5g and beidou integrated passive room division electromagnetic environment adaptive system

Abstract

The application belongs to the technical field of computer systems based on specific calculation models, and discloses a 5G and Beidou integrated passive room electromagnetic environment adaptive system, which comprises a bypass channel monitoring module, a knowledge base module, a cause-effect space-time consistency arbitrator, an intention deduction module and a strategy execution module; the system utilizes the cause-effect path atlas of the space-time logical relationship of event propagation recorded in the knowledge base module to arbitrate the collected bypass channel state information to confirm effective event fingerprints, and only after confirmation, matching and pre-judgment execution of the optimal response strategy are performed; the application moves the trigger point of system response from the result level of signal deterioration to the cause level of environmental event occurrence, and realizes the mode change from lagging passive compensation to cause-effect prediction.

Description

Technical Field

[0001] This invention relates to a passive indoor electromagnetic environment adaptive system integrating 5G and BeiDou, belonging to the field of computer system technology based on a specific calculation model. Background Technology

[0002] In current indoor wireless communication network deployments, a passive feedback adjustment method based on signal quality is commonly used to address the dynamic electromagnetic environment caused by factors such as the start-up and shutdown of personnel and mobile devices. This method continuously monitors the strength and quality of 5G or BeiDou signals and triggers a passive distributed antenna system or reconfigurable smart metasurface to make compensatory adjustments after detecting performance degradation. This method has proven effective in handling static or slowly changing environments and is a fundamental technology in this field.

[0003] However, in concealed indoor spaces such as underground locations and tunnels, when the system is used to provide positioning services for robots, equipment, and especially high-speed moving vehicles, the dynamic changes in the indoor electromagnetic environment are frequent and rapid. The aforementioned passive feedback-based adjustment method, due to its inherent operating mechanism, limits its application effectiveness. This mechanism dictates that the system response inevitably occurs after communication quality has deteriorated. For high-speed moving automated equipment, such as unmanned navigation vehicles navigating between metal shelves, this response delay may lead to communication interruptions or positioning signal jumps at critical path points, resulting in operational disruptions. Attempting to shorten the delay by increasing the signal monitoring frequency or the deployment density of monitoring points directly increases computing resources and hardware costs. This approach does not change the post-compensation working method but forces the system to face a direct technical trade-off between response time and deployment costs. Furthermore, this... The bottleneck in this technology is not simply a matter of control strategy. Even existing technologies that attempt to enhance the environmental tolerance of devices at the level of physical structure have failed to fundamentally solve the core problem of response lag. For example, Chinese utility model patent application CN217546619U discloses a multi-antenna calibration network device for 5G communication. This solution aims to improve the electromagnetic shielding and physical protection performance of the device itself by constructing a complex physical structure consisting of a reinforcing layer, a shock-absorbing layer, multiple anti-interference layers, an insulating layer, a waterproof and heat dissipation layer, etc. The essence of this design approach is still to create a more robust shield to passively resist environmental electromagnetic interference that has already occurred. It does not touch the level of early perception and prediction of environmental changes. Therefore, when facing high-speed dynamic scenarios, it still cannot avoid the inherent delay problem caused by compensation only after the actual deterioration of communication quality.

[0004] The root cause of this technological limitation lies in the fact that existing methods need to handle a physical inverse problem based on sparse and noisy samples, namely, real-time fitting of high-dimensional electromagnetic fields generated by the interaction of multiple objects in the environment. This has fundamental limitations in information acquisition and processing. Existing technologies mainly suffer from the following shortcomings: 1. The triggering of its adaptive adjustment is based on passive monitoring of communication signal degradation, resulting in a lag in response and making it difficult to meet the service continuity requirements of high-speed dynamic scenarios; 2. Its decision-making is based solely on the primary communication signal itself. When the signal is distorted due to deep fading or strong interference, the system will lose effective input for environmental perception and adaptive adjustment, resulting in insufficient reliability; 3. To shorten the response time, the system needs to continuously improve its fitting and predictive capabilities for complex electromagnetic physical fields, which correspondingly increases the complexity of the computational model and the requirements for edge computing resources, restricting its economic viability and the feasibility of large-scale deployment. Therefore, how to construct a computational model that enables its adjustment mechanism, causal event learning and identification, and thus achieve proactive and predictive adaptation to the electromagnetic environment without increasing resource consumption, becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a passive indoor electromagnetic environment adaptive system integrating 5G and BeiDou. Its main purpose is to solve the inherent response lag of existing technologies that rely on passive feedback adjustment of the communication signal itself, and the difficulty in making advance predictions of dynamic electromagnetic environments without increasing resource consumption.

[0006] To achieve the above objectives, the present invention provides a 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system, the system comprising:

[0007] The bypass channel monitoring module is configured to collect channel status information from multiple wireless bypass signal monitoring nodes covering the same indoor space;

[0008] The knowledge base module is configured to store event fingerprints corresponding to preset indoor environmental events, as well as a causal path graph, which records the spatiotemporal logical relationship of the event fingerprints corresponding to preset indoor environmental events propagating among multiple wireless bypass signal monitoring nodes.

[0009] The causal spatiotemporal consistency arbiter is configured to: before processing the channel state information collected by any wireless bypass signal monitoring node, first verify whether the channel state information is consistent with the spatiotemporal logical relationship of a certain event recorded in the causal path graph, and only when the verification is consistent will the channel state information be confirmed as a valid event fingerprint.

[0010] The intent deduction module is configured to match the valid event fingerprint with the event fingerprint stored in the knowledge base module only when a valid event fingerprint is received from the causal spatiotemporal consistency arbiter, so as to identify the preset indoor environmental event that is about to occur.

[0011] The strategy execution module is configured to extract and execute the optimal response strategy from the knowledge base module based on the recognition results of the intent inference module.

[0012] Preferably, the causal path graph stored in the knowledge base module is a topology graph established with multiple wireless bypass signal monitoring nodes as nodes; the spatiotemporal logical relationship specifically includes the event fingerprint corresponding to the preset indoor environmental event, the preset time delay collected sequentially on two adjacent wireless bypass signal monitoring nodes along the preset propagation path, and the signal strength evolution relationship presented on the two wireless bypass signal monitoring nodes.

[0013] Preferably, the system also includes a guardian module, which is configured to monitor the real-time quality of 5G or BeiDou signals; and when the real-time quality value is lower than a preset quality threshold or is not higher than the value before the execution of the optimal response strategy after the strategy execution module executes the optimal response strategy, the guardian module will be configured to switch the system's control mode to a control mode based on passive feedback adjustment of real-time quality.

[0014] Preferably, the system further includes a response feature excitation and a fuzzy arbitrator, which is configured such that: when the matching result of the intent inference module satisfies a preset fuzzy condition, i.e. , ,and At that time, the arbitration process is activated, in which and These represent the confidence levels of two candidate indoor environmental events. The confidence threshold is determined based on historical statistical data. The fuzzy judgment threshold is determined based on a preset percentage of the confidence threshold. After activation, the control strategy execution module applies a preset transient perturbation excitation configuration to the passive indoor distribution network. While applying the excitation configuration, the transient response characteristics of the bypass channel state information caused by the excitation configuration are collected by the bypass channel monitoring module. Based on the transient response characteristics, a unique indoor environmental event is determined from multiple candidate indoor environmental events that meet the fuzzy conditions.

[0015] Preferably, the instantaneous perturbation excitation configuration preset in the response feature excitation and fuzzy arbitrator is a set of excitation vectors whose parameter vectors are mathematically orthogonal to the parameter vectors of the optimal response strategy; the amplitude of the excitation vector is set such that when applied alone to the passive indoor distribution network, the absolute value of the real-time quality change of the 5G signal or BeiDou signal caused by it is lower than a perturbation tolerance determined based on the background noise level of the real-time quality.

[0016] Preferably, the system also includes a knowledge base confidence dynamic modeling engine, which is configured to: establish a dynamically updatable confidence score for each set of event fingerprints and the association mapping with the optimal response strategy stored in the knowledge base module; after each execution of the optimal response strategy by the strategy execution module, update the confidence score of the association mapping corresponding to the optimal response strategy based on the success rate of the event fingerprint matching corresponding to the execution, and a strategy effectiveness index calculated by the difference between the real-time quality of the 5G signal or BeiDou signal after execution and before execution; and when the confidence score of a certain set of association mappings is lower than a preset maintenance threshold, mark the set of association mappings as needing online optimization.

[0017] Preferably, the knowledge base confidence dynamic modeling engine is configured such that, when the intent inference module matches an event fingerprint corresponding to an association mapping marked as to be optimized online, it no longer directly executes the stored optimal response strategy, but triggers an online optimization process; the online optimization process includes: applying a series of controlled perturbation configurations near the parameters of the stored optimal response strategy, and searching for and determining a new optimal response strategy based on the real-time quality feedback of 5G signals or BeiDou signals, in order to update the knowledge base module.

[0018] Preferably, the system also includes an event entropy increase monitoring and prototype clustering engine, which is configured to: determine that event entropy increase has occurred when the intent inference module fails to successfully match the collected channel state information with any event fingerprint in the knowledge base module more than a preset matching failure number threshold within a preset time window; after determining that event entropy increase has occurred, collect and use an unsupervised clustering algorithm to perform online clustering on the continuously collected unmatched channel state information; and when the number of data points and the data point density of a newly formed cluster both exceed their respective preset stability condition thresholds, generate a new concept prototype representing a completely new type of indoor environmental event based on the cluster and output it.

[0019] Preferably, the core of the intent inference module is a fingerprint recognition model based on a specific computational model, which is either a lightweight convolutional neural network model or a pattern matching algorithm model based on dynamic time warping.

[0020] Preferably, the wireless bypass signal collected by the bypass channel monitoring module is a Wi-Fi signal, and the channel status information is Wi-Fi channel status information.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] 1. By collecting environmental information from non-primary channels through the bypass channel monitoring module, the causes of preset indoor environmental events are identified. The intention inference module, based on preset association mappings in the knowledge base, preemptively configures the optimal response strategy for the passive indoor distribution network before the primary channel signal quality is substantially affected. This operating mechanism shifts the trigger point of the system response from the result of signal quality deterioration to the cause of environmental events. This transforms the system's adaptation to environmental changes from delayed passive compensation to predictive adjustment based on event identification, thereby avoiding the inherent time delays in signal monitoring, analysis, and feedback control in existing technologies.

[0023] 2. By introducing response feature excitation and a fuzzy arbitrator, the system is endowed with a decision-making capability that avoids the fuzziness of passive perception. When the intention inference module is faced with multiple candidate events with similar characteristics and cannot make a unique judgment, the arbitrator does not seek to introduce new sources of perception information. Instead, it controls the strategy execution module to apply a perturbative excitation configuration to the passive indoor distribution network and simultaneously collects the transient response characteristics of the bypass channel caused by the excitation. Since different physical event entities have stable differences in their responses to the same electromagnetic perturbation, the system can make a high-confidence judgment based on the response characteristics. This mechanism of probing and reusing the execution end adds an active verification dimension to the decision-making process without changing the physical structure of the system, in order to deal with the inherent ambiguity of passive monitoring information.

[0024] 3. This invention utilizes a dynamic modeling engine for knowledge base confidence, enabling the system to perform online auditing and closed-loop correction of its own knowledge validity. After the strategy execution module completes a predictive adjustment, the system updates the confidence score of the corresponding association mapping in the knowledge base by combining the event fingerprint matching degree corresponding to the adjustment with the actual improvement of 5G or BeiDou signals. When the confidence score of a certain association mapping falls below a preset threshold due to its continuous decline in execution effect, the system can automatically trigger an online optimization process based on actual signal feedback to correct the strategy the next time the corresponding event is identified. This mechanism allows the system's knowledge base to adapt to gradual changes in the environment, avoiding the failure of static knowledge due to conceptual drift caused by the physical environment. The problem is that by setting up a causal spatiotemporal consistency arbitrator, a physical logic-based input information verification checkpoint is established at the front end of the core intent deduction module. Before performing event matching on the channel information collected by any monitoring node, the arbitrator first verifies whether the temporal characteristics of the information conform to the pre-set spatiotemporal logical relationship in the causal path graph of its propagation among multiple monitoring nodes. Only when a candidate event fingerprint shows a coherent and physically consistent propagation chain is it confirmed as a valid event and passed to the downstream for processing. This verification mechanism can distinguish between global channel changes caused by real physical events and local channel fluctuations caused by local electromagnetic interference based on the rationality of the source, thereby ensuring the authenticity of the system's decision-making basis. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the core workflow of the adaptive system based on cause prediction in this invention.

[0026] Figure 2 This is a comparison chart of the RSRP performance of predictive adjustment and passive feedback adjustment in this invention;

[0027] Figure 3 This is a timing diagram of the online adaptive optimization and closed-loop correction of the knowledge base in this invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the embodiments of this invention will be clearly and completely described below. It should be noted that the described embodiments are only some embodiments of this invention, not all embodiments. 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.

[0029] This invention provides a passive indoor electromagnetic environment adaptive system integrating 5G and BeiDou. This system aims to address the response lag problem caused by passive monitoring and feedback adjustment of the primary communication signal quality in highly dynamic indoor electromagnetic environments. The core mechanism of the system lies in shifting the trigger for adaptive adjustment from the result of communication quality deterioration to the physical events causing environmental changes. This constructs a predictive adjustment data processing closed loop based on event recognition and intent inference. The system mainly consists of a bypass channel monitoring module, a knowledge base module, and a causal spatiotemporal consistency mechanism. The system comprises a trimmer, an intent deduction module, and a strategy execution module. Working collaboratively, they transform the regression prediction problem of complex physical fields into a pattern recognition problem of environmental events. In a typical smart warehouse application scenario, the movement of an unmanned guided vehicle (AGV) generates repeatable, identifiable disturbances to the deployed Wi-Fi signals indoors. These disturbances occur much earlier than the substantial deep fading of 5G or BeiDou signals. This system utilizes this bypass channel information as a sentinel probe to predictively adjust the environmental risks of the primary channel. The bypass channel monitoring module's task is to construct a ring... In existing technologies, directly increasing the number of monitoring points or sampling frequencies for 5G or BeiDou signals to shorten response latency in the environmental change perception layer increases computational and hardware costs without altering the post-compensation mechanism. To circumvent this issue, the bypass channel monitoring module of this invention is configured to utilize existing standard Wi-Fi access points (APs) covering the same indoor space and collect Channel State Information (CSI) from these monitoring nodes. Specifically, CSI is a complex matrix describing the frequency domain response of a channel in an OFDM communication system, recording the amplitude and phase changes experienced by the signal along its propagation path. It can reflect multipath effects caused by physical events such as object movement and equipment startup / shutdown in the indoor environment with high sensitivity. This module configures the AP firmware to output CSI data streams at a sampling rate of at least 100Hz, providing high-temporal-resolution raw data input for subsequent event fingerprinting. By utilizing the deployed Wi-Fi infrastructure, this module provides the system with an information source capable of capturing dynamic environmental causes.

[0030] The knowledge base module, serving as the system's decision-making hub, aims to establish a deterministic correlation between causal events and optimal countermeasures. In a dynamic environment, simply collecting CSI data streams is insufficient; the key challenge lies in interpreting this data and linking it to the configuration parameters of a passive indoor distribution network capable of handling electromagnetic environment changes caused by the event. Therefore, the knowledge base module is configured to store two core data structures: one is multiple sets of event fingerprints corresponding to preset indoor environmental events, such as an AGV traveling along line 1 or the metal door of warehouse 1 opening. Each event fingerprint is a unique... The knowledge base module contains two main components: first, standardized multi-node CSI data sequences with time-varying fluctuation patterns; second, optimal response strategies for passive indoor distributed networks associated with each event fingerprinter, such as a set of phase shift matrix parameters that enable specific subarrays of reconfigurable smart metasurfaces (RIS) to generate specific beam patterns. In addition to the aforementioned mapping, the knowledge base module also stores a causal path graph. This graph uses multiple indoor Wi-Fi monitoring nodes as topology nodes and records the event fingerprint corresponding to each preset indoor environmental event, along with the preset time delay sequentially collected from two adjacent monitoring nodes along a specific propagation path. The knowledge base, which includes the signal strength evolution relationship at these two nodes, can be constructed through an association learning phase in the early stages of system deployment. During this phase, the system automatically records its side-channel fingerprints and identifies the corresponding optimal strategies while experiencing typical environmental events, thereby transforming the online decision-making problem into a lookup operation that can be implemented with lightweight computation.

[0031] Regarding the signal strength evolution relationship recorded in the causal path graph, the specific quantization calibration and online arbitration procedure is as follows: In the initial calibration phase of system deployment, a preset indoor environmental event, such as an unmanned navigation vehicle traveling along a preset path, is allowed to occur N times (N not less than 10). During this period, the bypass channel monitoring module records the characteristic peak amplitude of the event fingerprint at each event occurrence between two adjacent monitoring nodes. and The detected value is denoted as and Where i is the number of repetitions, from 1 to N, an offline processing program then calculates the amplitude ratio of these N repetitions. Statistical average with standard deviation The two calculation results are stored in the causal path graph as parameters of the signal strength evolution relationship of the event on that path segment. After the system enters online operation, when the causal spatiotemporal consistency arbiter detects a new candidate event, it calculates its position in real time. and Amplitude ratio at nodes and only if the Meet the conditions Only then is it determined that it conforms to the calibrated signal strength evolution relationship. The causal spatiotemporal consistency arbiter's function is to establish a physical logic-based input information verification checkpoint at the front end of the intent deduction module to solve the risk of CSI fingerprint spoofing caused by local electromagnetic interference. In a complex industrial environment, the instantaneous start and stop of a high-power motor may generate a signal fluctuation at a single monitoring node that is similar in waveform to the CSI fingerprint of a real physical event (such as the passage of an AGV). If this signal is transmitted downstream for processing without discrimination, it will cause the system to perform an incorrect prediction adjustment. To address this challenge, the causal spatiotemporal consistency arbiter in this solution is configured to first verify the rationality of the source of the channel state information collected by any monitoring node based on the causal path graph stored in the knowledge base module before processing it. The specific procedure is as follows: when the monitoring node When a candidate event fingerprint is initially identified at time t, the arbitrator does not immediately confirm it, but instead performs a two-way verification, that is, backtracks to verify the upstream nodes on the pre-defined propagation path of the event. Is it in ( The corresponding preceding part of the fingerprint has been observed near the time of the given time, and the downstream nodes of the path are prospectively verified. Is it about to ( When the subsequent part of the event fingerprint is observed near the time of the event, the channel state information is confirmed as a valid event fingerprint and is passed to the intent inference module only when a candidate event fingerprint shows a coherent propagation chain that conforms to the spatiotemporal logical relationship defined by the causal path graph on multiple monitoring nodes. Any isolated CSI fluctuation that does not meet this propagation characteristic, no matter how similar it is to the fingerprint in the database, will be judged as local interference and directly filtered out.

[0032] The intent deduction module's core function is to receive valid event fingerprints confirmed by a causal spatiotemporal consistency arbiter and efficiently and accurately identify them as specific indoor environmental events defined in the knowledge base. Considering the engineering requirement for the system to operate with low power on edge computing devices, the core fingerprint recognition model of this module is designed as either a lightweight convolutional neural network model or a pattern matching algorithm model based on Dynamic Time Warping (DTW). Taking the DTW algorithm model as an example, it measures the similarity between two time series by calculating the shortest curved path distance between the input CSI sequence and each standard event fingerprint sequence stored in the knowledge base. This algorithm allows for non-linear scaling of the time series on the time axis, thus effectively handling potential velocity differences that may exist when the same physical event occurs at different times. After receiving a valid event fingerprint, the module performs real-time matching calculations with all event fingerprints stored in the knowledge base and outputs one or more candidate indoor environmental event recognition results with the highest confidence. This process transforms the physically verified high-dimensional CSI data into data with clear physical characteristics. Meaningful environmental semantic information provides clear decision input for subsequent strategy execution. The strategy execution module, as the final execution end of the system, is responsible for converting the recognition results of the intent inference module into the physical configuration of the passive indoor distribution network to achieve predictive adjustment of the electromagnetic environment. When the module receives the event recognition result that the AGV is about to enter the No. 2 lane from the intent inference module, it will immediately extract the optimal response strategy uniquely corresponding to the event from the knowledge base module. For example, the strategy may contain a set of specific parameter instructions to control the phase of 128 units in the reconfigurable smart metasurface array at the lane entrance, so that it forms a beam that can focus the signal energy on a specific area inside the lane. The module then applies this set of optimal parameters to the passive indoor distribution network through the control interface before the AGV enters the lane and affects the 5G and Beidou signals. In this way, the system completes the response configuration in advance before the electromagnetic environment deteriorates, eliminating the time delay inherent in signal monitoring, analysis and feedback control in the traditional solution from the system, thereby ensuring the continuity of communication and positioning services in high-speed dynamic scenarios.To further enhance the system's robustness in complex environments, this invention may also include several optional enhanced functional modules. For example, a guardian module is configured to continuously monitor the real-time quality of 5G or BeiDou signals in the background. When the predictive adjustment of the intent inference engine is executed, and the real-time signal quality value is lower than a preset quality threshold determined based on historical statistical data (e.g., the Reference Signal Received Power (RSRP) is lower than -85dBm), or the real-time quality value is not higher than the value before execution, the guardian module will be triggered. It will automatically switch the system's control mode to a passive feedback adjustment mode based on real-time signal quality and simultaneously issue an alarm to the operation and maintenance system, indicating that a new, unknown environmental event not included in the knowledge base may have occurred, requiring learning and calibration. This mechanism provides a basic safety net for the system's predictive function, ensuring that the system's basic communication service quality is guaranteed under specific circumstances.

[0033] Furthermore, to address the decision ambiguity problem where the intention inference module cannot make a unique judgment when faced with multiple candidate events with similar characteristics, the system may also include a response feature excitation and fuzzy arbitrator. Its core mechanism lies in the fact that when the confidence levels of two candidate indoor environmental events... and All are above the confidence threshold (e.g., 0.8), and the absolute value of the difference between the two is less than a certain value based on... The 5% fuzzy judgment threshold is determined. When the preset fuzzy conditions are met. When the arbitrator is activated, it does not seek new perceived information but instead controls the strategy execution module to apply a preset, transient, perturbative excitation configuration to the passive indoor distribution network. The generation and selection procedure of the response feature excitation and the preset transient perturbative excitation configuration in the fuzzy arbitrator is as follows: for any pair of known indoor environmental events in the knowledge base that will produce similar event fingerprints... and (and its corresponding optimal response strategy) and In the offline calibration phase, the difference vector of its response strategy is first calculated. And generate a set of M elements (M not less than 20) and The set of candidate vectors composed of mathematically orthogonal unit excitation vectors. (where j is the candidate vector number, from 1 to M); subsequently, the system enters the stimulus testing process, in which events... and The process is repeated K times (K is not less than 10), and during each execution, each vector in the candidate set is processed sequentially. (The amplitude of the excitation is uniformly adjusted to meet the perturbation tolerance) and applied to the passive indoor distribution network. At the same time, the transient response characteristic energy integral value induced by this excitation is collected and calculated through the bypass channel monitoring module, thereby obtaining two sets of response datasets. and Ultimately, the system assigns each candidate vector... Calculate a separability index It is defined as the ratio of the square of the difference between the means of two sets of response data to the sum of the variances of the two sets of data, and a value is selected that makes this index... Candidate vectors that achieve the maximum value The instantaneous perturbation excitation configuration, ultimately used for online fuzzy arbitration, is fixed in the knowledge base. This excitation configuration is an excitation vector whose parameter vector is mathematically approximately orthogonal to the parameter vectors of the optimal response strategies for the two candidate events. Its amplitude is set to the absolute value of the real-time quality change of the 5G or BeiDou signal caused when applied alone, which is lower than a perturbation tolerance (e.g., 0.5dB) determined based on the signal background noise level. Simultaneously with the application of the excitation, the arbitrator rapidly acquires the transient response characteristics of the bypass channel state information triggered by the excitation through the bypass channel monitoring module. Since different physical event entities, such as the large metal surface of an AGV and the sparse metal frame of a cargo cart, have different scattering and reflection characteristics for the same electromagnetic perturbation, the system can determine the unique indoor environmental event from multiple candidate events that meet the fuzzy conditions based on the captured response characteristics through a pre-trained Support Vector Machine (SVM) classifier. This mechanism of probing and reusing the execution end adds an active verification dimension to the decision-making process.

[0034] To enable the system to perform online auditing and closed-loop correction of its own knowledge validity, addressing the concept drift problem caused by gradual environmental changes, the system may also include a dynamic modeling engine for knowledge base confidence. This engine establishes a dynamically updatable confidence score for each set of event fingerprints → optimal response strategies stored in the knowledge base. After each predictive adjustment by the strategy execution module, the engine updates the confidence score of the association mapping by combining the success rate of the event fingerprint matching corresponding to that execution with a strategy validity index calculated from the difference between the real-time quality of 5G or BeiDou signals before and after execution. When a certain set of associations... When the confidence level of a mapping falls below a preset maintenance threshold due to its continuously declining performance (e.g., when the effectiveness index is negative after five consecutive executions), the mapping set will be marked for online optimization. In the next instance where the intent deduction module matches an event fingerprint corresponding to the marked mapping, the system will no longer directly execute the stored old strategy. Instead, it will automatically trigger an online optimization process. This process involves applying a series of controlled perturbations near the stored optimal response strategy parameters and, based on real-time quality feedback from 5G or BeiDou signals, using gradient descent to search for and determine a new optimal response strategy to update the knowledge base module. This mechanism enables the system to... The system's knowledge base can adapt to gradual environmental changes, avoiding the problem of static knowledge becoming obsolete. Finally, to enhance the system's ability to autonomously discover and classify unknown environmental events, the system can also include an event entropy increase monitoring and prototype clustering engine. This engine is configured such that when the intent inference module fails to successfully match the collected channel state information with any event fingerprint in the knowledge base more than a preset matching failure threshold (e.g., 50 times) within a preset time window (e.g., 10 minutes), the system determines that event entropy increase has occurred, i.e., the unpredictability of the environment has increased. After determining that event entropy increase has occurred, the engine will be activated and begin collecting and using... A density-based streaming unsupervised clustering algorithm performs online clustering on continuously collected CSI fingerprint data that has not been successfully matched. This algorithm does not rely on any pre-labeling, but rather classifies data points autonomously based on the similarity of data points in the feature space. When the number of data points in a newly formed cluster exceeds the stability threshold (e.g., more than 100 points) and its data point density also exceeds a preset threshold, the engine determines that a stable and repeatable new type of indoor environmental event has been discovered. Based on this cluster, a centroid is calculated as a representative, a new concept prototype is generated, and it is output to the operation and maintenance management interface to prompt manual calibration of the corresponding optimal response strategy.

[0035] Example 1: This example demonstrates the specific operation of the technical solution in a particular industrial application scenario. In a continuously operating fully automated warehouse, a sorting robot relying on 5G communication for task scheduling and BeiDou signals for centimeter-level precise positioning is deployed in the densely packed shelving area. When an unmanned guided vehicle (AGV) fully loaded with metal-shielded goods is about to enter a narrow aisle (Aisle 2) consisting of 10-meter-high all-metal shelving, this routine operation poses a challenge to the indoor electromagnetic environment. The AGV's entry triggers a severe and instantaneous increase in multipath effects within the aisle, causing the 5G signal reference received power (RSRP) received by the sorting robot to drop by more than 20 dB within hundreds of milliseconds. Simultaneously, the carrier-to-noise ratio of the BeiDou positioning signal deteriorates, resulting in meter-level positioning jumps. For systems relying on passive feedback adjustment, this can only be addressed when the system detects... Compensation is only initiated after signal quality has deteriorated, and this inherent response delay is sufficient to cause the sorting robot to stop working due to communication interruption or collide with the shelf due to positioning errors. In the same scenario, the system of this invention operates as follows: when the AGV is 15 meters away from the entrance of lane 2 on the main aisle, its movement has begun to disturb the Channel State Information (CSI) of multiple Wi-Fi access points (APs) along the way. The bypass channel monitoring module in the system continuously collects CSI data streams from multiple monitoring nodes such as AP1, AP2, and AP3 at a frequency of 200Hz. At this time, the system does not monitor any quality indicators of 5G or Beidou signals because they have not yet been affected. As the AGV approaches, the causal spatiotemporal consistency arbiter in the system detects a CSI fluctuation pattern that first appears on AP1, and then, with a time delay matching its physical position and AGV speed. Subsequently, a similar evolution pattern was observed on AP2. Based on the causal path graph stored in the knowledge base module, the arbitrator confirmed that this was a propagation chain caused by a single, continuously moving physical object, consistent in spatiotemporal logic, rather than multiple isolated local electromagnetic interferences. Therefore, this multi-node CSI sequence was identified as a valid event fingerprint and transmitted to the intent deduction module. Upon receiving the valid event fingerprint, the intent deduction module immediately activated its internal pattern matching algorithm based on dynamic time warping, comparing the fingerprint sequence with the pre-stored event fingerprints in the knowledge base module, and matching the event fingerprint numbered E with a confidence level higher than 98%. The event fingerprint of E027 corresponds to the environmental semantics that a standard load AGV is about to enter lane 2. The output of this identification result indicates that the system has transformed a complex physical problem, namely, predicting how the AGV's entry into the lane will affect the electromagnetic field, into a more computationally simpler pattern recognition problem, namely, identifying the causal event itself. Subsequently, based on the identification result, the strategy execution module extracts the optimal response strategy associated with the E027 event fingerprint machine and numbered S027 from the knowledge base module. This strategy includes a set of phase shift matrix parameters for configuring the reconfigurable intelligent metasurface RIS array at the entrance of lane 2.

[0036] Approximately 500 milliseconds before the AGV's front end reaches the aisle entrance (about 2 meters away), and before it affects the 5G and BeiDou signals within the aisle, the strategy execution module applies the S027 strategy to the RIS array via the control bus. This instantaneously reshapes the array's beam pattern, creating a stable signal focus capable of diffracting into the aisle and covering the sorting robot's work area. Therefore, when the AGV finally enters the aisle, the RSRP fluctuation of the 5G signal received by the sorting robot is suppressed to within 3dB, the carrier-to-noise ratio of the BeiDou positioning signal remains stable without any positioning jumps, and the sorting robot's tasks can be executed continuously and uninterruptedly. The original potential collision risk is avoided. This process... The key lies in the fact that the causal spatiotemporal consistency arbiter provides highly realistic decision input to the intent inference module, which in turn shifts the trigger point for adaptive adjustment from a passive response to signal degradation to the pre-identification of the causes of environmental changes. The stable operation of this system in this scenario changes the technical approach to dealing with dynamic environments. It no longer attempts to catch up with a complex physical field change that has already occurred through higher-frequency monitoring, but instead deploys response strategies in advance by calculating and identifying the fingerprints of causal events in the bypass channel. The system's focus shifts from physical inverse solving in the field of signal processing to pattern recognition and predictive decision-making based on specific computational models.

[0037] Example 2: To objectively verify the effectiveness of the technical solution of this invention in dealing with sudden, high-speed changes in the indoor electromagnetic environment, this example designed and executed a comparative experiment. The purpose of the experiment was to quantitatively compare the predictive adjustment method of this invention with the passive feedback adjustment method of the prior art, and to assess the ability to maintain the quality of the communication link under the same dynamic interference event. The experiment was conducted in a 20m × 10m × 5m electromagnetic shielding room to simulate the warehouse environment in Example 1. A 5G micro base station and signal receiving terminal were deployed at one end of the shielding room to form the main communication link; a 15m long electromagnetic shielding link was laid along one side of the shielding room. A linear guide rail was used, along which a remotely controlled mobile vehicle equipped with a 1.5m x 1m metal plate moved at a speed of 2m / s to simulate the dynamic obstruction of the signal propagation path by an unmanned navigation vehicle. Three standard Wi-Fi wireless access points were deployed at equal intervals along the guide rail path as monitoring nodes for the bypass channel monitoring module. At the midpoint of the propagation path of the primary communication link, a reconfigurable intelligent metasurface RIS consisting of 64 units was set up as the control object of the passive indoor distribution network. The experiment was conducted in two groups: a control group using existing technology and an experimental group using the technical solution of this invention. The physical environment of the two groups of experiments was... The interference events and core hardware configurations were identical. The control group's system was configured in passive feedback adjustment mode, continuously monitoring the Reference Signal Received Power (RSRP) of the 5G signal receiving terminal. When the RSRP value fell below the preset -90dBm quality threshold due to vehicle obstruction, the RIS control system was triggered to perform a scan and optimization to find a new phase shift configuration to compensate for signal fading. The experimental group adopted the predictive adjustment mode of this invention, which did not monitor the RSRP value of the 5G signal, but instead used a bypass channel monitoring module to collect CSI data of the Wi-Fi signal and used an intent inference module to analyze the environment of vehicle movement. Event identification is performed. After identifying the event fingerprint that the vehicle is about to block the main communication link, the optimal response strategy is extracted and executed from the knowledge base module in advance, and the RIS is configured. During the experiment, a spectrum analyzer is used to record the dynamic changes of the RSRP value of the 5G signal receiving terminal of the control group and the experimental group at a time resolution of 10 milliseconds during the continuous vehicle crossing. After averaging the data of multiple measurements of each group, the performance of the two systems in dealing with the same interference event is obtained. The specific experimental data are shown in Table 1, which compares the received signal quality under the two modes.

[0038] Table 1: Comparison of received signal quality under two modes.

[0039]

[0040] Analysis of the data in Table 1 shows that at the critical time point when the vehicle enters the obstructed area, namely between 2000 and 2500 milliseconds, the RSRP value of the control group dropped sharply, reaching a minimum of -99.8 dBm. This signal strength is below the threshold for most communication protocols to function normally. Furthermore, after the vehicle leaves the obstructed area, it still takes approximately 1000 milliseconds to recover to a stable level. In contrast, the experimental group, by identifying the CSI event fingerprint at 1750 milliseconds (250 milliseconds before the vehicle affected the signal), executed the optimal response strategy of RIS in advance. The RSRP value of the system was only -84.5dBm at its lowest point during the entire occlusion event, remaining above the high-quality communication threshold throughout, with a fluctuation range of less than 2dB. This experimental result shows that the system of the present invention, by utilizing bypass channel information for event identification and intent inference, can complete predictive adjustments before the quality of the primary communication link is affected, thereby effectively avoiding deep signal fading caused by high-speed dynamic occlusion. Compared with existing technologies that rely on lag compensation for signal degradation, its performance in ensuring service continuity in high-speed dynamic scenarios has been improved.

[0041] To further verify the key role of the causal spatiotemporal consistency arbiter in ensuring the authenticity of predictive decisions in the present invention, the following comparative examples are set up.

[0042] Comparative Example 1: This comparative example aims to verify the actual performance of a simplified predictive system in the absence of a causal spatiotemporal consistency arbiter. The experimental system used in this comparative example has the same hardware configuration, knowledge base module, intent inference module, and strategy execution module as the experimental group in the aforementioned embodiments. The essential difference is that the causal spatiotemporal consistency arbiter is removed from the system structure of this comparative example. The channel state information (CSI) data streams from each monitoring node of the bypass channel monitoring module are directly input to the intent inference module for event fingerprint matching without any spatiotemporal logical relationship verification. The experimental environment is the same as the automated warehouse center scenario in the aforementioned embodiments. In this scenario, in addition to the unmanned guided vehicle (AGV), a six-axis articulated industrial robot is deployed near the monitoring node AP3 to perform palletizing operations. The knowledge base module has pre-stored the event fingerprint numbered E027, whose corresponding environmental semantics are that a standard load AGV is about to enter lane 2, and the optimal response strategy S027 associated with this event.

[0043] During a typical work cycle, when no AGV was scheduled to enter lane 2, the industrial robot performed a large-scale, high-torque rotary grasping motion. The motor of this motion started instantaneously, generating a localized, strong electromagnetic interference (EMI) at monitoring node AP3. The waveform characteristics of the CSI data caused by this interference in a single time series showed approximately 87% similarity to the leading part of the fingerprint of event E027 stored in the knowledge base. Due to the lack of a verification step by a causal spatiotemporal consistency arbiter, the intention inference module immediately identified this CSI data from AP3 as a valid trigger of event E027. Subsequently, based on this erroneous identification result, the strategy execution module extracted and executed the optimal response strategy for event S027 from the knowledge base module, performing a predictive configuration of the reconfigurable intelligent metasurface (RIS) array at the entrance of lane 2. In fact, no AGV approached or entered lane 2 within 5 minutes before and after this time point. This predictive adjustment was a false alarm caused by local interference. The specific system behavior and signal quality changes are recorded in Table 2.

[0044] Table 2: Record of false alarm events caused by local interference in Comparative Example 1.

[0045]

[0046] Analysis of the data in Table 2 shows that at T0+1.5 seconds, the system in this comparative example, unable to verify the physical source rationality of CSI information originating from a single monitoring node, misjudged passive local electromagnetic interference as a real AGV movement event and performed a completely unnecessary RIS network reconstruction. This erroneous predictive adjustment not only failed to bring any gain, but also caused a more than 3dB instantaneous degradation of the 5G signal RSRP in Lane 2 due to the alteration of the previously stable signal field distribution, resulting in unnecessary communication link disturbances to other equipment operating in the area. The experimental results indicate that if the signal fluctuations of the bypass channel are directly used for event matching without setting an arbitration mechanism such as the one proposed in this invention, which performs spatiotemporal consistency verification based on multi-node causal path graphs, the system will be unable to effectively distinguish between real global physical events and isolated local electromagnetic interference, leading to frequent decision-making errors and reducing the reliability and practicality of the system.

[0047] Example 3: This example combines Figures 1 to 3 A description of the 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system, such as... Figure 1As shown, the bypass channel monitoring module is responsible for collecting non-primary Wi-Fi channel status information (CSI) from the indoor environment. Before being processed, this information is first verified by the causal spatiotemporal consistency arbiter based on the causal path graph provided in the knowledge base module to confirm that it is a valid event fingerprint. Subsequently, the intent deduction module receives the valid event fingerprint and matches it with the event fingerprint database stored in the knowledge base module to identify specific physical cause events such as AGV movement. Finally, the policy execution module extracts and executes the optimal response policy from the knowledge base module based on the identification result, and performs predictive configuration for passive indoor distribution networks such as RIS, thereby shifting the trigger point of the system response from the result of the physical event to the cause level.

[0048] like Figure 2 As shown, this figure compares the received signal quality of the experimental group using the predictive adjustment method of this invention with that of the control group using the passive feedback adjustment method of the prior art when dealing with the same dynamic interference event. The horizontal axis of the figure represents time in ms, and the vertical axis represents the reference signal received power (RSRP) in dBm. The curves in the figure show that the RSRP value of the control group deteriorated sharply at 2000ms due to the event, reaching a low of nearly -100dBm. In contrast, the experimental group, because it performed predictive adjustment in advance at 1750ms, maintained an RSRP value of only -84.5dBm at its lowest point during the entire obstruction event, remaining above the high-quality communication threshold throughout, with stable fluctuations and no deep fading. Figure 3 As shown, the process begins when the strategy execution module reports the execution result to the knowledge base confidence dynamic modeling engine after completing a strategy execution. The engine combines the quality difference before and after execution obtained from 5G or BeiDou signals to calculate the strategy effectiveness index and update the confidence score of the corresponding associated mapping in the knowledge base module. If the score is lower than the preset maintenance threshold, the mapping is marked as needing online optimization. When the next intent inference module identifies the event to be optimized, the engine will trigger an online optimization process, that is, instruct the strategy execution module to apply a series of perturbation configurations and search for and determine a new optimal response strategy based on the collected real-time quality feedback, in order to update the knowledge base module and reset the confidence score, thereby completing a closed-loop adaptive correction of the knowledge base.

[0049] Example 4: This example aims to provide a detailed and reproducible engineering explanation of the construction of the internal knowledge base module and the calibration process of key operating parameters after the technical solution of the present invention is initially deployed in a completely new environment. In a specific application scenario, the system is deployed in a flexible manufacturing workshop where data collection has not been performed before. At this time, its knowledge base module is in an uninitialized state and does not store any association mapping between event fingerprints and optimal response strategies. In order for the system to enter a predictive adjustment mode, an association learning calibration procedure must first be executed. This procedure aims to establish a deterministic mapping relationship between recurring physical events in the environment, their CSI fluctuation patterns generated on the bypass channel, and a passive indoor distribution network configuration that can optimally respond to the event. The first step of this calibration procedure is the collection of event fingerprints and the generation of causal path maps. The calibration engineer selects a typical environmental event in the workshop, namely a material handling robot moving along the ABC path, and instructs the robot to operate under no-load conditions. Under these conditions, the robot repeatedly moves along the same path 10 times at its standard operating speed of 1.5 m / s. During this process, the bypass channel monitoring module deployed in the workshop, at a sampling rate of 200 Hz, completely records all CSI data streams output by the three Wi-Fi monitoring nodes AP-A, AP-B, and AP-C along the route during each movement. After the data acquisition is completed, an offline processing script is started. This script first uses the CSI data from the 5th run as a reference and employs a dynamic time warping algorithm to align the CSI data from the remaining 9 runs with this reference on the time axis to eliminate minor time differences during robot start-up and shutdown. After alignment, the script calculates the average amplitude and phase values ​​of these 10 CSI sequences at each sampling point, thereby generating a standardized CSI event fingerprint with a high signal-to-noise ratio and storing it in the knowledge base module. At the same time, the script also records the average timestamps of the fingerprint feature peaks appearing on AP-A, AP-B, and AP-C and calculates their propagation time delay between nodes. This information is then embedded in the causal path map.

[0050] The second step in the calibration procedure is to optimize the optimal response strategy and determine the key thresholds. For the material handling robot movement events for which fingerprint modeling has been completed in the previous step, the engineer places a fixed signal quality monitoring terminal in the center of a 5G weak signal coverage area in the BC segment of the path. Subsequently, the robot is instructed to run repeatedly on the path. However, during this run, the RIS array located near this area will enter parameter scanning mode. Specifically, the RIS control system will traverse the phase shift parameters of its 64 surface cells according to a preset codebook. Each traversal occurs when the robot passes through the key area. The scans were completed within the same time window, and the monitoring terminal recorded the minimum and average values ​​of the 5G signal RSRP under each scan configuration. After completing 32 scans with different coding configurations, the system compared all the recorded data and found that when the 19th coding book configuration was used, the average RSRP of the monitoring terminal was the highest, at -82.5dBm, and the minimum value remained at -86.0dBm, with no deep fading. Therefore, the system stored the phase shift matrix parameters corresponding to this 19th coding book as the optimal response strategy in the knowledge base module and established an association mapping with the previously collected CSI event fingerprints.

[0051] During this process, the system also uses the collected data to calibrate the preset quality threshold of the Guardian module. During the parameter scanning process, the system recorded that the lowest RSRP value caused by robot movement was -103.5 dBm when no RIS compensation was applied (i.e., the RIS was inactive). After applying the optimal response strategy, the lowest RSRP value was -86.0 dBm. To achieve a balance between ensuring system robustness and avoiding false alarms, one method for setting this threshold is to place it slightly below the midpoint between the worst and best cases. Based on this, the threshold can be set to -95 dBm. This value is significantly higher than the signal quality without compensation, effectively monitoring... While detecting situations where predictive adjustments fail, sufficient margins are maintained for optimal performance, avoiding false triggers caused by normal signal fluctuations. By repeatedly executing the above steps, all typical environmental events requiring predictive responses in the workshop are calibrated one by one. The system's knowledge base module is then fully constructed, and each key threshold is set based on actual field measurement data. This transforms the entire system from an uninitialized state into an operational system with the ability to recognize and predict specific environments. This calibration procedure transforms a deployment process into a standardized process with specific inputs, execution steps, and quantifiable outputs that can be performed by engineering technicians.

[0052] Example 5: This example aims to illustrate the autonomous learning and adaptation mechanism of the technical solution of the present invention when facing new and unknown environmental events not included in the knowledge base. In a production workshop that has completed initial calibration, due to production line adjustments, a newly added six-axis robotic arm begins to perform periodic grasping and rotation operations. Its movement generates stable and repeatable disturbances to the local electromagnetic environment that have not been recorded in the knowledge base. When the robotic arm runs for the first time, the bypass channel monitoring module collects the CSI fluctuations generated by its movement. However, when the intention inference module compares it with all existing CSI event fingerprints in the knowledge base, it cannot find any entries with similarity exceeding the preset matching threshold, resulting in continuous matching failures. When this matching failure state accumulates to more than the matching failure count threshold within a set time window, such as 5 minutes, such as 30 times, the system determines that an event entropy increase has occurred and automatically activates the event entropy increase monitoring and prototype clustering engine.

[0053] Once activated, the engine begins collecting continuously acquired, unmatched CSI time-series data as input samples. Internally, the engine runs a density-based streaming unsupervised clustering algorithm. This algorithm does not rely on any pre-labeling but classifies the CSI samples online based on their similarity in a multi-dimensional feature space. As the robotic arm performs its periodic operations, CSI samples representing its motion characteristics are continuously input into the engine, gradually forming a high-density data cluster in the feature space. When the number of data points in this newly formed cluster exceeds a stable threshold, such as reaching 100 sample points, and its internal data point density also exceeds a certain threshold... When the preset density threshold is reached, the engine determines that a new and stable indoor environmental event type has been discovered. At this time, the engine calculates the centroid or representative sample of the cluster and encapsulates it into a data object labeled as New Concept Prototype-001. Finally, the New Concept Prototype, along with its statistical information such as the frequency of its occurrence, is output to the system's operation and maintenance management interface for engineering technicians to calibrate or determine the corresponding optimal response strategy by executing an optimization procedure as described in Example 3. Once the calibration is completed, the New Concept Prototype and its response strategy are formally incorporated into the knowledge base module, thereby enabling the system to complete the discovery of unknown environmental changes and the updating of the knowledge base.

[0054] Example 6: This example aims to illustrate the specific working procedure of the response feature excitation and fuzzy arbitrator in the technical solution of the present invention, in order to solve the decision fuzziness problem faced by the system when the bypass channel event fingerprint features of two different physical events are similar. In a calibrated warehouse environment, the knowledge base module has stored two independent indoor environmental events. One is an unmanned guided vehicle (AGV) traveling along path 3, denoted as event E1, and the other is a manually pushed metal material cart traveling along path 3, denoted as event E2. Since both are metal objects and their running paths overlap, the CSI event fingerprints generated at the bypass channel monitoring node are similar in time domain waveform. When an object moves along path 3, the output result of the intention inference module simultaneously satisfies the preset fuzzy condition, namely the confidence of the two candidate events. and The scores were 0.92 and 0.89 respectively, both higher than the confidence threshold. The value is 0.85, and the absolute value of the difference between the two, 0.03, is less than the fuzzy judgment threshold. When the value is 0.05, the system cannot make a unique judgment, so it activates the response feature excitation and fuzzy arbitrator.

[0055] Once activated, the arbitrator first extracts a preset instantaneous perturbation excitation configuration associated with event pairs E1 and E2 from its internally stored perturbation excitation library. This excitation configuration is a vector parameter, which is determined during the offline calibration phase by obtaining the optimal response strategy vectors for E1 and E2 respectively. and Calculate the difference vector between the two. And solve for a relation with Mathematically orthogonal unit excitation vectors The amplitude of the vector is set such that, when applied alone, the absolute value of the real-time quality change of the 5G signal caused by it is less than the perturbation tolerance of 0.5 dB. The arbitrator then controls the strategy execution module to apply the excitation vector to the passive indoor distribution network. The duration is 50 milliseconds. Simultaneously with the application of the excitation configuration, the bypass channel monitoring module collects the transient response characteristics of the bypass channel CSI amplitude caused by this excitation configuration at a sampling rate of 1 kHz. Its core processing step is to calculate the integral value of the square of the first derivative of the CSI amplitude within this 50-millisecond time window. This integral value characterizes the channel's response energy to the perturbation excitation. Due to the physical differences in the scattering characteristics of the same electromagnetic perturbation between the large-area continuous metal body of the AGV and the sparse metal frame structure of the material cart, the former will produce a more concentrated and stronger transient response, while the latter will be relatively weak. An arbitrator inside... The pre-trained, single-layer decision tree classifier makes the final decision based on the response energy integral value. In this example, the measured response energy integral is 1.38 units, which is greater than the classification threshold of 1.15 determined based on historical statistical data to distinguish between E1 and E2. Therefore, the arbitrator identifies the current indoor environmental event as E1, i.e., the unmanned guided vehicle (AGV) travels along path 3. After the arbitrator completes the determination of the unique event, its output will overwrite the previous fuzzy judgment of the intent inference module. Based on this determined E1 event identification result, the policy execution module extracts and executes the optimal response policy corresponding to E1 from the knowledge base module. This allows for a high-confidence predictive adjustment. This procedure introduces a brief active detection on the basis of passive perception, using the different response characteristics of different physical entities to the same stimulus to eliminate the uncertainty of perception. Without adding new hardware, it adds an active verification dimension to the system's computational model.

[0056] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A passive indoor electromagnetic environment adaptive system integrating 5G and BeiDou, characterized in that, The system includes: The bypass channel monitoring module is configured to collect channel status information from multiple wireless bypass signal monitoring nodes covering the same indoor space; The knowledge base module is configured to store event fingerprints corresponding to preset indoor environmental events, and a causal path graph. The preset indoor environmental events are physical events caused by object movement or equipment start-up and shutdown. The causal path graph records the spatiotemporal logical relationship of the event fingerprint corresponding to the preset indoor environmental event propagating among multiple wireless bypass signal monitoring nodes. The spatiotemporal logical relationship specifically refers to the preset time delay that the event fingerprint corresponding to the preset indoor environmental event is sequentially collected at two adjacent wireless bypass signal monitoring nodes along the preset propagation path, and the signal strength evolution relationship presented at the two wireless bypass signal monitoring nodes. Among them, the signal strength evolution relationship parameter is the statistical average and standard deviation of the amplitude ratio of the characteristic peak amplitude of the event fingerprint at two adjacent wireless bypass signal monitoring nodes. The causal spatiotemporal consistency arbiter is configured to: before processing the channel state information collected by the wireless bypass signal monitoring node, first verify whether the channel state information is consistent with the spatiotemporal logical relationship of a certain event recorded in the causal path graph, and only when the verification is consistent will the channel state information be confirmed as a valid event fingerprint. The intent deduction module is configured to match the valid event fingerprint with the event fingerprint stored in the knowledge base module only when a valid event fingerprint is received from the causal spatiotemporal consistency arbiter, so as to identify the preset indoor environmental event that is about to occur. The strategy execution module is configured to extract and execute the optimal response strategy from the knowledge base module based on the recognition results of the intent inference module.

2. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The causal path graph stored in the knowledge base module is a topology graph built with multiple wireless bypass signal monitoring nodes as nodes.

3. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The system also includes a guardian module, which is configured to monitor the real-time quality of 5G or BeiDou signals. When the real-time quality value is lower than a preset quality threshold or is not higher than the value before the execution of the optimal response strategy after the strategy execution module executes the optimal response strategy, the guardian module will be configured to switch the system's control mode to a control mode based on passive feedback adjustment of real-time quality.

4. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The system also includes a response feature excitation and a fuzzy arbitrator, which are configured such that: when the matching result of the intent inference module satisfies a preset fuzzy condition, i.e. , ,and At that time, the arbitration process is activated, in which and These represent the confidence levels of two candidate indoor environmental events. The confidence threshold is determined based on historical statistical data. The fuzzy judgment threshold is determined based on a preset percentage of the confidence threshold. After activation, the control strategy execution module applies a preset instantaneous perturbation excitation configuration to the passive indoor distribution network. While applying the excitation configuration, the transient response characteristics of the bypass channel state information caused by the excitation configuration are collected by the bypass channel monitoring module. Based on transient response characteristics, a unique indoor environmental event is determined from multiple candidate indoor environmental events that meet fuzzy conditions. The transient perturbation excitation configuration is a set of excitation vectors, and the parameter vectors of the excitation vectors are mathematically orthogonal to the parameter vectors of the optimal response strategy. The amplitude of the excitation vector is set such that when applied alone to a passive indoor distribution network, the absolute value of the real-time quality change of the 5G signal or BeiDou signal caused by it is lower than a perturbation tolerance determined based on the background noise level of the real-time quality.

5. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The system also includes a knowledge base confidence dynamic modeling engine, which is configured to: establish a dynamically updatable confidence score for each set of event fingerprints and the association mapping between the optimal response strategy stored in the knowledge base module; after each execution of the optimal response strategy in the strategy execution module, update the confidence score of the association mapping corresponding to the optimal response strategy based on the success rate of the event fingerprint matching corresponding to this execution, and a strategy effectiveness index calculated by the difference between the real-time quality of the 5G signal or BeiDou signal after execution and before execution. And when the confidence score of a certain group of associated maps is lower than the preset maintenance threshold, the group of associated maps is marked as needing online optimization.

6. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 5, characterized in that, The knowledge base confidence dynamic modeling engine is configured such that when the intent inference module matches the event fingerprint corresponding to the association mapping marked as to be optimized online in the next time, it will not directly execute the stored optimal response strategy, but will trigger an online optimization process. The online optimization process includes: applying a series of controlled perturbation configurations near the parameters of the stored optimal response strategy, and searching for and determining a new optimal response strategy based on the real-time quality feedback of the 5G signal or BeiDou signal.

7. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The system also includes an event entropy increase monitoring and prototype clustering engine. This engine is configured to detect when the intent deduction module fails to successfully match the collected channel state information with any event fingerprint in the knowledge base module more than a preset failure rate within a preset time window, indicating an event entropy increase. Upon detection of an event entropy increase, the engine is activated to collect continuously collected, unmatched CSI fingerprint data and employs a density-based streaming unsupervised clustering algorithm to perform online clustering of the collected data. This unsupervised clustering algorithm does not rely on any pre-labeling but rather on the data itself. The similarity of points in the feature space is used for autonomous classification. When the number of data points in a newly formed cluster exceeds the stability condition threshold and its data point density also exceeds the preset threshold, the event entropy increase monitoring and prototype clustering engine determines that a stable and repeatable new type of indoor environmental event has been discovered. Based on the clustering, a centroid is calculated as a representative, a new concept prototype is generated and output to the operation and maintenance management interface, prompting manual calibration of the corresponding optimal response strategy. When the number of data points and the data point density of a newly formed cluster both exceed their respective preset stability condition thresholds, a new concept prototype representing a new type of indoor environmental event is generated based on the cluster and output.

8. The 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The core of the intent inference module is a fingerprint recognition model based on a specific computational model. The fingerprint recognition model can be a lightweight convolutional neural network model or a pattern matching algorithm model based on dynamic time warping.

9. A 5G and BeiDou integrated passive indoor electromagnetic environment adaptive system according to claim 1, characterized in that, The wireless bypass signal collected by the bypass channel monitoring module is a Wi-Fi signal, and the channel status information is Wi-Fi channel status information.

Images