Artificial intelligence-based dynamic electromagnetic environment characterization method and system

Abstract

The application provides a dynamic electromagnetic environment characterization method and system based on artificial intelligence, wherein the method comprises the following steps: acquiring multi-modal electromagnetic signals in a dynamic interference environment, generating a two-dimensional time-frequency graph based on the multi-modal electromagnetic signals, determining cooperative signals and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency graph, performing communication protocol characterization based on the cooperative signals to obtain signal coordination instructions, performing interference frequency point prediction based on the non-cooperative signals to obtain an adaptive frequency hopping sequence, controlling communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination instructions and the adaptive frequency hopping sequence, and using the two-dimensional time-frequency graph to clearly see the interference and distinguish cooperative signals from non-cooperative signals, and then adopting differentiated management strategies of protocol guidance clearing and active prediction avoidance for different types of signals, and finally successfully constructing a high-determinacy and anti-interference communication link in an extremely severe dynamic electromagnetic environment through the control of the synchronous frequency hopping of the communication nodes.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a dynamic electromagnetic environment characterization method and system based on artificial intelligence. Background Technology

[0002] In critical scenarios such as smart manufacturing workshops and power monitoring centers, the electromagnetic environment is exceptionally complex. This environment contains both non-cooperative signals such as broadband electromagnetic leakage and radiation from industrial motors and power lines, and cooperative signals from communication networks like Wi-Fi and 5G. These cooperative and non-cooperative signals intertwine and dynamically change in the time, frequency, and spatial domains, forming a high-density, heterogeneous interference pattern that seriously threatens the reliability and determinism of wireless communication.

[0003] Currently, the handling of complex interference mainly relies on static avoidance techniques or simple sensing techniques. However, static avoidance techniques, such as fixed spectrum allocation and static filters, cannot cope with the randomness and dynamism of interference sources. When the interference frequency is unknown or changes rapidly, the communication link is easily interrupted, resulting in poor reliability. Simple sensing techniques, such as spectrum sensing based on energy detection, can only determine whether there is a signal in the frequency band, but cannot identify the specific type and protocol of the signal. This leads to a single decision-making approach, which can only blindly and passively avoid interference and makes it difficult to carry out refined classification and management of interference. Summary of the Invention

[0004] This invention provides a dynamic electromagnetic environment characterization method and system based on artificial intelligence, which solves the problems of low wireless communication reliability caused by the single perception dimension, insufficient classification ability, and passive decision-making strategy in the existing technology under complex dynamic electromagnetic environments.

[0005] This invention provides a dynamic electromagnetic environment characterization method based on artificial intelligence, comprising: Acquire multimodal electromagnetic signals in a dynamic interference environment, and generate a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals; Based on the two-dimensional time-frequency diagram, the cooperative and non-cooperative signals in the dynamic interference environment are determined; Based on the cooperative signals, a communication protocol is characterized to obtain signal coordination instructions; Based on the non-cooperative signal, interference frequency point prediction is performed to obtain an adaptive frequency hopping sequence; Based on the signal coordination command and the adaptive frequency hopping sequence, the communication nodes in the dynamic interference environment are controlled to perform synchronous frequency hopping.

[0006] According to the present invention, a dynamic electromagnetic environment characterization method based on artificial intelligence is provided, wherein determining cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram includes: Based on the signal parameters of the multimodal electromagnetic signal, signal screening is performed to obtain suspected cooperative signals and suspected non-cooperative signals in the dynamic interference environment. Feature extraction is performed based on the two-dimensional time-frequency graph to obtain a high-dimensional feature vector; Based on the high-dimensional feature vector, the suspected cooperative signals and the suspected non-cooperative signals are classified to obtain the cooperative and non-cooperative signals in the dynamic interference environment.

[0007] According to the present invention, a dynamic electromagnetic environment characterization method based on artificial intelligence is provided, wherein the signal parameters include communication frequency band, signal bandwidth, and spectral morphology; The signal screening based on the signal parameters of the multimodal electromagnetic signal to obtain suspected cooperative signals and suspected non-cooperative signals in the dynamic interference environment includes: If the communication frequency band of any signal in the multimodal electromagnetic signals meets the preset frequency band and the signal bandwidth meets the preset communication standard, then any signal is identified as a suspected cooperative signal. If the communication frequency band of any signal in the multimodal electromagnetic signals is not fixed and the spectrum shape is irregular, then any signal is identified as a suspected non-cooperative signal.

[0008] According to the present invention, a dynamic electromagnetic environment characterization method based on artificial intelligence is provided, wherein feature extraction based on the two-dimensional time-frequency diagram to obtain a high-dimensional feature vector includes: Extract the deep abstract features of the two-dimensional time-frequency plot in the time-frequency domain; Based on the deep abstract features, a high-dimensional feature vector containing bandwidth and pulse shape attributes is generated.

[0009] According to the present invention, a dynamic electromagnetic environment characterization method based on artificial intelligence, wherein the step of predicting interference frequency points based on the non-cooperative signal to obtain an adaptive frequency hopping sequence includes: Extract the frequency variation features of the non-cooperative signal in the time dimension to generate a frequency sequence; Based on the jump characteristics of the frequency point sequence in the time dimension, predict the target interference frequency point at the next moment; Based on the target interference frequency and the current list of available channels, an adaptive frequency hopping sequence is generated to avoid the target interference frequency.

[0010] According to the present invention, a dynamic electromagnetic environment characterization method based on artificial intelligence is provided, wherein the communication protocol characterization based on the cooperative signal to obtain signal coordination instructions includes: The cooperation signal is characterized by a communication protocol to obtain the target communication protocol corresponding to the cooperation signal; Based on the protocol standard corresponding to the target communication protocol, a corresponding coordination busy tone signal or backoff guidance signaling is generated as the signal coordination instruction.

[0011] According to the present invention, a dynamic electromagnetic environment characterization method based on artificial intelligence is provided, wherein the multimodal electromagnetic signal is in-phase orthogonal data synchronously collected by multiple electromagnetic signal sensors deployed in the dynamic interference environment; The step of generating a two-dimensional time-frequency diagram based on the multimodal electromagnetic signal includes: The in-phase orthogonal data is segmented to obtain multi-group segmented data; The multi-component segmented data is subjected to Fourier transform to obtain the two-dimensional time-frequency diagram.

[0012] This invention also provides a dynamic electromagnetic environment characterization system based on artificial intelligence, comprising: The acquisition unit is used to acquire multimodal electromagnetic signals in a dynamic interference environment and generate a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals. The identification unit is used to determine the cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram. A characterization unit is used to characterize the communication protocol based on the cooperative signal to obtain signal coordination instructions; The prediction unit is used to predict the interference frequency point based on the non-cooperative signal to obtain an adaptive frequency hopping sequence; The frequency hopping unit is used to control the communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination command and the adaptive frequency hopping sequence.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the dynamic electromagnetic environment characterization method based on artificial intelligence as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the dynamic electromagnetic environment characterization method based on artificial intelligence as described above.

[0015] The present invention provides an artificial intelligence-based dynamic electromagnetic environment characterization method and system that uses a two-dimensional time-frequency diagram to identify interference and distinguish between cooperative and non-cooperative signals. Then, it adopts differentiated governance strategies such as protocol-guided clearing and active prediction and avoidance for different types of signals. Finally, by controlling the synchronous frequency hopping of communication nodes, a highly deterministic and interference-resistant communication link is successfully constructed in an extremely harsh dynamic electromagnetic environment. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the dynamic electromagnetic environment characterization method based on artificial intelligence provided by the present invention. Figure 2 This is a schematic diagram of the structure of the dynamic electromagnetic environment characterization system based on artificial intelligence provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] In critical application scenarios such as smart manufacturing workshops and power monitoring centers, the electromagnetic environment is exceptionally complex. This environment not only contains broadband electromagnetic leakage and radiation signals generated by industrial motors, power lines, and other equipment—collectively referred to as non-cooperative signals—but also a large number of communication networks deployed, including Wi-Fi, 5G (5th-Generation Mobile Communication Technology), and dedicated wireless data transmission networks; these network signals are referred to as cooperative signals. Cooperative and non-cooperative signals intertwine and dynamically change in the time, frequency, and spatial domains, forming a high-density, heterogeneous interference pattern that poses a severe challenge to the reliability and determinism of wireless communication.

[0020] Currently, handling complex interference mainly relies on static avoidance techniques or simple sensing techniques. However, these existing technologies have the following drawbacks: First, static avoidance techniques such as fixed spectrum allocation and static filters cannot cope with the randomness and dynamism of interference sources. When the interference frequency is unknown or changes rapidly, the communication link is easily interrupted, resulting in extremely low reliability. Second, simple sensing techniques such as energy detection-based spectrum sensing can only determine whether a signal "exists" or "is not present" in a certain frequency band, and cannot identify the specific type, protocol, and intent of the signal. This leads to a very simplistic decision-making strategy, resulting in blind, passive avoidance and an inability to perform refined classification and management of interference. Furthermore, existing technologies generally lack intelligent decision-making capabilities, making it difficult to learn patterns from historical interference data and predict future interference behavior. Instead, they remain in a state of reactive remediation after interference occurs.

[0021] In response, this invention provides a dynamic electromagnetic environment characterization method based on artificial intelligence. It aims to convert multimodal electromagnetic signals into two-dimensional time-frequency maps for deep perception and intelligent classification, and to perform protocol characterization on cooperative signals and predict future frequency points for non-cooperative signals. This enables differentiated interference processing that transforms passive blind avoidance into proactive intelligent management, thereby ensuring highly reliable and deterministic wireless communication in complex, dynamic, and harsh electromagnetic environments.

[0022] Figure 1 This is a flowchart illustrating the AI-based dynamic electromagnetic environment characterization method provided by this invention. The execution entity of this method is an AI-based dynamic electromagnetic environment characterization system (hereinafter referred to as the system). This method can be applied to complex, dynamic, and uncontrollable industrial-grade and mission-critical scenarios, such as smart manufacturing workshops and power monitoring centers. Through multimodal perception, intelligent identification, and proactive prediction, it establishes a highly reliable wireless communication link, thereby solving the problems of inefficient static avoidance, blind simple perception, and lack of predictive ability in existing technologies when dealing with complex electromagnetic environments. Figure 1 As shown, the method includes: Step 110: Acquire multimodal electromagnetic signals in a dynamic interference environment, and generate a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals; Step 120: Based on the two-dimensional time-frequency diagram, determine the cooperative and non-cooperative signals in the dynamic interference environment; Step 130: Based on the cooperative signals, the communication protocol is characterized to obtain the signal coordination instructions; Step 140: Based on the non-cooperative signal, predict the interference frequency point to obtain the adaptive frequency hopping sequence; Step 150: Based on signal coordination instructions and adaptive frequency hopping sequences, control the communication nodes in the dynamic interference environment to perform synchronous frequency hopping.

[0023] Specifically, when faced with interference from various frequency bands, cluttered spectra, and dynamic interference from external communication devices that crowd out the channel, the system first uses multiple distributed electromagnetic signal sensors, such as software-defined radio (SDR), to synchronously acquire multi-mode electromagnetic signals in the environment in real time at a sampling rate of no less than 20MHz, such as raw I / Q (In-phase / Quadrature) data.

[0024] Next, the system needs to perform transformation processing on these multimodal electromagnetic signals, such as Fourier transform, to convert them into a two-dimensional time-frequency graph that can intuitively display the energy distribution of the signal in the time and frequency domains. This process is like transforming intangible electromagnetic signals into tangible images, thus providing a reliable data foundation for subsequent deep feature extraction and accurate profiling.

[0025] After obtaining the two-dimensional time-frequency map, the system uses a pre-trained model to extract deep abstract features from the map, such as modulation scheme, bandwidth, and pulse shape, to accurately determine the cooperative and non-cooperative signals present in the dynamic interference environment. Cooperative signals refer to signals in the dynamic interference environment that conform to specific communication standards, have regular spectral patterns, and are interactive, such as Wi-Fi, 5G, and Bluetooth signals. Non-cooperative signals, on the other hand, refer to interference signals with variable frequency bands, chaotic spectra, high randomness, and inability to interact via protocols, such as high-power industrial motor noise and electromagnetic leakage from power lines. Here, the signal classification method based on multimodal features extracted from the two-dimensional time-frequency map completely eliminates the blindness of traditional single-energy detection, achieving accurate identification of interference sources.

[0026] Furthermore, after accurately classifying the signals, the system will adopt differentiated intelligent governance strategies. That is, for the identified cooperative signals, the system will perform in-depth communication protocol characterization, that is, deeply identify and analyze their specific communication protocol types, and generate corresponding signal coordination instructions accordingly.

[0027] The signal coordination command here refers to the control signaling used to guide external cooperative signal sources to actively back off, such as a virtual coordination busy tone signal or backoff guidance signaling. Through this protocol-level coordination and guidance, the system utilizes existing communication protocols, such as the Listen-Before-Speak (LBT) mechanism in the Wi-Fi standard, to cause external cooperative signals to mistakenly believe that the channel is congested and actively give up the channel, thereby creating a time window for its own communication.

[0028] Meanwhile, for non-cooperative signals that cannot interact via protocols, the system records their frequency sequence and predicts the timing of interference frequencies. By analyzing the historical hopping patterns of the frequency sequence, it accurately infers the location of the interference that the non-cooperative signal will encounter in the next moment. Based on this, the system can generate an adaptive frequency hopping sequence in advance to guide its own communication equipment to avoid interference. This achieves an upgrade from passively switching channels after interference occurs to proactively switching channels in advance based on the prediction of impending interference, greatly reducing the risk of communication interruption.

[0029] Following this, the system can then send out the generated signal coordination commands and adaptive frequency hopping sequences through a dedicated secure control channel, such as a ZigBee network, to control friendly communication devices / nodes located in dynamic interference environments, such as automated guided vehicles (AGVs) performing critical tasks, to synchronize frequency hopping. During this process, communication nodes can release signal coordination commands to clear cooperative signals, and strictly follow the adaptive frequency hopping sequence to synchronously and in advance switch to a safe, clean frequency point for communication before the predicted interference arrives.

[0030] The present invention provides an artificial intelligence-based dynamic electromagnetic environment characterization method that uses a two-dimensional time-frequency diagram to identify interference and distinguish between cooperative and non-cooperative signals. Then, it adopts differentiated governance strategies such as protocol-guided clearing and proactive prediction and avoidance for different types of signals. Finally, by controlling the synchronous frequency hopping of communication nodes, a highly deterministic and interference-resistant communication link is successfully constructed in an extremely harsh dynamic electromagnetic environment.

[0031] Based on the above embodiments, step 120 includes: Signal screening is performed based on the signal parameters of multimodal electromagnetic signals to obtain suspected cooperative signals and suspected non-cooperative signals in dynamic interference environments. Feature extraction is performed based on two-dimensional time-frequency graphs to obtain high-dimensional feature vectors; Based on high-dimensional feature vectors, suspected cooperative signals and suspected non-cooperative signals are classified to obtain cooperative and non-cooperative signals in dynamic interference environments.

[0032] Specifically, the process of determining cooperative and non-cooperative signals in a dynamic interference environment based on a two-dimensional time-frequency diagram may include: After acquiring the multimodal electromagnetic signals, the system first extracts their macroscopic signal parameters, which are various parameters that reflect the basic physical properties and statistical characteristics of the signals, such as communication frequency band, signal bandwidth, basic spectral morphology, and average power. Next, the system uses preset rules to quickly perform signal screening based on these parameters—a preliminary screening—to assess whether these signal parameters conform to certain known macroscopic laws. This initially divides the chaotic signals into two categories: one category consists of signals whose parameters appear to conform to specific communication laws or set conditions, i.e., suspected cooperative signals; the other category consists of signals with random parameters, non-fixed frequency bands, and irregular morphologies, i.e., suspected non-cooperative signals.

[0033] After signal screening, to overcome the limitations of manually designed macroscopic rules in handling complex features, the system further utilizes deep learning models, such as pre-trained Convolutional Neural Networks (CNNs), to perform deep feature extraction on the two-dimensional time-frequency graph. During this process, the model automatically scans the two-dimensional time-frequency graph, extracting deeper and more abstract subtle features of the electromagnetic signal in the time-frequency domain, such as bandwidth attributes, pulse shape attributes, and modulation latent features. The system then performs dimensionality reduction mapping and structured integration on these extracted complex features, ultimately generating a high-dimensional feature vector that represents the essential properties of the electromagnetic signal.

[0034] Following this, the system enters the fine screening stage. Using a pre-trained classification model, such as a Support Vector Machine (SVM), based on high-dimensional feature vectors, it performs high-precision secondary classification of the suspected cooperative and suspected non-cooperative signals obtained from the initial screening. Specifically, the classification model can correct potential misjudgments in the initial screening stage by analyzing the high-dimensional feature vectors, accurately determining whether the signal is friend or foe, thereby identifying truly existing, interactive cooperative signals and random, uncontrollable non-cooperative signals in a dynamic interference environment.

[0035] In this embodiment of the invention, a two-level classification architecture is introduced that combines initial screening based on signal parameters with precise classification based on high-dimensional feature vectors. This architecture utilizes the high efficiency and low computational overhead of initial screening based on macroscopic parameters, while leveraging the high accuracy of high-dimensional feature vectors in complex environments. This enables rapid and accurate classification of cooperative and non-cooperative signals, providing reliable data for the subsequent system to adopt differentiated avoidance and coordination strategies.

[0036] Based on the above embodiments, the signal parameters include communication frequency band, signal bandwidth, and spectrum configuration; Signal screening based on signal parameters of multimodal electromagnetic signals yields suspected cooperative and suspected non-cooperative signals in dynamic interference environments, including: If any signal in a multimodal electromagnetic signal meets the preset frequency band and the signal bandwidth conforms to the preset communication standard, the signal is identified as a suspected cooperative signal. In multimodal electromagnetic signals, if the communication frequency band of any signal is not fixed and the spectrum shape is irregular, the signal is identified as a suspected non-cooperative signal.

[0037] Specifically, the signal parameters used for initial screening include the communication frequency band, signal bandwidth, and spectral characteristics. These three are core indicators reflecting the macroscopic physical layer characteristics of electromagnetic signals. After acquiring multimodal electromagnetic signals, the system quickly analyzes the core indicators of any one of the signals and executes the following two parallel screening rules: Firstly, when a signal's communication frequency band meets a preset frequency band and its bandwidth conforms to a preset communication standard, the signal is identified as a suspected cooperative signal. That is, because legitimate communication devices must comply with specific radio management regulations and protocol specifications, the system detects the frequency location of the signal. Here, the preset frequency band refers to various known specific frequency ranges allocated for standard wireless network communication, such as 2.4GHz; and the preset communication standard refers to the industry-standard underlying communication protocol specifications, such as whether the bandwidth precisely conforms to the IEEE 802.11 Wi-Fi standard, or the specific bandwidth requirements of the 5G / Bluetooth protocol. When the system detects a signal that falls within the aforementioned fixed frequency band and whose occupied spectrum width matches the standard protocol, it is highly likely a signal emitted by a normally deployed, interactive communication node in the environment; therefore, the system marks it as a suspected cooperative signal.

[0038] Secondly, when the communication frequency band of the signal is not fixed and the spectral pattern is irregular, the signal is identified as a suspected non-cooperative signal. That is, because severe interference in industrial environments often does not follow any communication protocols, the system will capture anomalous electromagnetic energy exhibiting highly randomness. Here, "not fixed communication frequency band" means that the signal energy exhibits characteristics of random jumps, aimless movement, or no fixed center frequency in the frequency domain; while "irregular spectral pattern" means that it does not possess the flat-top, symmetrical, and specific roll-off regular shape of standard communication signals, often manifesting as broadband glitches, sudden chaotic spikes, or irregular energy leakage, such as the broadband electromagnetic noise generated by the frequent start-stop of high-power industrial motors in smart factory scenarios, or interference caused by power line radiation. Faced with such erratic signals that cannot be constrained by protocols, the system will classify and mark them as suspected non-cooperative signals for subsequent avoidance.

[0039] In this embodiment of the invention, by specifying the communication frequency band, signal bandwidth, and spectrum shape as the discrimination indicators for the initial screening of macroscopic rules, the significant differences between non-cooperative electromagnetic leakage and standard cooperative communication signals in the macroscopic physical layer of industrial sites are fully utilized. This macroscopic parameter direct screening strategy based on prior knowledge can quickly filter out interference sources with distinct characteristics with extremely low computing power overhead. This not only significantly reduces the amount of data processing in the subsequent fine screening stage, but also improves the real-time performance and response efficiency of the entire system.

[0040] Based on the above embodiments, feature extraction is performed based on the two-dimensional time-frequency graph to obtain a high-dimensional feature vector, including: Extracting deep abstract features of two-dimensional time-frequency plots in the time-frequency domain; Based on deep abstract features, a high-dimensional feature vector containing bandwidth and pulse shape attributes is generated.

[0041] Specifically, addressing the problem of low accuracy in traditional manual feature extraction methods due to severe signal overlap and distortion in complex electromagnetic environments, this invention employs a method where deep feature extraction is first performed in the time-frequency domain when extracting features from a two-dimensional time-frequency graph. This involves inputting the two-dimensional time-frequency graph into a pre-trained deep learning model, such as a convolutional neural network (CNN) model with multiple convolutional and pooling layers, for deep visual scanning. During this process, the model automatically extracts deep abstract features from the two-dimensional time-frequency graph in both the time and frequency domains. Unlike traditional shallow, intuitive parameters such as peak energy and average power, these deep abstract features refer to high-dimensional nonlinear features hidden deep within the texture of the two-dimensional time-frequency graph, difficult to describe with simple mathematical formulas, yet fundamentally reflecting the internal modulation patterns, transition rules, and energy distribution of electromagnetic signals. Here, through layer-by-layer propagation and nonlinear mapping of the deep learning model, the system can extract the essential features that truly distinguish different signal sources from the seemingly chaotic two-dimensional time-frequency graph (frequency domain waterfall plot).

[0042] After extracting deep features, the system further performs dimensionality reduction mapping and structured integration on these extracted deep abstract features, ultimately generating feature vectors that represent the essential attributes of electromagnetic signals. These feature vectors contain both bandwidth and pulse shape attributes. The bandwidth attribute reflects the width of the signal in the frequency domain, such as the broadband characteristic conforming to the 802.11 standard or the irregularly wide bandwidth of an industrial motor leakage. The pulse shape attribute reflects the contour characteristics of the signal during energy bursts in the time domain, such as the regular square pulse in communication protocols or the sharp, sudden, and distorted pulse generated by electrical sparks.

[0043] In this embodiment of the invention, a mechanism is adopted to extract deep abstract features in the time-frequency domain and generate high-dimensional feature vectors containing bandwidth and pulse shape attributes. This completely eliminates the limitations and fragility of traditional manually designed features, and fully leverages the superior performance of deep learning models in processing two-dimensional image textures. It can accurately extract the essential physical properties of signals from harsh and overlapping dynamic interference environments, thereby greatly improving the system's recognition accuracy and robustness against cooperative and non-cooperative signals.

[0044] Based on the above embodiments, step 140 includes: Extract the frequency variation features of non-cooperative signals in the time dimension to generate frequency sequence; Based on the jump characteristics of the frequency point sequence in the time dimension, predict the target interference frequency point at the next moment; Based on the target interference frequency and the list of currently available channels, an adaptive frequency hopping sequence is generated to avoid the target interference frequency.

[0045] Specifically, the process of predicting interference frequencies based on non-cooperative signals to obtain an adaptive frequency hopping sequence may include: After locking onto a non-cooperative signal, the system initiates a continuous time-domain tracking mechanism. That is, according to a set time window, it continuously records the core frequency position or energy peak frequency band of the non-cooperative signal at different historical moments, such as t-2, t-1, and t, thereby extracting the frequency variation characteristics of the non-cooperative signal in the time dimension. This characteristic reflects the dynamic physical law that the interference source is not static but rather roams freely across the spectrum. The system arranges this series of historical interference frequency data recorded over time in chronological order to construct a frequency sequence with time-dependent relationships. For example, it records the trajectory of the motor interference frequency jumping from 2.43 GHz to 2.44 GHz, and then to 2.45 GHz.

[0046] After acquiring sufficient historical tracking data, the system inputs the generated frequency sequence into a prediction model adept at processing time series data, such as a pre-trained Long Short-Term Memory (LSTM) network model, for deep analysis. Through deep learning, the prediction model uncovers and understands the hidden patterns and characteristics of the frequency sequence's transitions over time, distinguishing whether the non-cooperative signal exhibits linear sliding, periodic transitions, or complex transitions with some hidden Markov chain characteristics. Based on this in-depth understanding of patterns, the model can accurately predict the next moment, such as the specific frequency location at time t+1 where the interference source is most likely to transition and land, i.e., the future interference frequency (the target interference frequency at the next moment). For example, it can accurately predict that motor interference will occur at 2.46 GHz in the next second.

[0047] Once the target interference frequency is identified, the system immediately retrieves a pre-maintained list of currently available channels in its communication network—a set of legal communication channels that are currently unoccupied and have clean spectrum. The system compares the predicted target interference frequency with this list, selects a clean channel far from the target interference frequency, and generates an adaptive frequency-hopping sequence to guide the spatial positioning of its communication equipment. For example, if interference is predicted to strike 2.46GHz in the next second, the system will select the clean 2.40GHz channel from the available channels and generate a command for all AGVs to uniformly switch to 2.40GHz in the next second. Thus, when the interference actually occurs at 2.46GHz, the communication equipment is already communicating on a clean 2.40GHz channel, achieving perfect peak avoidance and mitigation.

[0048] In this embodiment of the invention, for non-cooperative signals that cannot perform protocol interaction, frequency point sequence extraction and timing pattern prediction in the time dimension are introduced. By completely subverting the traditional passive anti-interference mode of being forced to switch channels due to link breakage caused by interference, a proactive strategy of actively switching channels in advance when interference is predicted is completely overturned. This not only achieves predictive adaptive avoidance, but also minimizes the risk of communication interruption in extremely harsh electromagnetic environments, fundamentally ensuring the high determinism and high reliability of industrial-grade critical task wireless communication links.

[0049] Based on the above embodiments, step 130 includes: The cooperation signal is characterized by a communication protocol to obtain the target communication protocol corresponding to the cooperation signal; Based on the protocol standard corresponding to the target communication protocol, a corresponding coordination busy tone signal or backoff guidance signaling is generated as a signal coordination instruction.

[0050] Specifically, the process of characterizing the communication protocol based on the cooperative signal to obtain the signal coordination command may include: After classifying a signal as a cooperative signal, the system inputs it into a dedicated protocol identification module, such as a pre-trained shallow neural network classifier, to perform a more granular communication protocol representation of the signal. This communication protocol representation refers to accurately identifying the specific language or specification used by the signal. Specifically, the classifier analyzes the signal's microscopic manifestations in frame structure, pilot sequences, and physical layer header features to ultimately output the specific communication protocol followed by the cooperative signal, i.e., the target communication protocol. The target communication protocol covers various standardized wireless communication technologies commonly used in industrial environments, such as the IEEE 802.11 standard (Wi-Fi signal), the 3GPP standard (5G signal), and the IEEE 802.15.1 standard (Bluetooth signal).

[0051] After accurately identifying the target communication protocol, the system immediately invokes the underlying operational logic and media access control layer rules of that protocol, i.e., the protocol standard. As is well known, most standard communication protocols, especially Wi-Fi, strictly adhere to the Listen-Before-Speak or Carrier Sense Multiple Access (CSMA) mechanism to avoid data collisions. The system exploits this weakness by generating virtual coordination busy tones or sending backoff guidance signals in specific time slots based on the protocol standard.

[0052] Here, the coordination busy tone signal is a fake placeholder signal that misleads specific protocol devices into believing that the channel is currently extremely congested or occupied; the backoff guidance signal is a control frame broadcast directly to the environment, triggering the other device to start its backoff timer. The system transmits these virtual busy tones or signaling as a unified signal coordination command. This is similar to turning on a red light or setting up virtual roadblocks for vehicles that obey traffic rules (cooperation signal) on the road, using protocol guidance to induce other external communication devices in the environment to actively back off and give way, thereby proactively creating an interference-free time window for its own critical communication equipment.

[0053] In this embodiment of the invention, a refined governance mechanism combining communication protocol characterization and protocol guidance clearing is proposed for cooperative signals. By accurately identifying the target communication protocol and utilizing its inherent underlying channel contention mechanisms such as listening before speaking, a coordinated busy tone signal or backoff guidance signaling is generated in a targeted manner, successfully turning passive into active. This not only effectively avoids the additional overhead and link loss risk caused by frequent frequency hopping of the local communication equipment, but also maximizes the utilization of limited spectrum resources, achieving harmonious coexistence and efficient communication guarantee in a complex multi-network coexistence environment.

[0054] Based on the above embodiments, the multimodal electromagnetic signal is in-phase orthogonal data synchronously collected by multiple electromagnetic signal sensors deployed in a dynamic interference environment; In step 110, a two-dimensional time-frequency diagram is generated based on the multimodal electromagnetic signal, including: Segmenting in-phase orthogonal data yields multi-group segmented data. A two-dimensional time-frequency graph is obtained by performing a Fourier transform on the multi-group segmented data.

[0055] Specifically, the process of generating a two-dimensional time-frequency diagram based on multimodal electromagnetic signals includes the following steps: In places such as 500m 2In dynamic interference environments such as smart manufacturing workshops, the system spatially deploys multiple electromagnetic signal sensors, such as software-defined radio (SDR), distributed across the four corners and center of the workshop. These sensors work collaboratively to synchronously acquire electromagnetic signals within the space at a sampling rate of at least 20MHz. The raw I / Q data obtained is the multimodal electromagnetic signal. Unlike traditional scalars that only reflect signal strength, I / Q data contains information from two mutually orthogonal dimensions: the I-path (in-phase) and the Q-path (quadrature). It can completely preserve the physical baseband characteristics of the electromagnetic signal, such as instantaneous amplitude, phase, and frequency.

[0056] Since the I / Q data continuously and synchronously acquired by the sensor is an infinite and extremely large time-domain data stream, it cannot be directly used for static image feature extraction. Therefore, the system segments this continuous I / Q data stream, i.e., it applies a windowing technique, slicing the continuous data stream according to a fixed time window length or a set number of sampling points, thus obtaining multiple sets of segmented data. For example, the system can truncate every 1024 consecutive I / Q sampling points into a segment, forming an independent set of segmented data. In this way, the infinite time series is discretized into tiny time slices that are easy to perform matrix operations.

[0057] Furthermore, after slicing the data, the system then uses mathematical transformation tools, such as the short-time Fourier transform, to transform each group of segmented data one by one. This transforms each group of time-domain segmented data, containing 1024 sampling points, into its corresponding frequency-domain spectrum, thus resolving the frequency components and their energy levels contained in the signal within that extremely short time period. The system then stacks these spectral graphs, arranged in chronological order for each tiny time interval, along the time axis to ultimately render a two-dimensional time-frequency graph, also known as a spectral waterfall plot.

[0058] In a two-dimensional time-frequency graph, the horizontal axis represents time, the vertical axis represents frequency, and the color intensity or pixel brightness of the image represents the strength of electromagnetic energy, realizing cross-modal conversion from one-dimensional I / Q digital signals to two-dimensional intuitive electromagnetic images.

[0059] In this embodiment of the invention, I / Q data are synchronously collected by distributed sensors, ensuring the absolute integrity of electromagnetic signal characteristics and the synergy of spatial dimensions from the source of physical perception. Furthermore, through rigorous segmentation processing combined with Fourier transform operations, massive and abstract one-dimensional time-domain digital signals are successfully and efficiently converted into two-dimensional time-frequency diagrams rich in temporal dynamics and spectral details, providing an extremely high-definition and feature-rich data foundation for subsequent processing.

[0060] The following describes the AI-based dynamic electromagnetic environment characterization system provided by the present invention. The AI-based dynamic electromagnetic environment characterization system described below can be referred to in correspondence with the AI-based dynamic electromagnetic environment characterization method described above.

[0061] Figure 2 This is a schematic diagram of the structure of the dynamic electromagnetic environment characterization system based on artificial intelligence provided by the present invention, as shown below. Figure 2 As shown, the system includes: The acquisition unit 210 is used to acquire multimodal electromagnetic signals in a dynamic interference environment and generate a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals. The identification unit 220 is used to determine the cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram. The characterization unit 230 is used to characterize the communication protocol based on the cooperative signal to obtain signal coordination instructions; Prediction unit 240 is used to predict interference frequency points based on the non-cooperative signal to obtain an adaptive frequency hopping sequence; Frequency hopping unit 250 is used to control communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination command and the adaptive frequency hopping sequence.

[0062] The AI-based dynamic electromagnetic environment characterization system provided by this invention uses a two-dimensional time-frequency diagram to identify interference and distinguish between cooperative and non-cooperative signals. Then, it adopts differentiated governance strategies such as protocol-guided clearing and proactive prediction and avoidance for different types of signals. Finally, by controlling the synchronous frequency hopping of communication nodes, it successfully constructs a highly deterministic and interference-resistant communication link in an extremely harsh dynamic electromagnetic environment.

[0063] Based on the above embodiments, the identification unit 220 is used for: Based on the signal parameters of the multimodal electromagnetic signal, signal screening is performed to obtain suspected cooperative signals and suspected non-cooperative signals in the dynamic interference environment. Feature extraction is performed based on the two-dimensional time-frequency graph to obtain a high-dimensional feature vector; Based on the high-dimensional feature vector, the suspected cooperative signals and the suspected non-cooperative signals are classified to obtain the cooperative and non-cooperative signals in the dynamic interference environment.

[0064] Based on the above embodiments, the signal parameters include communication frequency band, signal bandwidth, and spectrum configuration; The identification unit 220 is used for: If the communication frequency band of any signal in the multimodal electromagnetic signals meets the preset frequency band and the signal bandwidth meets the preset communication standard, the signal is identified as a suspected cooperative signal. If the communication frequency band of any signal in the multimodal electromagnetic signal is not fixed and the spectrum shape is irregular, the signal is identified as a suspected non-cooperative signal.

[0065] Based on the above embodiments, the identification unit 220 is used for: Extract the deep abstract features of the two-dimensional time-frequency plot in the time-frequency domain; Based on the deep abstract features, a high-dimensional feature vector containing bandwidth and pulse shape attributes is generated.

[0066] Based on the above embodiments, the prediction unit 240 is used for: Extract the frequency variation features of the non-cooperative signal in the time dimension to generate a frequency sequence; Based on the jump characteristics of the frequency point sequence in the time dimension, predict the target interference frequency point at the next moment; Based on the target interference frequency and the current list of available channels, an adaptive frequency hopping sequence is generated to avoid the target interference frequency.

[0067] Based on the above embodiments, the characterization unit 230 is used for: The cooperation signal is characterized by a communication protocol to obtain the target communication protocol corresponding to the cooperation signal; Based on the protocol standard corresponding to the target communication protocol, a corresponding coordination busy tone signal or backoff guidance signaling is generated as the signal coordination instruction.

[0068] Based on the above embodiments, the multimodal electromagnetic signal is in-phase orthogonal data synchronously collected by multiple electromagnetic signal sensors deployed in the dynamic interference environment; The acquisition unit 210 is used for: The in-phase orthogonal data is segmented to obtain multi-group segmented data; The multi-component segmented data is subjected to Fourier transform to obtain the two-dimensional time-frequency diagram.

[0069] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an artificial intelligence-based dynamic electromagnetic environment characterization method. This method includes: acquiring multimodal electromagnetic signals in a dynamic interference environment and generating a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals; determining cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram; characterizing communication protocols based on the cooperative signals to obtain signal coordination instructions; predicting interference frequency points based on the non-cooperative signals to obtain an adaptive frequency hopping sequence; and controlling communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination instructions and the adaptive frequency hopping sequence.

[0070] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the dynamic electromagnetic environment characterization method based on artificial intelligence provided by the above methods, the method comprising: acquiring multimodal electromagnetic signals in a dynamic interference environment, and generating a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals; determining cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram; performing communication protocol characterization based on the cooperative signals to obtain signal coordination instructions; performing interference frequency point prediction based on the non-cooperative signals to obtain an adaptive frequency hopping sequence; and controlling communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination instructions and the adaptive frequency hopping sequence.

[0072] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the artificial intelligence-based dynamic electromagnetic environment characterization method provided by the methods described above. This method includes: acquiring multimodal electromagnetic signals in a dynamic interference environment and generating a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals; determining cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram; characterizing communication protocols based on the cooperative signals to obtain signal coordination instructions; predicting interference frequency points based on the non-cooperative signals to obtain an adaptive frequency hopping sequence; and controlling communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination instructions and the adaptive frequency hopping sequence.

[0073] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0074] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic electromagnetic environment characterization method based on artificial intelligence, characterized in that, include: Acquire multimodal electromagnetic signals in a dynamic interference environment, and generate a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals; Based on the two-dimensional time-frequency diagram, the cooperative and non-cooperative signals in the dynamic interference environment are determined; Based on the cooperative signals, a communication protocol is characterized to obtain signal coordination instructions; Based on the non-cooperative signal, interference frequency point prediction is performed to obtain an adaptive frequency hopping sequence; Based on the signal coordination command and the adaptive frequency hopping sequence, the communication nodes in the dynamic interference environment are controlled to perform synchronous frequency hopping.

2. The dynamic electromagnetic environment characterization method based on artificial intelligence according to claim 1, characterized in that, The determination of cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram includes: Based on the signal parameters of the multimodal electromagnetic signal, signal screening is performed to obtain suspected cooperative signals and suspected non-cooperative signals in the dynamic interference environment. Feature extraction is performed based on the two-dimensional time-frequency graph to obtain a high-dimensional feature vector; Based on the high-dimensional feature vector, the suspected cooperative signals and the suspected non-cooperative signals are classified to obtain the cooperative and non-cooperative signals in the dynamic interference environment.

3. The dynamic electromagnetic environment characterization method based on artificial intelligence according to claim 2, characterized in that, The signal parameters include the communication frequency band, signal bandwidth, and spectrum configuration; The signal screening based on the signal parameters of the multimodal electromagnetic signal to obtain suspected cooperative signals and suspected non-cooperative signals in the dynamic interference environment includes: If the communication frequency band of any signal in the multimodal electromagnetic signals meets the preset frequency band and the signal bandwidth meets the preset communication standard, then any signal is identified as a suspected cooperative signal. If the communication frequency band of any signal in the multimodal electromagnetic signals is not fixed and the spectrum shape is irregular, then any signal is identified as a suspected non-cooperative signal.

4. The method for characterizing dynamic electromagnetic environments based on artificial intelligence according to claim 2, characterized in that, The feature extraction based on the two-dimensional time-frequency graph to obtain a high-dimensional feature vector includes: Extract the deep abstract features of the two-dimensional time-frequency plot in the time-frequency domain; Based on the deep abstract features, a high-dimensional feature vector containing bandwidth and pulse shape attributes is generated.

5. The dynamic electromagnetic environment characterization method based on artificial intelligence according to any one of claims 1 to 4, characterized in that, The step of predicting interference frequency points based on the non-cooperative signal to obtain an adaptive frequency hopping sequence includes: Extract the frequency variation features of the non-cooperative signal in the time dimension to generate a frequency sequence; Based on the jump characteristics of the frequency point sequence in the time dimension, predict the target interference frequency point at the next moment; Based on the target interference frequency and the current list of available channels, an adaptive frequency hopping sequence is generated to avoid the target interference frequency.

6. The method for characterizing dynamic electromagnetic environments based on artificial intelligence according to any one of claims 1 to 4, characterized in that, The step of characterizing the communication protocol based on the cooperative signal to obtain signal coordination instructions includes: The cooperation signal is characterized by a communication protocol to obtain the target communication protocol corresponding to the cooperation signal; Based on the protocol standard corresponding to the target communication protocol, a corresponding coordination busy tone signal or backoff guidance signaling is generated as the signal coordination instruction.

7. The method for characterizing dynamic electromagnetic environments based on artificial intelligence according to any one of claims 1 to 4, characterized in that, The multimodal electromagnetic signals are in-phase orthogonal data synchronously collected by multiple electromagnetic signal sensors deployed in the dynamic interference environment; The step of generating a two-dimensional time-frequency diagram based on the multimodal electromagnetic signal includes: The in-phase orthogonal data is segmented to obtain multi-group segmented data; The multi-component segmented data is subjected to Fourier transform to obtain the two-dimensional time-frequency diagram.

8. A dynamic electromagnetic environment characterization system based on artificial intelligence, characterized in that, include: The acquisition unit is used to acquire multimodal electromagnetic signals in a dynamic interference environment and generate a two-dimensional time-frequency diagram based on the multimodal electromagnetic signals. The identification unit is used to determine the cooperative and non-cooperative signals in the dynamic interference environment based on the two-dimensional time-frequency diagram. A characterization unit is used to characterize the communication protocol based on the cooperative signal to obtain signal coordination instructions; The prediction unit is used to predict the interference frequency point based on the non-cooperative signal to obtain an adaptive frequency hopping sequence; The frequency hopping unit is used to control the communication nodes in the dynamic interference environment to perform synchronous frequency hopping based on the signal coordination command and the adaptive frequency hopping sequence.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the dynamic electromagnetic environment characterization method based on artificial intelligence as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the dynamic electromagnetic environment characterization method based on artificial intelligence as described in any one of claims 1 to 7.

Images