Intelligent canopy communication method based on 5G communication

Abstract

The present application relates to the field of information communication technology, and more particularly to a kind of intelligent canopy communication system and method based on 5G communication.The method comprises the following steps: multiple-source sensor array is deployed to canopy structure to collect environment, structure and personnel data, and data preprocessing is carried out, and heterogeneous original data set of canopy is constructed;The heterogeneous original data set of canopy is time-stamped alignment, and the synchronized canopy data set is obtained;Synchronization canopy data set is carried out heat matrix construction, and the dynamic heat matrix of canopy is obtained;Based on the dynamic heat matrix of canopy, principal component analysis is carried out, and the fusion feature matrix of canopy is obtained;Therefore, the closed-loop perception and control system based on multi-source fusion data is constructed, the problem of information isolation, response lag and control inaccuracy in traditional canopy management is solved, and the comprehensive perception ability and dynamic control efficiency of intelligent canopy in complex environment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information communication technology, and in particular to an intelligent tent communication system and method based on 5G communication. BACKGROUND

[0002] The existing tent communication system mostly relies on traditional wired control mode or low-power wireless local area network for equipment linkage and data transmission, which has problems such as limited bandwidth, high latency, and rigid network topology, and is difficult to adapt to the rapid deployment and flexible control requirements in high dynamic scenarios. Especially in emergency situations such as sudden environmental changes or large-scale crowd gathering, the existing system cannot realize low-latency collection and high-frequency synchronization of sensor data, resulting in a lag in response to core data such as structural stress, environmental indicators, and personnel trajectory. In addition, the existing control logic generally uses one-way instruction transmission and periodic feedback mechanism, lacks real-time modeling and iterative capability for feedback data, and cannot dynamically optimize the control strategy according to the actual execution result. At the same time, the current system has a coarse coding granularity for device control instructions, lacks multi-channel linkage expression of fine parameters in complex parameter scenarios (such as ventilation volume, lighting intensity, and temperature and humidity zoning), resulting in insufficient execution accuracy and environmental adaptability. More importantly, the traditional system generally lacks cross-domain information fusion capability, making it difficult to analyze and integrate decisions on three types of heterogeneous information: structure, personnel, and environment, resulting in a low overall intelligence level of the system and a lack of active prediction and adaptive control capability. In terms of data processing, the existing scheme mostly uses static threshold rules or simplified statistical models in the processes of feature extraction, noise filtering, and anomaly identification of sensor data, which cannot dynamically track the spatio-temporal linkage characteristics in complex scenarios and also cannot meet the rapid response requirements based on edge nodes. SUMMARY

[0003] Therefore, it is necessary to provide an intelligent tent communication system and method based on 5G communication to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, an intelligent tent communication method based on 5G communication, the method comprising the following steps:

[0005] Step S1: Deploying a multi-source sensor array on the tent structure to collect environmental, structural, and personnel data, and performing data preprocessing to construct a tent heterogeneous raw data set;

[0006] Step S2: Time stamp alignment of the tent heterogeneous raw data set to obtain a synchronized tent data set; heat matrix construction of the synchronized tent data set to obtain a tent dynamic heat matrix; principal component analysis based on the tent dynamic heat matrix to obtain a tent fusion feature matrix;

[0007] Step S3: Constructing a decision model using the canopy fusion feature matrix to obtain a canopy multi-objective decision model; outputting a device control parameter vector based on the canopy multi-objective decision model; vector encoding the device control parameter vector to obtain a 5G URLLC control instruction;

[0008] Step S4: Obtaining a canopy actuator list; searching for a canopy actuator communication link based on the canopy actuator list; issuing a 5G URLLC control instruction to the canopy actuator communication link based on the 5G network to obtain canopy execution feedback data; generating optimized parameters for the canopy execution feedback data and returning them to the canopy multi-objective decision model for optimization to obtain a canopy optimization decision model;

[0009] Step S5: Generating an AR navigation instruction set based on the canopy optimization decision model; triggering an audible and visual warning signal using the AR navigation instruction set and generating a reinforcement suggestion coordinate to obtain a canopy communication whole-process monitoring report.

[0010] The application has the beneficial effect that, by constructing a closed-loop data-driven link containing perception, processing, control and feedback, the intelligentization of the whole process from information collection to decision control of the tent structure is realized. In the data collection link, the system deploys a multi-source sensor array to collect multi-dimensional data such as structural stress, tent cloth deformation, environmental temperature and humidity and personnel trajectory in real time, and the obtained heterogeneous raw data forms a structured data set after unified preprocessing, providing unified data format support for subsequent modeling. Through timestamp alignment operation, the consistency of different data streams in the time dimension is ensured, and the timing deviation problem caused by asynchronous sampling is eliminated. Further, a dynamic thermal matrix is constructed for the synchronized data set, and principal component analysis (PCA) is applied to extract the core correlation features between multi-dimensional data, thereby reducing the interference of redundant information and constructing a fusion feature matrix with high interpretability. On this basis, the system establishes a multi-objective control decision model, and converts the output device control parameter vector of the model into 5G URLLC control instructions through vector encoding technology, realizing low-delay and high-reliability execution instruction issuing. At the execution end, the system dynamically constructs a communication link based on the actuator list, collects and returns feedback data, generates a residual matrix in combination with the theoretical control value, and thus continuously optimizes the model performance by using the gradient fine-tuning mechanism, thereby improving the accuracy of instruction execution. Finally, the system maps the optimized control results into an AR navigation instruction set, and generates reinforcement suggestions and alarm feedback by combining alarm thresholds and position information, effectively supporting the automatic generation and visual output of the communication whole-process monitoring report. This process realizes a complete closed loop of multi-source heterogeneous data fusion, dimension reduction, control, feedback and re-optimization at the data level, ensuring that the system has dynamic adaptability, fast responsiveness and high precision. Therefore, by constructing a closed-loop perception and control system based on multi-source fusion data, the application solves the problems of information isolation, response lag and inaccurate control in traditional tent management, and improves the comprehensive perception ability and dynamic control efficiency of the intelligent tent in complex environments.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: piezoelectric stress sensors are deployed at key nodes of the tent support, with a range of 0-50kN, a sampling frequency of 20Hz, an accuracy of 0.5%, and tent support stress time series data is collected;

[0013] Step S12: micro-deformation optical fiber sensors are deployed on the surface of the tent cloth, with a resolution of 0.01mm and a sampling frequency of 10Hz, and tent cloth deformation gradient data is collected;

[0014] Step S13: temperature and humidity-CO2 multi-parameter probes are uniformly deployed in the internal space, with a temperature accuracy of ±0.3℃ and a sampling frequency of 1Hz, and environmental state data is collected;

[0015] Step S14: embed the personnel safety cap with a UWB positioning tag, with a positioning accuracy of 10 cm, a refresh frequency of 5 Hz, and generate personnel coordinate trajectory data;

[0016] Step S15: use the wavelet threshold denoising algorithm to filter the support frame stress time series data, the tarp deformation gradient data, the environmental state data and the personnel coordinate trajectory data with a signal-to-noise ratio (SNR) greater than 30 dB, and compensate the baseline drift with a drift rate less than 0.03% / month, to obtain the tarp heterogeneous original data set.

[0017] The piezoelectric stress sensor and the micro-deformation optical fiber sensor work together to record the time series stress changes of the key nodes of the support frame and the micro-deformation response of the tarp surface at a sampling frequency of 20 Hz and 10 Hz, respectively, to provide continuous and stable data support for subsequent structural mechanics modeling and material response analysis. Secondly, by deploying a multi-parameter environmental probe with a temperature accuracy of ±0.3℃ and a sampling frequency of 1 Hz, the system realizes continuous monitoring of the internal thermal and humid environment of the tarp house, which can effectively reflect the influence of local temperature and humidity abnormalities on the structure and personnel behavior. In addition, the personnel trajectory data generated by the UWB positioning module based on 10 cm accuracy and 5 Hz refresh frequency not only provides high spatio-temporal resolution of crowd dynamic information, but also can be used for subsequent density estimation, risk prediction and navigation strategy formulation. In terms of data quality control, the wavelet threshold denoising technology is introduced to denoise the above four types of time series data, and the signal-to-noise ratio is constrained to be greater than 30 dB during the filtering process to ensure that the extracted features have high signal purity. At the same time, the baseline drift compensation (controlled to be less than 0.03% / month) of the sensor drift stabilizes the long-term data trend, enhancing the long-term comparability and modeling robustness of the data. Through the above full-link heterogeneous data acquisition and quality control process, the tarp heterogeneous original data set covering structural physical response, environmental microclimate change and personnel spatial behavior is finally formed, providing a unified standard, spatio-temporal continuity and high-precision data basis for subsequent multi-target perception fusion, intelligent control instruction generation and risk identification. Therefore, by constructing a multi-source high-precision data acquisition and denoising mechanism, the scheme solves the problems of missing information dimension, inconsistent data quality and serious perception blind area in traditional tarp monitoring, and improves the response sensitivity and processing accuracy of the tarp system to sudden structural abnormalities and environmental changes.

[0018] Preferably, the step S2 of constructing a thermal matrix for the synchronized tarp data set comprises:

[0019] Time dimension feature extraction is performed on the synchronized tarp data set, and a time axis label is added to obtain a tarp time axis feature vector.

[0020] The spatial relationship topology feature of the synchronized tarp data set is constructed, and a topology edge is initialized to obtain a tarp spatial topology relationship graph.

[0021] The personnel trajectory density distribution analysis is performed on the synchronized tent data set, and low-density data is removed to obtain a personnel heat matrix.

[0022] The application extracts and processes time-dimension features and time axis markers, periodically reconstructs data collected by various sensors, maps environment, structure and behavior data under different sampling frequencies to a standardized time axis, extracts key time features such as time interval change rate, mutation point identification and periodic disturbance, forms a time axis feature vector with dynamic evolution information, and thus enhances the time sequence sensitivity of the system to critical state changes. Secondly, in the spatial dimension, according to the internal structure arrangement of the tent, the physical distribution of the sensor and the construction logic of the support node, a spatial relationship topology graph containing nodes and edges is constructed, and an edge weight calculation method based on Euclidean distance or Manhattan distance is introduced in the graph initialization process, so that the spatial structure has a quantifiable adjacency relationship. The initialization of the topological edge not only reflects the relative position and correlation strength between the monitoring points, but also provides a basic structure constraint for the subsequent graph neural network or structure propagation model. Thirdly, in the behavior feature dimension, the system introduces a trajectory density estimation algorithm to perform kernel density estimation on the UWB positioning data, dynamically identifies the gathering hot area and flow path of personnel in space, and removes the data in the low-density area to remove noise, and finally constructs a high-resolution personnel heat matrix. The heat matrix not only reflects the real-time personnel distribution, but also can be used to identify potential congestion risks or abnormal behavior areas. Through the independent extraction and standardized expression of the above-mentioned three-dimensional features of time, space and behavior, the method realizes the structure cleaning, semantic enhancement and noise suppression of heterogeneous information at the data level, and improves the calculation effectiveness and analysis depth of the whole system in multi-target perception, behavior prediction and risk identification.

[0023] Preferably, the principal component analysis based on the dynamic heat matrix of the tent in step S2 comprises:

[0024] Align the tent time axis feature vector, the tent spatial topology relationship graph and the personnel heat matrix in three dimensions, and perform three-dimensional tensor splicing to obtain a tent three-dimensional feature space;

[0025] Perform principal component analysis dimension reduction on the tent three-dimensional feature space to obtain tent principal component dimension reduction data;

[0026] Perform 95% variance retention on the tent principal component dimension reduction data, and perform three-dimensional feature re-fusion to obtain a tent fusion feature matrix.

[0027] The application ensures that data from heterogeneous sources has consistent mapping relationship in three dimensions of time step, spatial node and personnel distribution density by aligning the canopy time axis feature vector, spatial topology relationship graph and personnel heat matrix in three-dimensional features, and then constructing a three-dimensional tensor structure with space-time collaborative semantics. The three-dimensional feature space not only can express the structural state change and personnel dynamic coupling relationship in the system completely, but also can be used as high-order input for deep learning or feature screening analysis. In order to avoid the redundant features and noise interference of the original tensor in the high-dimensional space, the system introduces principal component analysis (PCA) algorithm to reduce the dimension of the tensor in the data processing stage, obtains the contribution rate of each principal component through eigenvalue decomposition of the covariance matrix, and truncates the redundant dimensions according to the principle of reserving 95% of the cumulative variance contribution rate, to obtain a high-information-density reduced-dimension tensor. Then, the three-dimensional feature re-fusion operation is performed on the reduced-dimension result, and the compressed time, space and personnel behavior principal components are reconstructed into a low-dimensional dense tensor structure to form a canopy fusion feature matrix with traceability and distinguishability. The matrix significantly reduces the computational load in the subsequent model training and reasoning process while maintaining the integrity of the core semantic structure, improving the response rate and accuracy robustness of the control parameter output. Therefore, the method effectively overcomes the problems of inconsistent dimensions, redundant and complex fusion difficulty and other problems among heterogeneous data at the data level, realizes the deep collaborative expression and efficient compression of multi-source information, and enhances the identification ability and adaptability of the canopy system to complex events.

[0028] Preferably, the step S3 of constructing a decision model by using the canopy fusion feature matrix comprises:

[0029] The canopy fusion feature matrix is subjected to structural stress-deformation data extraction to obtain canopy structural stress-deformation data; and the canopy structural stress-deformation data is subjected to joint entropy calculation to obtain canopy safety risk theoretical grade data.

[0030] The canopy fusion feature matrix is subjected to environment-personnel data extraction to obtain canopy environment-personnel data; and the canopy environment-personnel data is subjected to temperature-humidity-CO2 comfort degree scoring to generate a canopy is moderate deviation value.

[0031] Pareto optimization is performed according to the canopy safety risk theoretical grade data and the canopy is moderate deviation value, and a model is constructed to output a canopy multi-objective decision model, wherein the constraint condition of the Pareto optimization is that the structural stress is less than 40kN and the upper limit of energy consumption is less than 3kW.

[0032] The present application extracts the structural stress and deformation information in the canopy fusion feature matrix, forms a multi-dimensional structure time series data vector, and quantifies the correlation and information uncertainty between each stress-deformation variable through a joint entropy calculation method to generate canopy safety risk theoretical grade data with physical risk interpretation, ensuring that the structure state evaluation has statistical robustness and physical consistency. In terms of environmental and personnel data processing, the system performs joint feature extraction on parameters such as temperature and humidity, CO2 concentration, and personnel distribution density, constructs a multi-dimensional vector describing the environmental experience, and introduces a temperature and humidity-comfort score model to quantify subjective feelings, and finally generates an "appropriate deviation value" index reflecting the degree of environmental deviation. To achieve joint solving of the two types of objective functions, the system introduces a Pareto optimization framework in the modeling stage, taking "structural stress less than 40 kN" and "upper limit of energy consumption not more than 3 kW" as hard constraints, and searching for the optimal compromise solution of structure safety level and environmental comfort deviation in the search space. The multi-objective non-inferior frontier surface model improves the comprehensive expression ability of conflicting objectives. The final generated canopy multi-objective decision model can output control suggestions that meet the constraint conditions based on inputting multi-source state characteristics, ensuring that a balanced mechanism is established among structural bearing capacity, energy consumption efficiency and personnel comfort, and realizing effective fusion control of high-dimensional heterogeneous data. Therefore, this method breaks through the problem of traditional control logic relying on a single index at the data level, realizes parallel optimization of structural safety and living experience, and significantly enhances the intelligent response ability and multi-scene adaptability of the canopy environmental system.

[0033] Preferably, step S4 comprises the following steps:

[0034] Step S41: obtaining a canopy actuator list; searching for a canopy actuator communication link based on the canopy actuator list; issuing a 5G URLLC control instruction based on the 5G network to the canopy actuator communication link to obtain canopy execution feedback data;

[0035] Step S42: obtaining theoretical execution value data; comparing the canopy execution feedback data and the theoretical execution value data by error difference to obtain a canopy model residual matrix;

[0036] Step S43: based on the canopy model residual matrix, the gradient backpropagation fine-tuning of the canopy multi-objective decision model is performed with a learning rate of 0.001 to obtain a canopy optimization decision model.

[0037] The application identifies all valid communication links by completely obtaining the list of canopy actuators and constructing a communication topology graph based on unique device identification. In the issuing stage, the 5G URLLC (Ultra-Reliable Low-Latency Communication) protocol is used to transmit device control parameters in a vector encoding manner for high-reliability instruction transmission, ensuring that different types of actuators (such as variable frequency fans, compressor refrigeration, intelligent light control, etc.) respond synchronously under low latency conditions. After execution is completed, the system constructs a canopy execution feedback data set from the real-time environmental parameter feedback data (such as wind speed, temperature, brightness, etc.) collected by each actuator. To realize error modeling, the system introduces theoretical execution value data, uses a multi-dimensional vector difference algorithm based on a time window to calculate the residual error of feedback values and theoretical values one by one, and eliminates abnormal noise points, finally generating a canopy model residual error matrix with time stamp, spatial identifier and error component. The residual error matrix is used as a key input for model optimization, and a back propagation algorithm with a fixed learning rate of 0.001 is used to correct the gradient of the canopy multi-objective decision model, automatically update the weight coefficients and feature mapping relationship of the control parameters in the model, and generate an optimized decision model that adapts to the current environmental dynamics. This optimization mechanism not only enables the system to have real-time self-repairing capability for multi-source control accuracy deviation, but also improves the generalization performance of the model under new data distribution, thereby maintaining the response speed and control robustness of the canopy environment regulation in dynamic load, sudden environmental disturbance or personnel flow drastic change scenarios, significantly improving the intelligent level and sustainable operation and maintenance capability of the overall system.

[0038] Preferably, step S41 comprises the following steps:

[0039] Step S411: Obtain a list of canopy actuators; search for a communication link of the canopy actuator based on the list of canopy actuators;

[0040] Step S412: Issue 5G URLLC control instructions based on the 5G network to the communication link of the canopy actuator, including adjusting the ventilation volume by adjusting the speed of the variable frequency fan to obtain ventilation environment feedback data; adjusting the temperature value by adjusting the power of the low air conditioner compressor to obtain canopy air conditioner adjustment data; changing the light intensity of the LED array to obtain canopy light adjustment data;

[0041] Step S413: Superimpose the ventilation environment feedback data, the canopy air conditioner adjustment data and the canopy light adjustment data to obtain the canopy execution feedback data.

[0042] The application completes the rapid search and verification of the communication link by obtaining the complete canopy executor list, using the unique device identifier and the communication protocol parameter, and ensures the efficient transmission path of the control instruction. The low latency and high reliability communication based on the 5G URLLC protocol ensures the real-time and accurate delivery of the instruction from the central decision unit to the specific executor (including the frequency conversion fan, air conditioner compressor and LED light source array). The specific encoding of the instruction includes fine-grained adjustment of the frequency conversion fan speed, which is used to dynamically control the air flow and generate ventilation environment feedback data in the form of time sequence signal reflecting the change of wind speed and pressure; the power of the air conditioner compressor is fine-tuned to realize the dynamic adaptation of the temperature set point and form the temperature regulation feedback data, and the environmental temperature change curve is returned through the sensor network; the light intensity of the LED array is adjusted, and the feedback data is represented as spatial mapping information of light distribution. Subsequently, the system adopts a weight superposition algorithm to normalize the three types of feedback data, and introduces a weight factor based on environmental demand, time period and safety level to realize the fusion representation of multi-dimensional data and construct a standardized canopy execution feedback data set. The data set not only maintains the dynamic time sequence characteristics and spatial distribution characteristics of the feedback of each subsystem, but also reflects the relative importance of the adjustment of each environmental parameter through weight assignment, supporting the optimization iteration of the subsequent multi-objective decision model and the update of the accurate control strategy. Therefore, the system realizes the closed-loop interaction from the instruction issuing to the execution feedback at the data level, and guarantees the real-time, accuracy and adaptability of the canopy environment control. Preferably, step S5 comprises the following steps:

[0043] Step S51: generating an AR navigation instruction set based on the canopy optimization decision model;

[0044] Step S52: if the canopy safety risk theoretical level data is greater than or equal to level 3, triggering an audible and visual alarm signal based on the AR navigation instruction set through the canopy executor communication link, and generating reinforced recommendation coordinates to obtain intelligent canopy communication optimization data;

[0045] Step S53: constructing a canopy communication whole-process monitoring report based on the intelligent canopy communication optimization data.

[0046] The application realizes a high dynamic risk linkage response mechanism based on data-driven by fusing the optimization decision result with AR navigation, man-machine warning system and data archiving mechanism, thereby establishing a self-adaptive, multi-dimensional collaborative monitoring feedback system in the tent operation environment. At the data level, first, based on the AR navigation instruction set generated by the tent optimization decision model, not only the spatial pose, personnel density and dynamic risk level are highly decoupled, but also the features of codeable, reportable and interactive are possessed, which is convenient for low-delay visual loading and real-time semantic feedback in terminal equipment. When the safety risk theory level value is greater than or equal to 3, the system will automatically filter the high-priority navigation content associated with the risk level and convert it into a control instruction under the 5G URLLC protocol structure, ensuring the high reliability and millisecond transmission delay of the instruction link and improving the response rate in the burst state. At the same time, the signal frequency, intensity and duration triggered by the sound and light warning system are linked and set with the risk classification label in the navigation instruction set, realizing differentiated prompt strategies in different risk scenarios. The generation of reinforced recommendation coordinates is based on the interpolation mapping and main core area calibration of the structural stress main distribution area and the high-density point of personnel aggregation, and the output result has coordinate-level positioning accuracy, which can be directly called by maintenance personnel or intelligent assembly units to realize the point reinforcement of the structure state. Finally, the system encapsulates the execution feedback, path adjustment record, alarm response event, coordinate activation data and user response delay collected from the whole process from AR navigation instruction triggering to executor response, constructs a standardized communication whole-process monitoring report, and forms a data closed loop in the system, which is convenient for subsequent calling in modeling retraining, trend prediction and operation and maintenance evaluation modules. This process not only improves the semantic coupling degree between multi-source heterogeneous data, but also significantly enhances the linkage response capability of the system to high-risk situations. Preferably, step S51 comprises the following steps:

[0047] Step S511: acquiring real-time personnel trajectory data;

[0048] Step S512: inputting the real-time personnel trajectory data into the tent optimization decision model to obtain tent local area density data, when the tent local area density data is greater than 0.5 person / m 2 For high risk, personnel path intersection conflict point prediction is performed to obtain personnel guidance prediction data;

[0049] Step S513: generating a turning angle based on the personnel guidance prediction data and outputting a walking speed recommendation to obtain an AR navigation instruction set.

[0050] The application constructs a two-dimensional trajectory matrix in the form of a timing sequence with a sampling period of 200 ms based on high-frequency real-time trajectory data collected by the UWB positioning module, and uses a sliding window density kernel estimation algorithm to count the number of personnel in each local space unit (such as a 2m*2m grid) in a unit of time, thereby calculating the local area density data. After the density value exceeds the preset safety threshold (0.5 person / m 2 ) value, the system automatically determines that the area is a high-risk gathering area, and calls the spatial path topology graph to perform path intersection conflict detection, uses a path overlap prediction algorithm based on trajectory intersection metrics to identify mobile trend points with high overlap, and generates personnel guidance prediction data. The prediction data includes the occurrence probability, expected conflict time, participant ID and speed vector of each intersection conflict point, and performs local disturbance calculation on individual paths in the form of a vector field. Subsequently, the system calculates the optimal turning angle for each person according to the prediction data, and generates a walking speed suggestion value based on the current pedestrian density distribution, the speed of the individual and the environmental traffic conditions, thereby constructing an AR navigation instruction set containing multi-dimensional action instructions. The instruction set can be synchronized in real time with the AR device worn by the user to prompt the avoidance direction and movement rhythm, thereby ensuring the smoothness of the pedestrian flow while avoiding potential collision risks in the intersection paths. Overall, this method realizes a closed-loop control path from trajectory collection, density modeling, conflict prediction to instruction generation at the data level, significantly enhancing the refinement and intelligence level of crowd management in the tent space.

[0051] In the present specification, an intelligent tent communication system based on 5G communication is provided for performing the intelligent tent communication method based on 5G communication described above, and the intelligent tent communication system based on 5G communication comprises:

[0052] A multi-source data acquisition and heterogeneous construction module is configured to acquire environment, structure and personnel data by deploying a multi-source sensor array in the tent structure, perform data preprocessing, and construct a tent heterogeneous raw data set;

[0053] A space-time synchronization and feature fusion module is configured to align the tent heterogeneous raw data set by time stamping to obtain a synchronized tent data set, construct a dynamic heat matrix of the tent based on the synchronized tent data set, and perform principal component analysis based on the dynamic heat matrix of the tent to obtain a tent fusion feature matrix;

[0054] A multi-target intelligent decision-making and control instruction generation module is configured to construct a decision-making model using the tent fusion feature matrix to obtain a tent multi-target decision-making model, output a device control parameter vector based on the tent multi-target decision-making model, and perform vector encoding on the device control parameter vector to obtain a 5G URLLC control instruction;

[0055] The link control and model closed-loop optimization module is used for obtaining a tent executor list; searching a tent executor communication link based on the tent executor list; issuing a 5G URLLC control instruction based on the 5G network to the tent executor communication link to obtain tent execution feedback data; generating optimized parameters for the tent execution feedback data, and returning the optimized parameters to a tent multi-objective decision model for optimization to obtain a tent optimization decision model;

[0056] The augmented reality navigation and intelligent early warning module is used for generating an AR navigation instruction set based on the tent optimization decision model; triggering an audible and light warning signal by using the AR navigation instruction set, and generating a reinforcement suggestion coordinate to obtain a tent communication whole-process monitoring report.

[0057] The present application has the beneficial effect that, by constructing a closed-loop data-driven link containing perception, processing, control and feedback, the whole-process intelligentization of the tent structure from information collection to decision control is realized. In the data collection link, a multi-source sensor array is deployed to collect multi-dimensional data such as structural stress, tarp deformation, environmental temperature and humidity, and personnel trajectory in real time. The obtained heterogeneous raw data forms a structured data set after unified preprocessing, providing unified data format support for subsequent modeling. Through timestamp alignment operation, the consistency of different data streams in the time dimension is ensured, and the timing deviation problem caused by asynchronous sampling is eliminated. Further, a dynamic thermal matrix is constructed for the synchronized data set, and principal component analysis (PCA) is applied to extract the core correlation features between multi-dimensional data, thereby reducing the interference of redundant information and constructing a fusion feature matrix with high interpretability. On this basis, the system establishes a multi-objective control decision model, and converts the output device control parameter vector of the model into a 5G URLLC control instruction through vector encoding technology, realizing low-delay and high-reliability execution instruction issuing. At the execution end, the system dynamically constructs a communication link based on the executor list, collects and returns feedback data, and generates a residual matrix in combination with the theoretical control value, so as to continuously optimize the model performance by using the gradient fine-tuning mechanism, and improve the accuracy of instruction execution. Finally, the system maps the optimized control result into an AR navigation instruction set, and generates reinforcement suggestions and alarm feedback by fusing alarm thresholds and position information, effectively supporting the automatic generation and visual output of the communication whole-process monitoring report. This process realizes a complete closed loop of multi-source heterogeneous data fusion, dimension reduction, control, feedback and re-optimization at the data level, ensuring that the system has dynamic adaptability, fast responsiveness and high precision. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 It is a step flowchart of an intelligent tent communication method based on 5G communication;

[0059] Figure 2 It is Figure 1 It is a detailed implementation step flowchart of step S4 in the method;

[0060] The objectives, functional characteristics and advantages of the present application will be further described with reference to the embodiments and in conjunction with the accompanying drawings. DETAILED DESCRIPTION

[0061] The technical method of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0062] In addition, the accompanying drawings are only schematic illustrations of the present application, and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated description thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities, which do not necessarily have to correspond to physically or logically independent entities. The functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0063] It should be understood that although the terms "first", "second" and the like can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the exemplary embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0064] To achieve the above-mentioned object, please refer to Figures 1 to 2 The method comprises the following steps:

[0065] Step S1: A multi-source sensor array is deployed on the canopy structure to collect environmental, structural and personnel data, and data preprocessing is performed to construct a canopy heterogeneous raw data set;

[0066] Step S2: The canopy heterogeneous raw data set is time-stamped and aligned to obtain a synchronized canopy data set. The synchronized canopy data set is subjected to heat matrix construction to obtain a canopy dynamic heat matrix. Principal component analysis is performed based on the canopy dynamic heat matrix to obtain a canopy fusion feature matrix;

[0067] Step S3: a decision model is constructed by using the canopy fusion feature matrix, a canopy multi-objective decision model is obtained, a device control parameter vector is output based on the canopy multi-objective decision model, the device control parameter vector is vector encoded, and a 5G URLLC control instruction is obtained;

[0068] Step S4: a canopy actuator list is acquired, a canopy actuator communication link is searched based on the canopy actuator list, a 5G URLLC control instruction is issued to the canopy actuator communication link based on the 5G network, canopy execution feedback data is obtained, optimized parameters are generated based on the canopy execution feedback data, and the optimized parameters are fed back to the canopy multi-objective decision model for optimization, and a canopy optimization decision model is obtained;

[0069] Step S5: an AR navigation instruction set is generated based on the canopy optimization decision model, an audible and light warning signal is triggered by using the AR navigation instruction set, reinforced suggestion coordinates are generated, and a canopy communication whole-process monitoring report is obtained.

[0070] In the embodiment of the application, reference Figure 1 As shown in the figure, it is a step flow diagram of an intelligent canopy communication method based on 5G communication, in the example, the intelligent canopy communication method based on 5G communication comprises the following steps:

[0071] Step S1: a multi-source sensor array is deployed on a canopy structure to collect environment, structure and personnel data, data preprocessing is performed, and a canopy heterogeneous original data set is constructed;

[0072] In the embodiment of the present application, a multi-source sensor array needs to be arranged at the key load-bearing components, node connection areas and main entrances of the canopy, which includes but is not limited to temperature and humidity sensors, anemometers, structural strain gauges, tilt sensors, ultrasonic range finders, video image acquisition devices and human infrared thermal sensing detectors, etc., to ensure the coverage of three core elements of environmental parameters, structural response information and personnel dynamic data. Various sensors continuously or quasi-real-time data acquisition according to the established sampling frequency, generating multi-dimensional raw data streams. Due to the high heterogeneity of different sensors in physical characteristics, data structure and transmission protocol, a unified multi-source data access interface needs to be established, and a standardized data format packaging mechanism is used to normalize the expression of raw data, thereby constructing a unified buffer pool of multi-modal raw data. On this basis, a distributed timestamp calibration mechanism is used to assign a unified collection time label to various data streams, to realize the basic guarantee of subsequent synchronization processing. In addition, preliminary data filtering and denoising operation is also needed for structural data, and methods such as sliding window median filtering and wavelet transform denoising are used to eliminate the pseudo signals introduced by vibration, communication delay or external interference. Environmental parameter data needs to be subjected to section average and volatility analysis to identify the mutation abnormal area. For human activity data, image target recognition algorithm is combined to extract hot spots and preliminarily divide the motion trajectory. Finally, a canopy heterogeneous raw data set with time label, spatial positioning, modal classification and preliminary cleaning features is formed, laying a foundation for subsequent spatio-temporal alignment and feature fusion processing.

[0073] Step S2: time stamp alignment of the canopy heterogeneous raw data set to obtain a synchronized canopy data set; thermal matrix construction of the synchronized canopy data set to obtain a canopy dynamic thermal matrix; principal component analysis based on the canopy dynamic thermal matrix to obtain a canopy fusion feature matrix;

[0074] In the embodiment of the present application, based on the heterogeneous raw data set constructed in step S1, high-precision timestamp calibration and alignment of various data streams are required. Due to the asynchronous sampling characteristics of different sensor devices, a network time protocol (NTP) or high-precision time synchronization protocol (such as IEEE 1588) is used to realize a full-network time synchronization mechanism, and each data packet is bound to a unified time reference. Subsequently, interpolation mapping, missing compensation and time sliding window correction algorithms are used to reposition the data with time offset, ensuring that all data frames are aligned according to a unified time axis, and finally a synchronized cabin data set is constructed. On this basis, in order to capture the spatial distribution characteristics of the internal environment and structural dynamic state of the cabin, a thermal matrix construction mechanism is introduced. This process first maps the data in three-dimensional space according to the spatial coordinate system of the sensor deployment, projects different modal data according to their measurement point positions to a unified spatial grid; then, a weighted superposition model is used to normalize various modal data, and a spatial thermal weight distribution is constructed through a modal weight adjustment factor, thereby generating a dynamic thermal matrix of the cabin that can evolve over time. This thermal matrix expresses the response intensity of different spatial regions in the integrated parameters at each time in the form of a two-dimensional tensor. In order to reduce the data dimension and enhance the processing efficiency, principal component analysis (PCA) is used to reduce the dimension of the dynamic thermal matrix. The specific operation includes constructing the covariance matrix and eigenvalue decomposition of the thermal matrix sequence, mapping the original high-dimensional tensor to the fusion feature space by retaining the principal component vectors whose cumulative variance contribution rate is greater than a certain threshold (such as 95%), and obtaining a representative cabin fusion feature matrix. This feature matrix not only retains the key structural information of multi-modal data in the time-space dimension, but also provides a data basis for the construction of subsequent intelligent decision and control models.

[0075] Step S3: using the cabin fusion feature matrix to construct a decision model to obtain a cabin multi-objective decision model; outputting a device control parameter vector based on the cabin multi-objective decision model; vector encoding the device control parameter vector to obtain a 5G URLLC control instruction;

[0076] In the embodiment of the application, key decision factors are extracted from a high-dimensional feature vector space, a multi-objective trade-off model is constructed, and data instructions that can be used for 5G URLLC control links are generated through coding strategies. First, on the basis of completed principal component analysis, the obtained tent fusion feature matrix expresses the information of multiple physical dimensions such as environmental changes, structural responses and personnel behaviors in the form of multivariate joint features after dimension reduction. Next, a decision model needs to be constructed based on the feature matrix. The model adopts a multi-objective optimization strategy to configure weights for control targets (such as structural stability, personnel density balance, ventilation rate, emergency response time, etc.). To construct the model, a linear weighted model, a fuzzy comprehensive evaluation method or a neural network architecture based on gradient descent is usually introduced to adjust each objective function in a unified optimal solution domain. The model training stage can be iteratively converged through historical stress response data, scene change records and sensor measurement feedback, and finally a mapping relationship is formed with the input being the fusion feature vector and the output being the control variable sequence. The output result is a multi-dimensional device control parameter vector, which includes fine control indicators such as temperature control system start-stop threshold, open-closed window execution angle, emergency lighting trigger level, air flow adjustment range, etc. In order to adapt to the requirements of 5G URLLC (Ultra-Reliable Low-Latency Communication) protocol, real-time encoding processing needs to be performed on the control parameter vector. The encoding method includes constructing a sparse vector index structure, performing data compression and error tolerance processing, and generating control instruction frames combined with short frame encoding mechanism. In this encoding process, the instruction data needs to meet the communication requirements of compactness, fast decodability and strong robustness. The control instruction after encoding is used as the 5G URLLC control instruction issued in the subsequent communication link, providing basic data-driven support for accurate device execution. This process constitutes a key central link from multi-modal data to target-driven control.

[0077] Step S4: obtaining a tent actuator list; searching for a tent actuator communication link based on the tent actuator list; issuing 5G URLLC control instructions to the tent actuator communication link based on the 5G network, obtaining tent actuator feedback data; generating optimization parameters for the tent actuator feedback data and returning them to the tent multi-objective decision model for optimization, obtaining a tent optimization decision model;

[0078] In the embodiment of the present application, by querying the list of actuator devices configured in the tent structure, a list of actuators including electric window openers, exhaust fans, lighting devices, temperature control units, reinforcement support drive assemblies, etc. is obtained, and a unique identifier is assigned to each type of actuator and a control authority and response parameter mapping table is established. Then, the system constructs a communication path identification process based on the actuator list, uses a link addressing algorithm based on a graph structure and a channel reachability detection mechanism to filter and confirm the reachable communication path of each actuator, and forms a logical control link topology graph. On this basis, using the URLLC (Ultra-Reliable Low Latency Communication) channel in the 5G network, the control instructions that have been vector encoded are distributed to each target actuator through the network scheduling mechanism, completing the parallel sending of instructions. Each actuator changes state after receiving the instructions and transmits execution feedback data in real time. The feedback data includes execution state identification (such as success, delay, failure), execution delay time, physical action response value (such as angle change, wind speed change, support displacement), and electrical characteristic parameters (such as current fluctuation, drive energy consumption, etc.), all feedback signals are encapsulated and reported to the edge computing node or central processing unit according to the pre-defined protocol in the 5G backhaul link. Then, the system uses these feedback data and the original control parameters for comparison and analysis, identifies the control effect deviation and uses a constraint optimization algorithm (such as a multi-objective adjustment model based on Lagrange multipliers or a gradient enhancement learning mechanism) to correct errors, generating an optimized parameter vector. Finally, the optimized parameters will be returned to the original tent multi-objective decision model as adjustment factors, realizing the local update of the model structure and the dynamic self-adaptation of the control parameter space, thereby forming an iterative optimization model based on real-time feedback, i.e. the tent optimization decision model, to guide the generation and issuance of subsequent control instructions, realizing a complete data closed loop from control execution to model evolution.

[0079] Step S5: generating an AR navigation instruction set based on the tent optimization decision model; triggering an audible and visual alarm signal using the AR navigation instruction set, and generating reinforcement suggestion coordinates to obtain a tent communication whole-process monitoring report.

[0080] In this embodiment of the invention, an optimized tent multi-objective decision-making model is invoked. Under the current sensor data input, a set of structural response prediction values ​​and personnel behavior offset indicators are output. The system then uses a rule engine and a feature weight table to jointly determine the spatial areas requiring guidance and intervention. Subsequently, the system encodes these spatial locations requiring operational responses into a set of three-dimensional coordinate points and maps them to the spatial index structure in the digital tent model, forming a set of reinforcement suggestion coordinates that can be used for positioning and navigation. Based on this, the system uses the SLAM (Simultaneous Localization and Mapping) positioning mechanism and real-time pose data from IMU sensors to determine the actual relative spatial position of the user terminal or operator within the tent. It then combines this with navigation logic to generate an AR navigation instruction set containing "path nodes, turning instructions, and behavioral action prompts." This instruction set uses a structured data format and includes fields such as guidance sequence number, action type identifier (e.g., turning, stopping, checking, support reinforcement), target coordinates, text prompt semantics, audio encoding index, and graphics rendering resource index, which serve as input parameters for the AR engine. On the rendering side, the system calls the 3D modeling engine to map the visual objects in the navigation command set to virtual components in the augmented reality scene. By overlaying this onto the real-time camera feed, it displays audio-visual guidance elements (such as flashing arrows, warning icons, and voice announcements) in structural hotspot areas to the user, thus achieving visual guidance. Regarding the warning mechanism, the command set can trigger the audio-visual alarm system based on model weight priority, calling alarm devices bound to corresponding spatial coordinates. It also controls the alarm activation sequence, intensity level, and duration according to the navigation guidance process, forming a time-linked guidance feedback loop. Finally, the system integrates all process data, including navigation generation basis, alarm trigger records, coordinate response status, and user behavior path data, to generate a standardized, structured, and traceable tent communication full-process monitoring report, providing data support for subsequent retrospective analysis and safety assessment.

[0081] Preferably, step S1 includes the following steps:

[0082] Step S11: Deploy piezoelectric stress sensors at key nodes of the tent support structure, with a range of 0-50kN, a sampling frequency of 20Hz, and an accuracy of 0.5%, to collect time-series data of stress on the support structure.

[0083] Step S12: Deploy a micro-deformation fiber optic sensor on the surface of the tarpaulin with a resolution of 0.01 mm and a sampling frequency of 10 Hz to collect tarpaulin deformation gradient data;

[0084] Step S13: Distribute temperature, humidity and CO2 multi-parameter probes evenly in the internal space, with a temperature accuracy of ±0.3℃ and a sampling frequency of 1Hz, to collect environmental status data;

[0085] Step S14: embedding a UWB positioning tag in the personnel safety helmet, positioning accuracy 10 cm, refresh frequency 5 Hz, generating personnel coordinate trajectory data;

[0086] Step S15: using a wavelet threshold denoising algorithm to filter the support stress time series data, the tarp deformation gradient data, the environment state data and the personnel coordinate trajectory data with a signal-to-noise ratio (SNR) greater than 30 dB, and compensating the baseline drift with a drift rate less than 0.03% / month, to obtain a tarp heterogeneous original data set.

[0087] In the embodiment of the application, piezoelectric stress sensors with a range of 0-50 kN are arranged at the key nodes of the tarp support, and the stress on the support is sampled in real time at a high frequency of 20 Hz to generate high-precision stress time series data streams with an upper limit of error of 0.5%; at the same time, micro-deformation fiber sensors are uniformly arranged on the surface of the tarp, and the gradient change value of the micro tensile deformation is collected by using the deformation resolution of 0.01 mm and the sampling frequency of 10 Hz to capture the stress response of the flexible material under different working conditions. At the environmental level, the stable environmental parameter change data of the tarp interior is obtained by arranging a temperature and humidity multi-parameter probe (temperature accuracy ±0.3℃, sampling frequency 1 Hz) to build the coupling relationship between the modeling environment field and the structure behavior; the personnel behavior data is realized by embedding a high-precision UWB positioning tag in the safety helmet, and the spatial resolution reaches 10 cm with a refresh frequency of 5 Hz, which can be used to generate time-continuous two-dimensional or three-dimensional coordinate trajectory data to depict the activity distribution situation of the personnel in the tarp. The above-mentioned various heterogeneous original data is uniformly subjected to a wavelet threshold denoising process after collection, a multi-scale wavelet decomposition method with a hard threshold function is selected to perform compression and reconstruction operations on the time series data in different scale spaces, high-frequency interference components with low signal-to-noise ratio are screened out, and low signal-to-noise data segments are removed by a preset threshold function with an SNR greater than 30 dB to improve the overall structure availability of the data and the subsequent modeling stability. In addition, the system performs baseline correction operation on the long-term observation values of all sensors, and the slowly varying noise with a drift rate less than 0.03% / month in the observation sequence is compensated by exponential smoothing and regression by using the drift trend of the time series sampling to weaken the error accumulation effect caused by the thermal drift, aging or interference accumulation of the sensor. Finally, after the synchronization of all data is completed, the system integrates the stress data, deformation data, environmental parameters and personnel trajectory to construct a unified format of the heterogeneous original data set, which provides standardized and verifiable multi-source input samples for the subsequent analysis steps.

[0088] Preferably, the step S2 of constructing a thermal matrix for the synchronized tarp data set comprises:

[0089] The synchronized tarp data set is subjected to time dimension feature extraction and time axis labeling to obtain a tarp time axis feature vector.

[0090] The spatial relationship topology feature of the synchronized tent data set is constructed, and the topology edge is initialized to obtain a tent spatial topology relationship graph.

[0091] The personnel trajectory density distribution analysis is performed on the synchronized tent data set, and the low-density data is removed to obtain a personnel heat matrix.

[0092] In the embodiment of the present application, in the aspect of time dimension feature extraction, the system performs sliding window segmentation processing on various time series signals based on multi-source synchronized data, extracts statistical feature quantities (such as mean, standard deviation, kurtosis, skewness, periodicity, etc.) in each time period, and combines empirical mode decomposition (EMD) and Fourier transform to extract representative low-frequency trends and high-frequency disturbance factors in the signal, forming a time feature subset describing the dynamic change law of different structure / environment / personnel variables. On this basis, by performing unified timestamp alignment processing on various types of sensing data, a standardized time axis is established, the time nodes corresponding to different data sources are marked, and a “tent time axis feature vector” with a unified indexing system is constructed, so that the order is maintained and the window sliding operation is performed during subsequent fusion analysis. Secondly, in the aspect of spatial relationship topology construction, the initial node graph is constructed using the physical connection information between the sensor deployment coordinates and the structure members, each sensor is regarded as a graph node, the topology edge set is defined according to its spatial proximity, structure level relationship or stress propagation path, the edge weight is initialized by the minimum spanning tree algorithm or distance attenuation function, and a “tent spatial topology relationship graph” reflecting the distribution logic of the tent structure is formed, and the data format is adjacency matrix or sparse graph structure. Finally, for personnel behavior heat feature construction, the personnel position information in the time series is rasterized based on the UWB trajectory data, the three-dimensional coordinates are projected onto the two-dimensional plane to construct a spatial grid unit, the personnel stay frequency in each unit per unit time is accumulated to form a trajectory density matrix; then the kernel density estimation method is applied for spatial smoothing and distribution evaluation, the grid regions with an average density lower than a set threshold are removed to reduce the interference of abnormal or edge activity data on the overall heat pattern, and finally a structured “personnel heat matrix” is generated, which is visualized as a density heat map combined with space-time, used for subsequent correlation modeling between structure response and environmental change. The overall process maintains the compressibility of the data format and the consistency of the interface, ensuring that the time-space-personnel three types of features are cross-analyzed in a unified indexing system.

[0093] Preferably, the principal component analysis based on the dynamic heat matrix of the tent in step S2 comprises:

[0094] The tent time axis feature vector, the tent spatial topology relationship graph and the personnel heat matrix are aligned in three dimensions, and three-dimensional tensor splicing is performed to obtain a tent three-dimensional feature space.

[0095] Perform principal component analysis dimension reduction on the three-dimensional feature space of the canopy to obtain canopy principal component dimension reduction data;

[0096] Perform 95% variance retention on the canopy principal component dimension reduction data, and perform three-dimensional feature re-fusion to obtain a canopy fusion feature matrix.

[0097] In the embodiment of the application, in the three-dimensional feature alignment stage, the time axis feature vector (time dimension), the spatial topology relationship graph (structure space dimension) and the personnel heat matrix (space dimension) obtained in the early stage are multi-dimensionally synchronized. Specifically, the system constructs an index mapping relationship matrix to ensure that the three types of features establish a corresponding relationship under a unified time node and a spatial reference coordinate, and normalizes different data scales to have consistent numerical dimensions. On this basis, by three-dimensional tensor splicing operation, the three data sets of different sources are represented by tensorization with time-space-behavior as three-dimensional index axes to form a canopy three-dimensional feature space tensor with shape=[T,S,A] (T is the number of time steps, S is the number of spatial nodes, and A is the number of behavior or state features), and the tensor elements store the state indicators under a specific time, a specific spatial position and a specific behavior parameter. Subsequently, in order to reduce the redundancy of high-dimensional data and extract the main change trend, principal component analysis (PCA) operation is performed on the three-dimensional feature space tensor. Since tensor data cannot be directly applied to traditional two-dimensional PCA, the three-dimensional tensor is first unfolded into a two-dimensional matrix along each mode (mode-1, mode-2, mode-3) by flattening or tensor unfolding method, and the covariance matrix is calculated. On this basis, the first several principal components with the largest eigenvalues are extracted, and according to the retention standard of cumulative contribution rate reaching 95%, the first k principal components are selected to construct a transformation matrix to perform linear projection on the original tensor data to generate a low-dimensional dense representation, i.e. the canopy principal component dimension reduction data. Finally, for the feature space after dimension reduction, in order to restore the original structural dependence between dimensions, the system performs a three-dimensional feature re-fusion operation. The specific method includes: reconstructing the three-axis index system based on the physical attribute labels of the dimension reduction principal components, combining the time period statistical characteristics, spatial topology constraints and personnel behavior patterns to reconstruct the low-dimensional data; using weighted fusion or deep feature fusion method to cross merge different feature subspace information such as time axis change rate, space edge strength and behavior density trend, so as to form a canopy fusion feature matrix with compression and representation, which is used for subsequent control model training and decision process. In the whole process, the establishment and retention of tensor structure, the explainability control of principal component, and the feature consistency maintenance in the re-fusion process constitute the key links of data processing.

[0098] Especially important is that the specific process of three-dimensional tensor splicing includes:

[0099] The time domain peak factor of the canopy time axis feature vector is advanced to obtain a time domain feature subvector; five main frequency components of the canopy time axis feature vector are extracted to obtain a frequency domain feature subvector;

[0100] The time domain feature subvector and the frequency domain feature subvector are spliced to obtain an enhanced time feature vector;

[0101] The enhanced time feature vector is graph structure coded to obtain a spatial topology feature matrix;

[0102] The personnel heat matrix is density normalized to obtain personnel normalized density data;

[0103] The personnel normalized density data, the spatial topology feature matrix and the canopy spatial topology relationship graph are matched in a coordinate system grid, and the feature dimensions are spliced to obtain a canopy three-dimensional feature space.

[0104] In the embodiment of the application, the time domain signal processing operation is performed on the constructed canopy time axis feature vector, the peak factor extraction algorithm is used to identify the sharp peak anomaly and mutation trend in the signal, and the time domain feature subvector reflecting the local fluctuation of time sequence is generated. Subsequently, the original time vector is analyzed by short-time Fourier transform (STFT), the five main frequency components with the largest energy proportion are extracted to constitute the frequency domain feature subvector, and the periodicity and oscillation characteristics of the signal in the frequency domain are described. The above two subvectors are spliced to form an enhanced time feature vector, and a graph structure coding method (such as the adjacency matrix construction idea of graph convolution network GCN) is introduced to map it to a spatial topology feature matrix to reflect its associated properties in the spatial node. At the same time, the heat matrix obtained from the personnel trajectory data is analyzed for density distribution, the maximum and minimum value normalization method is used to map the personnel density to the closed interval [0, 1], the outliers are eliminated, the comparability is improved, and the personnel normalized density data is obtained. Then, according to the predefined spatial coordinate grid standard, the personnel density data, the spatial topology feature matrix and the constructed canopy spatial topology relationship graph are rasterized and remapped and aligned in the coordinate system, and the consistent operation in the spatial dimension is completed by the interpolation method or the mean value grid registration. Finally, the multi-dimensional splicing operation (such as the tensor splicing function concat(axis=2)) is used to uniformly combine in the feature dimension to generate the canopy three-dimensional feature space tensor with time, frequency, space and personnel dynamic information, which provides high-coupling and high-expression input data for subsequent dimension reduction and modeling.

[0105] Especially important is that the three-dimensional feature re-fusion includes:

[0106] The canopy principal component dimension reduction data is subjected to 95% variance retention and variance contribution rate screening to obtain canopy principal component screening data;

[0107] The canopy main component screening data is grouped in physical meaning to obtain canopy physical meaning grouping data, wherein the physical meaning grouping includes stress / deformation correlation, temperature / humidity / CO2 correlation and density / moving speed correlation;

[0108] The canopy physical meaning grouping data is calculated for structure safety feature*environment comfort feature interaction to obtain a safety-environment coupling factor; and the canopy physical meaning grouping data is calculated for people flow dynamic feature*structure safety feature to obtain a people flow-structure interference factor;

[0109] The safety-environment coupling factor and the people flow-structure interference factor are spliced to obtain a canopy fusion feature matrix.

[0110] In the embodiment of the application, based on the canopy main component dimension reduction data after principal component analysis (PCA) dimension reduction, the variance contribution rate corresponding to each main component is counted, and a 95% cumulative variance threshold is set as a retention standard to screen out redundant dimensions, thereby obtaining canopy main component screening data with representative feature intensity, significant and high information integrity. Next, based on the source of each main component feature and the physical quantity corresponding relationship in the original data, semantic reconstruction and physical dimension calibration are carried out, and three logical groups are divided, i.e. structure stress and deformation related feature group, temperature and humidity and environment density related feature group, and people moving speed related feature group, to generate canopy physical meaning grouping data with physical interpretation. Subsequently, coupling modeling is carried out around two key dimensions of structure safety and environmental comfort, multivariate interaction analysis is carried out on the structure stress / deformation group and the temperature / humidity / density group, the weighted normalized product method or the bivariate standard deviation synergy degree calculation method is used to construct the safety-environment coupling factor, and the collaborative evolution characteristics between physical load and living conditions are reflected; and further interference analysis is carried out on the people flow dynamic feature group and the structure stress / deformation group to construct the people flow-structure interference factor, and the influence mechanism of crowd activity on structure stress stability is quantified through time-varying correlation coefficient, dynamic cross-correlation coefficient and the like. Finally, the two interaction factors are integrated into a unified tensor structure through the original tensor channel splicing method (such as splicing axis along the feature dimension) to generate a canopy fusion feature matrix with comprehensive expression ability of safety state, environmental comfort and crowd disturbance, thereby providing high-dimensional linkage input feature basis for a multi-objective decision model.

[0111] Preferably, the step S3 of constructing a decision model by using the canopy fusion feature matrix comprises:

[0112] The canopy fusion feature matrix is subjected to structure stress-deformation data extraction to obtain canopy structure stress-deformation data; and the canopy structure stress-deformation data is subjected to joint entropy calculation to obtain canopy safety risk theoretical grade data.

[0113] The canopy environment-personnel data is extracted from the canopy fusion feature matrix to obtain canopy environment-personnel data; the canopy environment-personnel data is subjected to temperature and humidity-CO2 comfort score to generate a canopy is moderate deviation value;

[0114] According to the canopy safety risk theory grade data and the canopy is moderate deviation value, perform Pareto optimization, and perform model construction, output a canopy multi-objective decision model, wherein the constraint condition of the Pareto optimization is that the structure stress is less than 40 kN and the upper limit of energy consumption is less than 3 kW.

[0115] In the embodiment of the application, the structure stress related indexes (such as stress peak value, loading frequency, deformation gradient) and the corresponding material response variables (such as maximum displacement, stress concentration area distribution density) are analyzed from the canopy fusion feature matrix, and such structure stress-deformation data is uniformly modeled as a state joint distribution function. The dependence between each stress and deformation variable is quantitatively processed by using the joint distribution function, the joint entropy (Joint Entropy) calculation method is introduced, the system uncertainty contained in different stress deformation combinations is measured by constructing a two-dimensional joint probability density function and introducing the Shannon entropy calculation formula, and then the uncertainty index value of the structure state is generated, and the safety risk theory grade score data of the canopy is obtained by constructing a mapping function in combination with the structure fatigue limit data. At the same time, the environment variables (temperature, humidity, air distribution, etc.) and the personnel state variables (distribution density, moving speed, residence time, etc.) are extracted from the fusion feature matrix to form an environment-personnel coupling data set. The data set integrates the instantaneous environment indexes and personnel activity trajectories of each measuring point in a space-time matching manner, and the comfort score of the temperature and humidity conditions of each space unit is calculated according to the ASHRAE standard or the PMV-PPD model, and the deviation calculation between the multi-point thermal comfort value and the expected value is generated, and then a spatially distributed moderate deviation quantitative index is formed to reflect the state deviation degree of the human-environment coupled system. When constructing the multi-objective decision model, the system takes the above-mentioned “canopy safety risk theory grade data” and “canopy moderate deviation value” as the double objective functions, and establishes a Pareto boundary optimization framework. Based on the non-dominated solution set, the solution space that satisfies the constraint condition “structure stress <40 kN, energy consumption <3 kW” is selected, and the non-linear optimization search of the objective function is performed. By introducing evolutionary calculation methods (such as NSGA-II) or gradient-based Pareto frontier approximation algorithms, a set of multiple local optimal solutions is found to cover possible strategy outputs under different weight combinations, and finally a dynamically updated canopy multi-objective decision model is formed based on the compromise solution mapping output. Each output in the model corresponds to one of the multiple objective compromise solutions in the control boundary, and has a traceable target constraint path.

[0116] As an example of the present application, reference is made toFigure 2 As shown, the step S4 includes:

[0117] Step S41: Obtain a canopy actuator list; search for a canopy actuator communication link based on the canopy actuator list; and issue a 5G URLLC control instruction based on the 5G network to the canopy actuator communication link to obtain canopy execution feedback data;

[0118] Step S42: Obtain theoretical execution value data; compare the canopy execution feedback data and the theoretical execution value data by error difference to obtain a canopy model residual matrix;

[0119] Step S43: Gradient back propagation fine-tuning with a learning rate of 0.001 is performed on the canopy multi-objective decision model based on the canopy model residual matrix to obtain an optimized canopy multi-objective decision model.

[0120] In the embodiment of the application, the canopy actuator list is extracted from the device management database or the actuator registry, which contains the identification information, function type and communication parameters of each actuator. Then, based on the list, a communication link search module is started, combined with the actuator network topology and the 5G communication protocol stack, to dynamically establish or confirm a low-delay and high-reliability URLLC (Ultra-Reliable Low Latency Communications) control channel, to ensure the real-time transmission and response capability of the instructions. After the control instructions are encapsulated in an encoded format, they are issued to the corresponding actuators through the 5G network. The actuators return feedback data after completing the action, which covers the execution status code, response timestamp and measurement parameters of the execution effect, forming a canopy execution feedback data set. Through the data comparison module, the pre-defined theoretical execution value data set is used as the reference standard, and the real-time acquired execution feedback data is mapped and compared according to time and actuator identification. The comparison uses an error difference algorithm to calculate the point-by-point deviation of the actual execution result from the theoretical expectation, forming a model residual matrix. Each element of the matrix represents the error amount of a specific time point and actuator, and the error types cover multi-dimensional information such as position deviation, response delay and action completion degree. The residual matrix not only reflects the deviation distribution of the execution level, but also implies the difference characteristics between the model prediction and the actual system behavior. The residual matrix is introduced into the training mechanism of the canopy multi-objective decision model as an error feedback used for the loss function part of supervised learning. By setting a gradient descent strategy with a learning rate of 0.001, the residual information is transmitted layer by layer to the model weight adjustment unit using the back propagation algorithm. This fine-tuning process refines the internal parameters of the model, improving the model's ability to adapt to dynamic changes in the actual execution environment. Gradient calculation solves the partial derivative of error items in the residual matrix to the model parameters, ensuring that the model minimizes the error while maintaining overall stability. Finally, the optimized canopy multi-objective decision model is output, completing the iterative optimization of the system-level closed loop.

[0121] Preferably, step S41 comprises the following steps:

[0122] Step S411: acquire a canopy actuator list; search for a canopy actuator communication link based on the canopy actuator list;

[0123] Step S412: based on the 5G network, issue 5G URLLC control instructions to the canopy actuator communication link, including adjusting the speed of the frequency conversion fan to adjust the ventilation volume, obtaining ventilation environment feedback data; adjusting the power of the low-altitude air conditioner to adjust the temperature value, obtaining canopy air conditioner adjustment data; changing the light intensity of the LED array to obtain canopy light adjustment data;

[0124] Step S413: superimpose the ventilation environment feedback data, the canopy air conditioner adjustment data and the canopy light adjustment data by weight to obtain the canopy execution feedback data.

[0125] In the embodiment of the application, the canopy actuator list is called from the actuator management system, which contains the identification code, type attribute and network communication parameter of each actuator in detail. Based on the list, the communication management module automatically locates and activates the 5G communication link of each actuator through network topology analysis and communication protocol stack scheduling, guarantees the low delay and high reliability of the link, and ensures that the control instructions can be accurately and timely transmitted to each actuator. 5G URLLC control instructions are issued to different types of actuators, including: sending speed adjustment instructions to the frequency conversion fan actuator to control the ventilation volume, collecting air flow rate, temperature and humidity parameters fed back by the ventilation environment sensor to form ventilation environment feedback data; sending power adjustment commands to the low-altitude air conditioner actuator, monitoring and collecting corresponding indoor temperature change data to generate air conditioner adjustment data; adjusting the light intensity of the LED array, collecting light sensor data feedback to form light adjustment data. Each feedback data contains a timestamp and a device identifier to ensure the timing consistency and traceability of multi-source data. In the data fusion module, the above three types of feedback data are processed by weight assignment and superposition. The weight distribution is dynamically adjusted based on the relative importance of the actuator to the environment and the signal-to-noise ratio of the feedback data, and the ventilation environment feedback data, the air conditioner adjustment data and the light adjustment data are fused by weighted average method to form a unified canopy execution feedback data set. The feedback data set is stored in a structured multi-dimensional data form, has timing and spatial correlation, and provides comprehensive and accurate execution effect evaluation basis for subsequent model analysis and decision optimization.

[0126] Preferably, step S5 comprises the following steps:

[0127] Step S51: generating an AR navigation instruction set based on the canopy optimization decision model;

[0128] Step S52: If the canopy safety risk theoretical grade data is greater than or equal to level 3, trigger the sound and light warning signal based on the AR navigation instruction set through the canopy actuator communication link, and generate reinforcement suggestion coordinates to obtain intelligent canopy communication optimization data;

[0129] Step S53: Construct a canopy communication whole-process monitoring report based on the intelligent canopy communication optimization data.

[0130] In the embodiment of the application, the multi-dimensional decision variable output by the canopy optimization decision model is converted into an AR navigation instruction set by using an instruction mapping algorithm, the instruction set is encoded in a structured vector form, and contains spatial coordinates, action types and timestamp information, so as to ensure the space-time consistency and traceability of the instruction. The instruction set data is encapsulated into a data packet suitable for the actuator communication link through a protocol conversion module, so as to ensure the integrity and safety in the subsequent transmission process. The canopy safety risk theoretical grade data is subjected to threshold determination, and when the grade value reaches or exceeds level 3, an alarm mechanism based on the AR navigation instruction set is triggered. The mechanism issues the instruction set to the corresponding actuator node through the canopy actuator communication link, and the actuator activates the sound and light alarm device according to the instruction, so as to realize the physical triggering of the alarm signal. At the same time, the system generates reinforcement suggestion coordinate data based on the spatial positioning algorithm and the structural stress data, and the data includes three-dimensional spatial position and suggestion operation parameters, which are archived as part of the intelligent canopy communication optimization data. The data integrates real-time monitoring data and model optimization results to form multi-dimensional alarm response information, so as to ensure the pertinence and timeliness of the instruction. At the data aggregation level, based on the intelligent canopy communication optimization data, the data integration and formatting technology is used to construct a canopy communication whole-process monitoring report including alarm event records, actuator response data and reinforcement suggestions. The report data structure includes time sequence logs, spatial distribution maps and state change matrices, supports multi-dimensional analysis and tracking. The data storage adopts a standardized database format, supports subsequent retrieval and historical backtracking analysis, and realizes closed-loop monitoring and data-driven management of the whole communication process.

[0131] Preferably, step S51 comprises the following steps:

[0132] Step S511: Obtain real-time personnel trajectory data;

[0133] Step S512: Input the real-time personnel trajectory data into the canopy optimization decision model to obtain canopy local area density data; 2 For high risk, personnel path intersection conflict point prediction is performed to obtain personnel guidance prediction data;

[0134] Step S513: Generate a turning angle based on the personnel guidance prediction data, and output a walking speed suggestion to obtain an AR navigation instruction set.

[0135] In the embodiments of the present application, the real-time trajectory data of personnel obtained by the multi-source positioning sensor and the UWB positioning tag records the two-dimensional or three-dimensional spatial coordinates of the personnel inside the tent in time series form, and a high-frequency updated personnel trajectory data set is constructed. The data set is processed by time sequence synchronization and noise filtering to ensure that the space-time accuracy of the trajectory points meets the subsequent density calculation requirements. The processed real-time personnel trajectory data is input into the density estimation algorithm module in the tent optimization decision model. Based on the kernel density estimation method, the module performs real-time statistics on the number of people per unit area in the specified local area, generating continuous density distribution data. By setting a density threshold (such as 0.5 people per square meter), dynamic identification of high-risk dense areas is achieved. Once the local area density exceeds the threshold, the system automatically starts the path intersection conflict point prediction module, which combines trajectory history data and kinematic models to calculate possible personnel path intersection points and their occurrence probabilities through trajectory clustering and collision prediction algorithms, generating predicted personnel guidance data. Based on the predicted personnel guidance data, turning angle instructions are generated using geometric calculation and motion planning techniques, and a step speed adjustment scheme suitable for the current density and path conditions is output by the walking speed suggestion model. The above turning angle and speed suggestion data are converted into structured AR navigation instruction sets through encoding. The instruction set includes spatial positioning parameters, navigation path adjustment schemes, and speed control information, supporting real-time guidance of personnel by AR devices. Throughout the process, the space-time continuity and accuracy of the data are strictly guaranteed to ensure the accuracy and timeliness of the navigation instructions.

[0136] In the present specification, an intelligent tent communication system based on 5G communication is provided for performing the above-mentioned intelligent tent communication method based on 5G communication, which comprises:

[0137] A multi-source data acquisition and heterogeneous construction module is configured to acquire environment, structure and personnel data by deploying a multi-source sensor array on the tent structure, and to perform data preprocessing and construct a tent heterogeneous raw data set;

[0138] A space-time synchronization and feature fusion module is configured to align the tent heterogeneous raw data set by time stamping to obtain a synchronized tent data set, construct a dynamic heat matrix of the tent based on the synchronized tent data set, and obtain a tent fusion feature matrix based on principal component analysis of the dynamic heat matrix of the tent;

[0139] A multi-target intelligent decision and control instruction generation module is configured to construct a decision model using the tent fusion feature matrix to obtain a tent multi-target decision model, output a device control parameter vector based on the tent multi-target decision model, and obtain a 5G URLLC control instruction by vector encoding the device control parameter vector;

[0140] The link control and model closed-loop optimization module is used to obtain a tent executor list; the tent executor communication link is searched based on the tent executor list; the 5G URLLC control instruction is issued to the tent executor communication link based on the 5G network, and tent execution feedback data is obtained; the tent execution feedback data is used to generate an optimization parameter, and the optimization parameter is returned to the tent multi-objective decision model for optimization, and a tent optimization decision model is obtained;

[0141] The augmented reality navigation and intelligent early warning module is used to generate an AR navigation instruction set based on the tent optimization decision model; an audible and light warning signal is triggered by using the AR navigation instruction set, and a reinforcement suggestion coordinate is generated, and a tent communication whole-process monitoring report is obtained.

[0142] Therefore, from any viewpoint, the embodiments should be considered as being exemplary and non-limiting, the scope of the application being defined by the appended claims and not by the above description, and all the changes falling within the meaning and the scope of the equivalent elements of the patent file are therefore intended to be comprised in the present application.

[0143] The above description is merely illustrative of the application and is not intended to limit the scope of the application. Various modifications of the embodiments will be readily apparent to persons skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A smart tent communication method based on 5G communication, characterized in that, Includes the following steps: Step S1: Deploy a multi-source sensor array on the tent structure to collect environmental, structural, and personnel data, and perform data preprocessing to construct a heterogeneous raw dataset of the tent. Step S2: Timestamp align the heterogeneous original tent dataset to obtain a synchronized tent dataset; A thermal matrix was constructed from the synchronized tent dataset to obtain the dynamic thermal matrix of the tents; Principal component analysis was performed based on the dynamic thermal matrix of the tents to obtain the tent fusion feature matrix. Principal component analysis based on the tent's dynamic thermodynamic matrix includes: The tent time axis feature vector, tent spatial topology diagram and personnel heat matrix are aligned in three dimensions and then stitched together in three dimensions to obtain the tent three-dimensional feature space. Principal component analysis was performed on the three-dimensional feature space of the tent to reduce its dimensionality, resulting in principal component dimensionality-reduced data of the tent. The principal component data of the tents were dimensionality reduced with 95% variance preserved, and the three-dimensional features were then fused to obtain the tent fusion feature matrix. Step S3: Construct a decision model using the tent fusion feature matrix to obtain a multi-objective decision model for the tent; output the equipment control parameter vector based on the multi-objective decision model for the tent; perform vector encoding on the equipment control parameter vector to obtain 5G URLLC control commands. The decision model constructed using the tent fusion feature matrix includes: Structural stress-deformation data is extracted from the tent's fusion feature matrix to obtain tent structural stress-deformation data; joint entropy is calculated from the tent structural stress-deformation data to obtain tent safety risk theoretical level data; Environmental-personnel data is extracted from the tent's fusion feature matrix to obtain tent environment-personnel data; this data is then further analyzed for temperature and humidity. The comfort rating for the generated tent is a moderate deviation value. Based on the theoretical level data of tent safety risks and the moderate deviation value of tent, Pareto optimization is performed and a model is built to output a multi-objective decision model for tent. The constraint condition of Pareto optimization is that the structural stress is less than 40kN and the energy consumption limit is less than 3kW. Step S4: Obtain the tent actuator list; search for tent actuator communication links based on the tent actuator list; send 5G URLLC control commands to the tent actuator communication links based on the 5G network to obtain tent execution feedback data; generate optimization parameters from the tent execution feedback data and send them back to the tent multi-objective decision model for optimization to obtain the tent optimization decision model. Step S5: Generate an AR navigation instruction set based on the tent optimization decision model; trigger an audible and visual alarm signal using the AR navigation instruction set, and generate reinforcement suggestion coordinates to obtain a full-process monitoring report of tent communication.

2. The intelligent tent communication method based on 5G communication according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Deploy piezoelectric stress sensors at key nodes of the tent support structure, with a range of 0-50kN, a sampling frequency of 20Hz, and an accuracy of 0.5%, to collect time-series data of stress on the support structure. Step S12: Deploy a micro-deformation fiber optic sensor on the surface of the tarpaulin with a resolution of 0.01 mm and a sampling frequency of 10 Hz to collect tarpaulin deformation gradient data; Step S13: Evenly distribute temperature and humidity within the interior space. A multi-parameter probe with a temperature accuracy of ±0.3℃ and a sampling frequency of 1Hz is used to collect environmental condition data. Step S14: Embed a UWB positioning tag in the personnel's safety helmet, with a positioning accuracy of 10cm and a refresh rate of 5Hz, to generate personnel coordinate trajectory data; Step S15: Use wavelet threshold denoising algorithm to filter the support stress time series data, tent deformation gradient data, environmental state data and personnel coordinate trajectory data to achieve a signal-to-noise ratio (SNR) greater than 30dB, and compensate for baseline drift with a drift rate of less than 0.03% / month to obtain the tent heterogeneous original dataset.

3. The intelligent tent communication method based on 5G communication according to claim 1, characterized in that, Step S2 involves constructing a heat matrix for the synchronized tent dataset, including: Temporal features were extracted from the synchronized tent dataset, and timeline labeling was performed to obtain the tent timeline feature vector. Spatial relationship topology features are constructed from the synchronized tent dataset, and topology edges are initialized to obtain a tent spatial topology graph. A population trajectory density distribution analysis was performed on the synchronized tent dataset, and low-density data was removed to obtain a population heat map matrix.

4. The intelligent tent communication method based on 5G communication according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Obtain the tent actuator list; search for the tent actuator communication link based on the tent actuator list; send 5G URLLC control commands to the tent actuator communication link based on the 5G network to obtain tent execution feedback data; Step S42: Obtain theoretical execution value data; compare the tent execution feedback data with the theoretical execution value data using error difference analysis to obtain the tent model residual matrix; Step S43: Based on the tent model residual matrix, perform gradient backpropagation fine-tuning on the tent multi-objective decision model with a learning rate of 0.001 to obtain the tent optimization decision model.

5. The intelligent tent communication method based on 5G communication according to claim 4, characterized in that, Step S41 includes the following steps: Step S411: Obtain the tent actuator list; search for tent actuator communication links based on the tent actuator list; Step S412: Send 5G URLLC control commands to the tent actuator communication link based on the 5G network, including adjusting the ventilation volume of the variable frequency fan to obtain ventilation environment feedback data; adjusting the temperature value of the low air conditioning compressor to obtain tent air conditioning adjustment data; and changing the light intensity of the LED array to obtain tent lighting adjustment data. Step S413: Weight and superimpose the ventilation environment feedback data, tent air conditioning adjustment data, and tent lighting adjustment data to obtain tent execution feedback data.

6. The intelligent tent communication method based on 5G communication according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Generate an AR navigation instruction set based on the tent optimization decision model; Step S52: If the theoretical safety risk level of the tent is greater than or equal to level 3, trigger an audible and visual alarm signal through the tent actuator communication link based on the AR navigation instruction set, and generate reinforcement suggestion coordinates to obtain intelligent tent communication optimization data. Step S53: Construct a full-process monitoring report for tent communication based on intelligent tent communication optimization data.

7. The intelligent tent communication method based on 5G communication according to claim 6, characterized in that, Step S51 includes the following steps: Step S511: Obtain real-time personnel trajectory data; Step S512: Input real-time personnel trajectory data into the tent optimization decision model to calculate local area density, obtaining local area density data for the tent. When the local area density data is greater than 0.5 people / If the risk is high, then the intersection and conflict points of personnel paths are predicted to obtain personnel guidance prediction data; Step S513: Generate a turning angle based on the predicted data of the person being guided, and output a walking speed suggestion to obtain the AR navigation instruction set.

8. A smart tent communication system based on 5G communication, characterized in that, For executing the 5G-based smart tent communication method as described in claim 1, the 5G-based smart tent communication system comprises: The multi-source data acquisition and heterogeneous construction module is used to collect environmental, structural and personnel data by deploying a multi-source sensor array on the tent structure, and to perform data preprocessing to construct a heterogeneous raw dataset for the tent. The spatiotemporal synchronization and feature fusion module is used to align the timestamps of the heterogeneous original tent datasets to obtain a synchronized tent dataset; to construct a heat matrix from the synchronized tent dataset to obtain a dynamic heat matrix of the tents; and to perform principal component analysis based on the dynamic heat matrix of the tents to obtain a fused feature matrix of the tents. The multi-objective intelligent decision-making and control command generation module is used to construct a decision model using the tent fusion feature matrix to obtain a tent multi-objective decision model; output the equipment control parameter vector based on the tent multi-objective decision model; and perform vector encoding on the equipment control parameter vector to obtain 5G URLLC control commands. The execution link control and model closed-loop optimization module is used to obtain the tent actuator list; search for the tent actuator communication link based on the tent actuator list; send 5G URLLC control commands to the tent actuator communication link based on the 5G network to obtain tent execution feedback data; generate optimization parameters from the tent execution feedback data and send them back to the tent multi-objective decision model for optimization to obtain the tent optimization decision model. The augmented reality navigation and intelligent early warning module is used to generate AR navigation command sets based on the tent optimization decision model; it uses the AR navigation command sets to trigger audible and visual alarm signals and generate reinforcement suggested coordinates to obtain a full-process monitoring report of tent communication.

Images