel space construction method as described in any one of claims 1-9, characterized in that, The system includes: The multi-domain data acquisition module is configured to acquire source data from electromagnetic physical space, electromagnetic digital space, and electromagnetic parallel space through a sensor network deployed in different spatial dimensions. The data processing and synchronization module is configured to clean, normalize, and extract features from the source data to generate a standardized ternary spatial dataset and to achieve real-time data synchronization between the electromagnetic physical space and the electromagnetic digital space. The digital modeling and simulation module is configured to construct and dynamically update the device digital twin model, the electromagnetic environment model, and the electromagnetic digital space application scenario model that integrates the two, based on the standardized ternary spatial dataset. The parallel evolution and decision-making module is configured to migrate the electromagnetic digital space application scenario model to the electromagnetic parallel space to form multiple agent models with autonomous decision-making capabilities, drive the agent models to perform multi-branch interactive evolution, and generate decision instructions based on the evolution results. The closed-loop control and optimization module is configured to issue and execute the decision command, collect execution effect data, and iteratively optimize the model in the digital modeling and simulation module and the surrogate model in the parallel evolution and decision module based on the execution effect data.

Images