Multi-modal data driven latent disaster intelligent perception and emergency data engineering system

Abstract

The application discloses a multi-modal data-driven potential disaster intelligent perception and emergency data engineering system and relates to the technical field of intelligent information processing.The system comprises a metasurface communication module, a cloud collaborative module and a rescue decision module.The metasurface communication module uses a disaster risk probability graph and terrain height data to calculate the phase gradient between communication nodes, generates a beamforming matrix, performs low-power non-line-of-sight data transmission, monitors link quality and feeds back communication energy consumption, and generates encrypted data packets.The cloud collaborative module decrypts the encrypted data packets, performs three-dimensional disaster situation deduction, and generates a disaster influence heat map.The rescue decision module uses an improved path planning algorithm to perform path optimization on the disaster influence heat map, generates an optimized rescue path, and, in combination with communication energy consumption data and backhaul terrain parameters, generates updated communication power and redundancy configuration parameters.The application greatly improves the reliability and energy utilization efficiency of emergency communication, ensures the stability of communication links in key areas, and has redundancy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent information processing, and in particular to a multi-modal data driven potential disaster intelligent perception and emergency data engineering system. BACKGROUND

[0002] In recent years, with the development of sensing technology, communication technology and high-performance computing, the real-time monitoring and emergency response capability for natural disasters has been significantly improved. The emergence of multi-modal heterogeneous data acquisition and fusion technology enables the integration of multi-source data from satellite remote sensing, ground sensor networks, unmanned aerial vehicle platforms and weather radars under the same spatio-temporal reference, thereby providing a richer information base for disaster prediction and risk assessment. At the same time, geomechanical modeling and high-performance numerical solving methods are constantly improving, making it possible to simulate disaster evolution based on physical mechanisms; emerging metasurface communication and beamforming technology provides a new means for high-reliability, low-power data transmission in complex terrain and harsh environments. In addition, the development of cloud collaborative computing and intelligent decision-making algorithms provides a higher level of automation and intelligence for the whole-link information processing and command and dispatch of disaster emergency.

[0003] However, in the existing technical system, the spatio-temporal alignment of multi-modal data and the energy consumption constraint calculation of multi-source physical models, as well as the adaptive optimization of communication links in dynamic disaster environments, are often implemented independently, lacking real-time closed-loop collaboration under the same architecture. For example, in disaster simulation calculation, the dynamic adaptation ability of physical modeling and numerical solving to energy consumption is insufficient, which may lead to a decrease in calculation efficiency in energy-limited environments; in terms of communication, although beamforming technology can optimize link quality, parameter adjustment often fails to comprehensively optimize in combination with disaster risk spatial distribution and energy state. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a multi-modal data driven potential disaster intelligent perception and emergency data engineering system to solve the problems of insufficient collaboration between multiple links and dynamic optimization of energy consumption.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] The application provides a multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, which comprises a data processing module, a multi-modal heterogeneous data acquisition module, a time-space alignment processing module, a standardized time-space matrix generation module, an energy management module, an energy state real-time monitoring module, a calculation mode selection module, a physical neural operator accelerator driving module, an energy control instruction output module, a physical calculation module, a three-dimensional geomechanics equation utilization module, a standardized time-space matrix analysis modeling module, a gridization module, an equation discretization module, an accelerated solution module, a mechanical field distribution extraction module, a instability probability calculation module, a disaster risk probability graph generation module, an ultra-surface communication module, a communication node phase gradient calculation module, a beamforming matrix generation module, a low-power non-line-of-sight data transmission module, a link quality monitoring module, a communication energy consumption feedback module, an encrypted data packet generation module, a cloud-side collaborative module, a three-dimensional disaster situation deduction module, a disaster impact heat map generation module, a rescue decision module, an improved path planning algorithm utilization module, an optimized rescue path generation module, a communication power and redundancy configuration parameter generation module, and an updated communication power and redundancy configuration parameter generation module.

[0008] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the standardized time-space matrix generation module comprises the following specific steps,

[0009] The multi-modal heterogeneous data is time-synchronized by using a dynamic time warping algorithm, and a time-aligned data sequence is outputted; and the time-aligned data sequence is converted into spatial coordinates by combining RTK positioning information, and time-space aligned data is generated.

[0010] The time-space aligned data is subjected to format standardization processing, and a standardized time-space matrix is formed.

[0011] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the energy control instruction output module comprises the following specific steps,

[0012] According to the time and position synchronization information of the standardized time-space matrix, the output power and super capacitor power are real-time collected by using an energy monitoring sensor, and are subjected to preprocessing to form an energy state parameter;

[0013] According to the energy state parameter, the current energy sufficiency degree and task calculation demand are determined, a calculation mode selection logic is determined, a matched calculation mode is determined, and the selected calculation mode is coded into a calculation mode signal;

[0014] The calculation mode signal is converted into a corresponding energy control instruction, and is sent to the physical neural operator accelerator for driving operation, and the energy control instruction is outputted.

[0015] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the disaster risk probability graph is generated, and the specific steps are as follows,

[0016] The standardized space-time matrix and energy control instruction are analyzed, the space-time field variable and solving configuration are extracted, the three-dimensional calculation grid is mapped and updated, and the discrete grid model is generated through spatial interpolation assignment;

[0017] The initial mechanical condition and boundary condition are applied to the discrete grid model, the coupling equation, time step and solver parameter are set according to the solving configuration, and the discretized equation group is formed;

[0018] The discretized equation group is loaded into the physical neural operator accelerator for numerical solving, and the mechanical field distribution is extracted;

[0019] The mechanical field distribution is calculated and spatially mapped to generate a disaster risk probability graph.

[0020] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the standardized space-time matrix and energy control instruction are analyzed, the space-time field variable and solving configuration are extracted, the three-dimensional calculation grid is mapped and updated, and the discrete grid model is generated through spatial interpolation assignment, and the specific steps are as follows,

[0021] The standardized space-time matrix and energy control instruction are analyzed, the space-time field variable and solving configuration are extracted, the three-dimensional calculation grid is mapped and updated, and the discrete grid model is generated through spatial interpolation assignment, and the specific steps are as follows,

[0022] According to the three-dimensional calculation region and solving configuration, an initial grid structure is generated, and a three-dimensional calculation grid is output;

[0023] The space-time field variable is mapped to the three-dimensional calculation grid, and the grid density is adjusted according to the disaster risk probability graph, the mechanical field distribution and the energy state parameter of the last calculation period, and an updated three-dimensional calculation grid is output;

[0024] The updated three-dimensional calculation grid is subjected to spatial interpolation, and the space-time field variable is assigned to the grid nodes to generate a discrete grid model.

[0025] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the beamforming matrix is generated, and the specific steps are as follows,

[0026] The terrain height data in the existing digital terrain model database is used to form a terrain risk data set in combination with the disaster risk probability graph;

[0027] The spatial coordinate difference and height difference between the communication nodes are calculated using the terrain risk data set to obtain the distance and height difference matrix between the nodes.

[0028] The phase gradient of the communication node is calculated by using the distance and height difference matrix between nodes, and is mapped to the array element to generate a phase modulation value, and is combined to form a beamforming matrix.

[0029] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the encrypted data packet is generated, and the specific steps are as follows,

[0030] Data transmission is performed by using the beamforming matrix and the disaster risk probability graph to obtain a transmission state.

[0031] According to the transmission state, the communication link quality is evaluated and the communication energy consumption is calculated to generate communication energy consumption data.

[0032] The communication energy consumption data is used to encrypt the disaster risk probability graph to generate an encrypted data packet.

[0033] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the disaster impact thermal map is generated, and the specific steps are as follows,

[0034] The encrypted data packet is decrypted to output a decrypted disaster risk probability graph.

[0035] The decrypted disaster risk probability graph is mapped to a discrete grid model node, the initial value of the mechanical field and the spatial boundary constraint condition are initialized, three-dimensional disaster deduction is performed, and preliminary disaster simulation data is generated.

[0036] The preliminary disaster simulation data and the meteorological and hydrological observation data measured and collected by the field sensor are fused and processed, three-dimensional disaster deduction is performed again, and a disaster impact thermal map is generated.

[0037] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the optimized rescue path is generated, and the specific steps are as follows,

[0038] The risk level and obstacle information of the path planning area are extracted from the disaster impact thermal map to generate path planning input data.

[0039] According to the path planning input data, the initial parameters and heuristic functions of the improved path planning algorithm are initialized, path search is performed, candidate path cost is evaluated, path selection is optimized, and a preliminary rescue path is generated.

[0040] The preliminary rescue path is smoothed and risk adjusted to generate an optimized rescue path.

[0041] As a preferred scheme of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, the specific steps of generating the updated communication power and redundancy configuration parameters are as follows,

[0042] The high-risk area coordinates are extracted from the optimized rescue path, the beamforming matrix control parameter is triggered for correction, and the corrected communication parameters are generated.

[0043] The communication link quality indicators are monitored and collected, the link power consumption and energy consumption are calculated in combination with the corrected communication parameters, the energy consumption feedback information is generated, and the comprehensive energy state data is generated after fusion processing.

[0044] The adaptive optimization algorithm is used to analyze the comprehensive energy state data, dynamically adjust the communication power distribution ratio and the redundancy setting of the beamforming matrix, and generate the updated communication power and redundancy configuration parameters.

[0045] The beneficial effects of the present application are: by triggering the beamforming matrix control parameter correction in real time, and dynamically adjusting the communication power and redundancy in combination with the link quality monitoring and adaptive optimization algorithm, a closed-loop optimization mechanism of disaster situation awareness, task demand and communication resource management is formed, the reliability and energy utilization efficiency of emergency communication are greatly improved, the communication link in the key area is stable and has redundancy, and resources are saved in the low-risk area, which is suitable for emergency rescue scenes with limited energy and complex communication environment. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0047] Fig. 1 The schematic diagram of the multi-modal data-driven potential disaster intelligent perception and emergency data engineering system in the present application.

[0048] Fig. 2 The flowchart of generating the disaster risk probability map in the present application.

[0049] Fig. 3 The flowchart of generating the beamforming matrix in the present application.

[0050] Fig. 4 The flowchart of generating the disaster influence heat map in the present application. DETAILED DESCRIPTION

[0051] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0052] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. The present application, however, can be practiced in a variety of ways beyond the specific embodiments described herein without departing from the scope of the present application, and it is understood that variations can be made in view of what is described herein, by individuals skilled in the art, without departing from the spirit and scope of the present application.

[0053] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate or alternative to other embodiments.

[0054] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a multi-modal data-driven potential disaster intelligent perception and emergency data engineering system, comprising the following steps:

[0055] The data processing module collects multi-modal heterogeneous data and performs spatio-temporal alignment processing to generate a standardized spatio-temporal matrix.

[0056] The dynamic time warping algorithm is used to synchronize the time of the multi-modal heterogeneous data, output a time-aligned data sequence, and convert the time-aligned data sequence into spatial coordinates in combination with the RTK positioning information to generate spatio-temporal aligned data.

[0057] Specifically, the multi-modal heterogeneous data includes meteorological sensor measurement data, hydrological sensor measurement data, geological sensor measurement data, and satellite remote sensing data.

[0058] The dynamic time warping algorithm is used to compare the collected multi-modal heterogeneous data in sequence according to the time stamp, perform nonlinear stretching or compression on the time series of different data sources, align the features corresponding to each time point, and output a time-aligned data sequence.

[0059] The time-aligned data sequence is matched with the longitude, latitude and elevation coordinates in the RTK positioning information, the data at each time step is mapped to the corresponding spatial coordinates to form spatio-temporal aligned data, which contains time information and spatial coordinate information, for example, the temperature, humidity and vibration data collected by different sensors in the same second are respectively mapped to the same geographic location coordinates to realize unified representation of time and space.

[0060] The spatio-temporal aligned data is subjected to format standardization processing to form a standardized spatio-temporal matrix.

[0061] Specifically, the spatio-temporal alignment data is standardized, including unit unification, value normalization and missing value filling of the feature data of each time step, arranging different data types of features in column order, and generating a unified data table structure.

[0062] The time information, spatial coordinates and various feature values are organized into a matrix form, with the rows of the matrix corresponding to the time steps and the columns corresponding to the spatial positions and feature channels. For example, temperature, humidity, vibration and displacement data are arranged in a fixed order into the matrix columns to form a standardized spatio-temporal matrix.

[0063] The energy management module monitors the energy state in real time according to the standardized spatio-temporal matrix, selects a calculation mode and drives the physical neural operator accelerator to output energy control instructions.

[0064] According to the time and position synchronization information of the standardized spatio-temporal matrix, the output power and supercapacitor capacity are collected in real time by the energy monitoring sensor and preprocessed to form energy state parameters.

[0065] Specifically, according to the time and position synchronization information of the standardized spatio-temporal matrix, the output power and supercapacitor capacity are collected in real time by the energy monitoring sensor, and the collected voltage, current and capacity signals are subjected to noise filtering and signal smoothing processing, such as moving average method or weighted filtering method for processing of continuous sampling data. The power and capacity at each time are normalized according to the sampling time, the units and value ranges are unified, the processed power values and capacity values are arranged according to the time steps and spatial positions, and the energy state parameters including the output power and supercapacitor capacity of each time step are formed.

[0066] According to the energy state parameters, the current energy sufficiency and task calculation demand are determined, the matching calculation mode is determined by the mode selection logic, and the selected calculation mode is encoded into a calculation mode signal.

[0067] Specifically, the energy state parameters of each time step are compared with the preset output power threshold and supercapacitor capacity threshold to determine whether the current energy meets the load and energy storage requirements, thereby generating an energy sufficiency classification result.

[0068] According to the task calculation demand, the calculation task type and the corresponding calculation load are matched, for example, high-load tasks are matched with high energy state, and low-load tasks are matched with medium-low energy state, and the matching calculation mode of each time step is determined by the mode selection logic. The selected calculation mode is converted into a calculation mode signal according to a preset encoding rule, such as binary or multi-bit encoding method, and a calculation mode signal sequence corresponding to each time step is output.

[0069] It should also be noted that the specific steps of the output power threshold and the super capacitor power threshold: according to the task calculation demand and the energy supply capacity, the output power thresholds corresponding to the high, medium and low energy states are preset, for example, the high energy state threshold is 80%, and the low energy state threshold is 20%; At the same time, according to the super capacitor charging and discharging characteristics and the continuous energy supply time requirement, the super capacitor power thresholds corresponding to the high, medium and low energy states are preset, for example, the high energy state threshold is 80%, and the low energy state threshold is 20%; In each time step, the real-time collected output power and super capacitor power are compared with the preset super capacitor power threshold, and it is judged whether the current energy state belongs to high, medium or low energy state, so as to provide basis for subsequent calculation mode selection.

[0070] The preset encoding rule establishes a mapping relationship between each calculation mode and the corresponding encoding value according to the calculation mode classification in the standardized space-time matrix and the task calculation demand; The encoding value can be expressed in binary or decimal, for example, the high-performance calculation mode is encoded as "11", the energy-saving mode is encoded as "01", and the ordinary calculation mode is encoded as "10"; The encoding rule is derived from the analysis of the functional characteristics of the calculation mode and the energy management strategy, so as to ensure that the encoding can uniquely identify each calculation mode and adapt to the energy control instruction transmission and analysis requirements.

[0071] The calculation mode signal is converted into corresponding energy control instruction, and sent to the physical neural operator accelerator to drive operation, and the energy control instruction is output.

[0072] Specifically, the calculation mode signal is analyzed according to the preset encoding rule, and is mapped into the parameter value corresponding to the energy control instruction, for example, the encoding "11" is converted into the high-performance calculation mode instruction, and the encoding "01" is converted into the energy-saving mode instruction; The mapped energy control instruction is sent to the physical neural operator accelerator through the control interface of the physical neural operator accelerator to drive operation, and the physical neural operator accelerator adjusts the calculation resource allocation and energy use strategy according to the received energy control instruction, completes the operation control, and outputs the energy control instruction.

[0073] It should also be noted that the physical neural operator accelerator adopts existing high-performance computing hardware and programmable accelerator technology, combines neural network computing optimization scheme, realizes fast processing of complex numerical operation and equation solving through hardware level parallel computing and instruction set optimization, realizes relying on existing GPU, FPGA or ASIC accelerator platform, using mature hardware interface and control protocol to drive operation, without building non-public computing device, ensuring that the energy control instruction can be efficiently executed and the calculation result is output.

[0074] The physical computing module uses three-dimensional geomechanics equations to perform analytical modeling, gridding, equation discretization and accelerated solving on the standardized space-time matrix under the constraint of the energy control instruction, extracts the mechanical field distribution and calculates the instability probability, and generates a disaster risk probability map.

[0075] It should be noted that the standardized space-time matrix of multi-modal heterogeneous data is closely combined with the energy control instruction, analytical modeling, gridding and equation discretization solving are realized based on three-dimensional geomechanics equations, and mechanical field distribution and instability probability are efficiently extracted through accelerated calculation, thereby generating a refined disaster risk probability map, realizing real-time closed-loop combination of data-driven physical modeling and disaster risk quantification, and significantly improving the accuracy and calculation efficiency of disaster risk prediction.

[0076] The beneficial effects are that by coupling the standardized space-time matrix of multi-modal heterogeneous data with the energy control instruction, analytical modeling, gridding and discretization solving are performed using three-dimensional geomechanics equations, which can extract the mechanical field distribution and calculate the instability probability in real time and high precision, thereby generating a disaster risk probability map. Unlike the disaster risk assessment in the prior art which usually relies on a single data source or static empirical model, the deep integration of multi-source data-driven, physical modeling and accelerated calculation is realized, which not only improves the accuracy and spatial resolution of disaster risk prediction, but also significantly reduces the calculation delay, realizes the real-time and dynamic adjustability of disaster risk quantification, and embodies the innovation in disaster intelligent perception and emergency data processing.

[0077] The standardized space-time matrix and the energy control instruction are analyzed, the space-time field variables and solving configuration are extracted, the three-dimensional calculation grid is mapped and updated, and the discrete grid model is generated through spatial interpolation assignment.

[0078] Further, the standardized space-time matrix and the energy control instruction are analyzed, the space-time field variables and calculation requirement parameters are extracted, and the space-time field variables and solving configuration are output.

[0079] Specifically, the time information, position coordinate information, meteorological sensor measurement data, hydrological sensor measurement data and geological sensor measurement data in the standardized space-time matrix are read, and the time information, position coordinate information, meteorological sensor measurement data, hydrological sensor measurement data and geological sensor measurement data are indexed and matched according to the space-time coordinate correspondence relationship; the energy control instruction is analyzed, the calculation mode code and energy constraint information are extracted, and the calculation resource and energy demand required by each spatial node are calculated by combining the task calculation requirement data in the standardized space-time matrix; the solving configuration is generated according to the calculation requirement parameters and the node position, including the time step, the equation type, the boundary condition and the grid division scheme; the space-time field variables and the solving configuration corresponding to each spatial node are output.

[0080] An initial mesh structure is generated according to the three-dimensional calculation region and the solving configuration, and a three-dimensional calculation mesh is output.

[0081] Specifically, the voxel blocks of the three-dimensional calculation region are divided according to the spatial boundary information of the three-dimensional calculation region, the block size, and the time step, the equation coupling parameter, and the solver setting in the solving configuration, the grid nodes uniformly distributed along the X-axis, the Y-axis, and the Z-axis are generated, the connection relationship between the nodes is established to form the initial mesh structure, and the initial mesh structure is output as the three-dimensional calculation mesh.

[0082] The spatiotemporal field variables are mapped to the three-dimensional calculation mesh, the grid density is adjusted according to the disaster risk probability map, the mechanical field distribution, and the energy state parameter of the previous calculation period, and the updated three-dimensional calculation mesh is output.

[0083] Specifically, the spatiotemporal field variables in the standardized spatiotemporal matrix are subjected to trilinear interpolation in space according to the node coordinates of the three-dimensional calculation mesh, and the interpolation results are written into the grid node attributes of the three-dimensional calculation mesh.

[0084] The risk value of the disaster risk probability map at the center, the stress or displacement value of the mechanical field distribution at the grid node, and the energy state parameter corresponding to the output power and the super capacitor power of the time step are obtained for each grid node of the three-dimensional calculation mesh, and the normalized value of the energy state parameter is calculated.

[0085] The risk value, the mechanical field gradient amplitude, and the normalized value of the energy state parameter are used to determine the local refinement or coarsening operation, for example, when the risk value exceeds the example high-risk standard or the mechanical field gradient amplitude exceeds the example high-gradient standard and the energy state parameter normalized value exceeds the example high-energy standard, it is marked for refinement, and when the risk value is lower than the example low-risk standard and the mechanical field gradient amplitude is lower than the example low-gradient standard or the energy state parameter normalized value is lower than the example low-energy standard, it is marked for coarsening.

[0086] Local octant or binary refinement is performed on the cubic block marked for refinement, new grid nodes are inserted, and the node connection relationship is updated, adjacent cubic blocks marked for coarsening and meeting the merging conditions are merged and redundant nodes are removed, and adjacent cubic blocks are adjusted for level consistency to maintain smooth transition of the grid resolution; finally, the updated three-dimensional calculation mesh is output.

[0087] The updated three-dimensional calculation mesh is subjected to spatial interpolation, and the spatiotemporal field variables are assigned to the grid nodes to generate a discrete grid model.

[0088] Specifically, according to the node coordinates of the updated three-dimensional calculation grid, the space interpolation processing is performed on the space-time field variables in the standardized space-time matrix, and the sensor measurement data at each time step and position is mapped to the corresponding grid node.

[0089] The existing three-dimensional interpolation method, such as trilinear interpolation or Kriging interpolation, is used to calculate the field variable values at the grid nodes, and the interpolated displacement, stress, temperature or measurement parameter is assigned to each grid node to generate a discrete grid model.

[0090] The initial mechanical conditions and boundary conditions are applied to the discrete grid model, and the coupling equation, time step and solver parameters are set according to the solving configuration to form a discrete equation system.

[0091] Specifically, according to the grid node information of the discrete grid model, the initial mechanical condition parameters and boundary condition parameters in the standardized space-time matrix are read, and the initial displacement, velocity, stress and temperature are assigned to the grid nodes of the discrete grid model, and the displacement constraint, force or pressure boundary condition is applied to the boundary nodes of the discrete grid model.

[0092] According to the equation type, time step and solver parameters in the solving configuration, the discretized three-dimensional geomechanical partial differential equation is assembled according to the grid node number, and combined with the initial conditions and boundary conditions, the mechanical coupling relationship between nodes is established, the combination of stiffness matrix, mass matrix and force vector is formed, the coupling equation is obtained, and the discrete equation system is output for numerical solution.

[0093] It should be noted that the solving configuration is derived from the standard numerical calculation method for solving three-dimensional geomechanical equations and engineering experience parameters, including the type of discretization method, time step setting, solver selection, boundary conditions and initial conditions parameters of the coupling equation, and the calculation accuracy and stability requirements obtained through experimental verification in similar geomechanical calculation tasks, which are used to ensure that the numerical solution of the discrete equation system can meet the accuracy and efficiency of disaster risk probability calculation.

[0094] The discrete equation system is loaded into the physical neural operator accelerator for numerical solution, and the mechanical field distribution is extracted.

[0095] Specifically, the discrete equation system is converted into a numerical input format recognizable by the physical neural operator accelerator according to the solver parameters and time step requirements, and the discrete equation system is loaded into the calculation program of the physical neural operator accelerator. The numerical iterative solution is performed in the physical neural operator accelerator according to the calculation mode and parallel operation strategy, and the space-time field variables of the grid nodes are periodically updated until the solving convergence condition is reached. The mechanical field distribution of each grid node, such as node displacement, stress and strain value, is output.

[0096] It should be noted that in the physical neural operator accelerator, when solving the discretized equation system iteratively, the change of the mechanical field distribution of each grid node in the three-dimensional calculation grid is calculated continuously to determine whether the change meets the preset convergence threshold. If the change of the mechanical field distribution of all grid nodes meets the convergence threshold, the solution is considered complete; if the change of the mechanical field distribution of some grid nodes does not meet the convergence threshold, the mechanical field distribution of the grid nodes is updated according to the time step and the next iteration is performed, and the process is repeated until the change of the mechanical field distribution of all grid nodes meets the convergence requirement, and finally the converged extraction mechanical field distribution is output.

[0097] The preset convergence threshold is determined according to the mechanical solution accuracy, the stability of the discretized equation, and the reliability of the disaster risk prediction, for example, the maximum allowed difference of the displacement or stress change of the grid node; specifically, the accuracy requirement is determined according to the resolution of the three-dimensional calculation grid and the time step, the convergence threshold is set for the sensitive area in combination with the disaster risk probability map, and the change of the mechanical field distribution of each grid node is compared with the convergence threshold in real time during the iterative solution process to determine the convergence.

[0098] The instability probability of the mechanical field distribution is calculated and spatially mapped to generate a disaster risk probability map.

[0099] Specifically, the mechanical field distribution obtained by the discretized equation system is analyzed node by node, and the node mechanical feature vector is formed according to the stress, strain and displacement data of the grid node; the node mechanical feature vector is compared with the preset instability judgment condition one by one, the instability probability is calculated node by node using the over-limit mapping method, and the instability probability of the grid node is arranged according to the node number to form a node instability probability matrix.

[0100] The node instability probability matrix is mapped to the spatial position coordinates of the three-dimensional calculation grid, a continuously distributed disaster risk probability map is generated through interpolation processing, and the disaster risk probability map is spatially smoothed and visually output to form a disaster risk probability map.

[0101] It should be noted that the preset instability judgment condition is determined according to existing geomechanics theory, three-dimensional geomechanics equation solving experience and disaster historical observation data, including the strength limit of rock and soil materials, friction angle, cohesion and parameters of typical instability modes, combined with the characteristics of the mechanical field distribution and the stress-strain critical value, obtained by field survey and experimental determination method, forming a set of judgment conditions that can be used to calculate the instability probability of the grid node.

[0102] The metasurface communication module calculates the phase gradient between communication nodes using the disaster risk probability map and terrain height data, generates a beamforming matrix, performs low-power non-line-of-sight data transmission, monitors link quality and feeds back communication energy consumption, and generates encrypted data packets.

[0103] The terrain height data in the existing digital terrain model database is adopted, and the terrain risk data set is formed by combining the disaster risk probability map.

[0104] Specifically, the terrain height data in the existing digital terrain model database is extracted, the terrain height information of the corresponding region is screened according to the spatial range of the disaster risk probability map, the terrain height of each grid node is matched with the disaster risk probability of the corresponding node, the terrain risk data set is generated according to the position of the grid node, for example, the terrain height value and the disaster risk probability value are formed into a two-dimensional table or a three-dimensional array for storage, and finally the terrain risk data set which can be used for subsequent communication node phase gradient calculation is formed.

[0105] The spatial coordinate difference and the height difference between the communication nodes are calculated by using the terrain risk data set, and the distance and height difference matrix between the nodes are obtained.

[0106] Specifically, according to the spatial coordinates and the terrain height of each communication node in the terrain risk data set, any two communication nodes are taken in turn and , the horizontal distance and the height difference between the nodes are calculated, the horizontal distance calculation expression is:

[0107]

[0108] Among them, represents the horizontal distance between the communication node and the communication node , the value of represents the value of the spatial coordinates of the communication node in the direction of the axis, the value of represents the value of the spatial coordinates of the communication node in the direction of the axis, the value of represents the value of the spatial coordinates of the communication node in the direction of the axis, the value of represents the value of the spatial coordinates of the communication node in the direction of the axis, the value of represents the number of the first communication node participating in the calculation, the value of represents the number of the second communication node participating in the calculation, the value range of and is , wherein is the total number of communication nodes;

[0109] The height difference calculation expression is:

[0110]

[0111] wherein, represents a communication node and a communication node , represents a communication node , represents a communication node ;

[0112] All communication node combinations are calculated to obtain a distance and height difference matrix between nodes .

[0113] The phase gradient of the communication node is calculated using the distance and height difference matrix between nodes, and is mapped to the array element to generate the phase modulation value, and is combined to form the beamforming matrix.

[0114] Specifically, the target communication link pair is selected one by one according to the link identifier, and the link geometry information sequence is generated based on the horizontal distance and height difference of each link pair; with the array reference element as the zero phase reference, the phase gradient increment along the horizontal and vertical directions of the array is calculated for each link according to the link geometry information sequence, and the phase gradient sequence is obtained;

[0115] According to the array element number, the phase gradient sequence is mapped to each array element, the phase increment is accumulated and phase normalization and quantization (such as discretization according to the preset quantization bit width) are performed, and the phase modulation value sequence of the array element is obtained;

[0116] Finally, according to the two-dimensional index corresponding to the array element number and the link number, the array element phase modulation value corresponding to each link is combined by column, and the beamforming matrix is output.

[0117] Data transmission is performed using the beamforming matrix and the disaster risk probability map to obtain the transmission state.

[0118] Specifically, the disaster risk probability map is sliced and packed according to the grid index of the disaster risk probability map to generate disaster risk probability map data slices; lossless compression and packet encapsulation are performed on the disaster risk probability map data slices, and a timestamp and a spatial coordinate label are attached to form a to-be-sent data stream;

[0119] The beamforming matrix is loaded at the sending end, the baseband symbol stream corresponding to the to-be-sent data stream is precoded according to the beamforming matrix, and mapping is completed according to the transmit array radio frequency channel;

[0120] Channel coding and modulation (such as LDPC coding and QPSK modulation) are used to generate radio frequency signals, which are transmitted through a wireless link; timing and carrier synchronization, demodulation and decoding are completed at the receiving end, and the disaster risk probability map data slices are recovered according to the sequence number and CRC check of the disaster risk probability map data slices and the receiving result is recorded;

[0121] The effective frame count, frame loss count, retransmission times, average round-trip delay, and instantaneous throughput of the sending end and the receiving end are counted, and the beam number and transmission power record of the beamforming matrix are combined to generate a transmission state.

[0122] According to the transmission state, the communication link quality is evaluated and the communication energy consumption is calculated to generate communication energy consumption data.

[0123] Specifically, the effective frame count, frame loss count, retransmission times, average round-trip delay, instantaneous throughput, beam number, and transmission power in the transmission state are read and analyzed; the frame success rate is calculated according to the effective frame count and the frame loss count, the communication link quality level is evaluated in combination with the average round-trip delay and the instantaneous throughput, and the link performance difference is recorded by beam number during the evaluation process;

[0124] The total transmission energy is calculated using the transmission power, retransmission times, and frame transmission time, the reception energy is calculated in combination with the receiving end working time and the receiving power consumption, the transmission energy and the reception energy are accumulated to obtain the communication energy consumption; the communication link quality level and the communication energy consumption are associated to generate communication energy consumption data containing time stamp, spatial coordinates, link quality level, and communication energy consumption value.

[0125] Using the communication energy consumption data, the disaster risk probability map is encrypted to generate an encrypted data packet.

[0126] Specifically, the communication energy consumption data and the disaster risk probability map are read, the available encryption overhead is evaluated according to the communication energy consumption data, and the symmetric encryption algorithm and key length are selected;

[0127] The session key is derived by the existing key derivation function (such as HKDF) according to the preset master key and the time stamp and spatial coordinate information of the disaster risk probability map, and the initialization vector is generated by the true random number generator;

[0128] The disaster risk probability map is divided into fragments, and each fragment is encrypted using the selected symmetric encryption algorithm, the derived session key, and the initialization vector, and a message authentication code is generated and attached with a time stamp and a spatial coordinate label;

[0129] The encrypted fragments, encryption algorithm identifier, derived session key parameter, time stamp, and spatial coordinate label are encapsulated in a message format to output an encrypted data packet.

[0130] It should be noted that the preset master key is derived from a high-strength random number sequence generated by a key management center that has passed national commercial password detection and authentication, a 256-bit binary key is generated by a hardware true random number generator, is securely distributed by the key management center and is written into a secure storage area for encryption processing through a secure channel conforming to the national SM4 encryption algorithm, is loaded and solidified before deployment, and ensures the uniqueness and unpredictability of the preset master key during the entire use cycle.

[0131] The cloud collaboration module decrypts the encrypted data packet and performs three-dimensional disaster scenario deduction to generate a disaster impact heat map.

[0132] The encrypted data packet is decrypted to output a decrypted disaster risk probability map.

[0133] Specifically, the message header and payload of the encrypted data packet are read, and the encryption algorithm identifier, session key derivation parameter, initialization vector, message authentication code, data slice, timestamp and spatial coordinate label are parsed out;

[0134] Based on the preset master key and the derived session key parameter, the HKDF is used to derive the session key; then, based on the parsed initialization vector and the derived session key, the encryption algorithm (such as AES-GCM or ChaCha20-Poly1305) indicated in the message header is used to perform message authentication code verification and decryption operation on each data slice, and the integrity and sequence number (such as CRC check or sequence number comparison) of the decrypted slice are verified;

[0135] All verified decrypted slices are reorganized according to the timestamp and spatial coordinate label and combined into a complete disaster risk probability map data stream according to the original slice order;

[0136] The combined disaster risk probability map data stream is losslessly decompressed (if compressed), and the decrypted disaster risk probability map is output.

[0137] The decrypted disaster risk probability map is mapped to the discrete grid model nodes, the initial value of the mechanical field and the spatial boundary constraint condition are initialized, and three-dimensional disaster scenario deduction is performed to generate preliminary disaster simulation data.

[0138] Specifically, the decrypted disaster risk probability map is registered with the spatial coordinates of the discrete grid model nodes, and the risk value of each discrete grid model node is calculated based on trilinear spatial interpolation and written into the node attribute;

[0139] According to the node attribute, each discrete grid model node is given a mechanical field initial value, including node displacement initial value, stress initial value and velocity initial value;

[0140] Identify the boundary nodes of the discrete grid model and impose spatial boundary constraints according to the solving configuration, such as fixed displacement boundary, force / pressure boundary or seepage boundary parameters; according to the solving configuration, use explicit or implicit time integration scheme to carry out three-dimensional disaster deduction on the discretized equation set, update the mechanical field of the discrete grid model nodes according to time step and carry out necessary numerical stability and convergence check;

[0141] Organize and export the field variables of the discrete grid model nodes at each time step according to spatial grid points and time sequence to generate preliminary disaster simulation data.

[0142] Fuse the preliminary disaster simulation data with the meteorological and hydrological observation data measured by the field sensors, and carry out three-dimensional disaster deduction again to generate disaster impact heat map.

[0143] Specifically, read the preliminary disaster simulation data and the meteorological and hydrological observation data measured by the field sensors, and perform spatial and temporal interpolation registration of the meteorological and hydrological observation data measured by the field sensors to the discrete grid model nodes according to time stamp and spatial coordinates;

[0144] Calculate the node-by-node deviation of the observation value and the preliminary disaster simulation data at the registered nodes, and use existing data assimilation methods (such as least squares incremental correction or ensemble Kalman filter) to apply the deviation as a correction term to the mechanical field initial value and boundary conditions of the discrete grid model nodes, and output the corrected discrete grid initial value and boundary conditions;

[0145] Based on the corrected discrete grid initial value and the solving configuration, perform three-dimensional disaster deduction according to the time integration scheme specified by the solving configuration, update the grid node field variables step by step until the simulation period is completed, and export the time series field data of the three-dimensional disaster deduction;

[0146] According to the influence index calculation rules, perform spatial and temporal aggregation and rasterization processing on the time series field data output by the three-dimensional disaster deduction, and perform spatial interpolation and necessary smoothing processing on the rasterized influence value, and finally generate the disaster impact heat map.

[0147] It should be noted that the source of the influence index calculation rules is the published disaster assessment industry standards, meteorological and hydrological monitoring technical specifications, and historical disaster statistical analysis results, and the calculation factors and weight distribution are determined combined with the typical physical characteristics of the target disaster type, for example, according to the national or industry disaster risk assessment standards to determine rainfall intensity, runoff, slope stability coefficient, etc. as evaluation factors, and reference the influence range and loss degree caused by the same disaster in the historical disaster database to determine the weight and range of each evaluation factor, forming a quantifiable and repeatable influence index calculation rule.

[0148] The rescue decision module uses an improved path planning algorithm to optimize the disaster influence thermal map, generates an optimized rescue path, and combines communication energy consumption data and backhaul terrain parameters to generate updated communication power and redundancy configuration parameters.

[0149] It should be noted that the multi-dimensional information fusion of the disaster influence thermal map, communication energy consumption data and backhaul terrain parameters, the dynamic optimization of the rescue path is realized by using the improved path planning algorithm, and the communication power distribution and beamforming redundancy are dynamically adjusted to realize the closed-loop collaborative optimization of the rescue action and communication support.

[0150] The beneficial effects are that by integrating the disaster influence thermal map, communication energy consumption data and backhaul terrain parameters in the rescue decision module, the unified optimization of rescue path optimization and communication resource collaborative scheduling is realized. Unlike the existing technology which usually relies on only single geographic information or disaster risk information for path planning, the communication power distribution and beamforming redundancy can be dynamically adjusted to keep the communication link in the key area stable in the high-risk area, while saving energy in the low-risk area, thereby forming a closed-loop optimization mechanism of disaster situation awareness, path planning and communication management. The efficiency and communication reliability of emergency rescue are significantly improved, which is suitable for disaster scenarios with limited energy and complex communication environment, and embodies the innovation and practicality of the integration of path planning and communication support.

[0151] The risk level and obstacle information of the path planning area are extracted from the disaster influence thermal map to generate path planning input data.

[0152] Specifically, the risk value of each grid in the disaster influence thermal map is read, and the risk value is classified according to the preset risk level interval to generate the risk level of the path planning area. At the same time, the grid with obstacles is determined according to the comparison of the risk value and the obstacle judgment threshold, and the spatial coordinates and attribute information of the obstacles are recorded. The risk level and obstacle information are organized according to the grid order to form the path planning input data.

[0153] It should also be noted that the preset risk level is divided into different levels according to the geological disaster, meteorological and hydrological historical data statistics and disaster influence index calculation rules, such as low, medium and high levels, to reflect different degrees of potential danger; the obstacle judgment threshold is determined according to the field sensor measurement data and historical obstacle distribution characteristics, for example, by analyzing the obstacle probability distribution, the obstacle judgment threshold is set to an example value of 0.7 to ensure that the path planning input data can accurately reflect the actual obstacle distribution situation.

[0154] According to the path planning input data, the initial parameters and heuristic function of the improved path planning algorithm are initialized, the path search is performed and the candidate path cost is evaluated, the path selection is optimized, and the preliminary rescue path is generated.

[0155] Specifically, the risk level and obstacle information of each grid in the path planning input data are read, the improved path planning algorithm is initialized according to the heuristic function and initial parameters, the start point and the end point grids are added to the search queue, the path cost of the candidate path in the queue is calculated in turn, including the path length, the risk level cumulative value and the obstacle avoidance cost, the optimal candidate path is selected according to the cost sorting, and the search queue and the path state are updated, and the search is repeated until the end point is reachable, and finally the preliminary rescue path sorted by the optimal cost is output.

[0156] For example, the example path length weight in the heuristic function is 0.5, the risk level weight is 0.3, and the obstacle avoidance weight is 0.2.

[0157] The heuristic function is derived from the classical heuristic search method in the field of artificial intelligence path planning, such as Manhattan distance, Euclidean distance or weighted distance function used in A* algorithm, which is improved in combination with the risk level and obstacle information in the path planning input data to guide the search to preferentially explore areas with high safety and low path cost, and to realize effective path search.

[0158] The preliminary rescue path is smoothed and adjusted in risk to generate an optimized rescue path.

[0159] Specifically, the node coordinates of the preliminary rescue path are read, the line segments between adjacent nodes are smoothed using a curve fitting method, the nodes in the curve located in the high-risk area or close to the obstacle are adjusted according to the path risk level and obstacle judgment threshold in the path planning input data, the node coordinates are recalculated and the path order is updated to form the optimized rescue path.

[0160] The high-risk area coordinates are extracted from the optimized rescue path to trigger the correction of the beamforming matrix control parameters to generate corrected communication parameters.

[0161] Specifically, the node coordinates of the optimized rescue path are read, the nodes located in the high-risk area are selected and the spatial coordinates are extracted, the phase modulation value and amplitude distribution of the beamforming matrix are adjusted according to the high-risk area coordinates, the control parameters of the array elements are recalculated, and the corrected communication parameters are generated.

[0162] The communication link quality indicators are monitored and collected, the link power consumption and energy consumption are calculated in combination with the corrected communication parameters, the energy consumption feedback information is generated, and the comprehensive energy state data is generated after fusion processing.

[0163] Specifically, the communication link quality indicators are collected and parameters such as signal strength, signal-to-noise ratio and bit error rate are recorded, the instantaneous power consumption and cumulative energy consumption of each communication link are calculated according to the modified communication parameters, the energy consumption feedback information is generated by integrating the power consumption and energy consumption of each link, the energy consumption feedback information is fused with the communication link quality indicators, and the comprehensive energy state data is obtained.

[0164] The adaptive optimization algorithm is used to analyze the comprehensive energy state data, dynamically adjust the communication power distribution ratio and the redundancy setting of the beamforming matrix, and generate updated communication power and redundancy configuration parameters.

[0165] Specifically, the comprehensive energy state data is read and normalized, the communication link power consumption, energy consumption and signal quality indicators are taken as input features, the adaptive optimization algorithm is used to calculate the adjustment amount of the communication power distribution ratio and the optimization value of the beamforming matrix redundancy, and the communication power distribution ratio and the beamforming matrix redundancy setting are updated according to the optimization results, and finally the updated communication power and redundancy configuration parameters are generated for subsequent communication control.

[0166] It should be noted that the optimized rescue path and the updated communication power and redundancy configuration parameters are obtained by multi-modal fusion processing of the disaster influence thermal map, communication energy consumption data and backhaul terrain parameters, which realizes intelligent sensing and emergency data engineering operation in potential disaster environment, and the communication power and redundancy parameters are used as feedback signals to dynamically adjust the rescue path planning and communication support strategy.

[0167] In summary, the present application realizes the closed-loop optimization mechanism of disaster situation awareness, task demand and communication resource management by real-time triggering of beamforming matrix control parameter modification, and dynamically adjusting the communication power and redundancy in combination with link quality monitoring and adaptive optimization algorithm, greatly improves the reliability and energy utilization efficiency of emergency communication, ensures the stability and redundancy of communication link in critical areas, and saves resources in low-risk areas, and is suitable for emergency rescue scenes with limited energy and complex communication environment.

[0168] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalent, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A multimodal data-driven intelligent perception and emergency data engineering system for potential disasters, characterized in that: include, The data processing module collects multimodal heterogeneous data and performs spatiotemporal alignment processing to generate a standardized spatiotemporal matrix; The energy management module monitors the energy status in real time based on the standardized spatiotemporal matrix, selects the calculation mode, drives the physical neural operator accelerator, and outputs energy control commands. The physical calculation module uses three-dimensional geomechanical equations to perform analytical modeling, gridding, equation discretization and accelerated solution of the standardized spatiotemporal matrix under the constraint of energy control commands. It extracts the mechanical field distribution and calculates the instability probability to generate a disaster risk probability map. The metasurface communication module uses disaster risk probability maps and terrain height data to calculate the phase gradient between communication nodes, generate a beamforming matrix, perform low-power non-line-of-sight data transmission, monitor link quality and feedback communication energy consumption, and generate encrypted data packets. The cloud-based collaborative module decrypts encrypted data packets and performs 3D disaster simulation to generate a heat map of disaster impact. The rescue decision-making module uses an improved path planning algorithm to optimize the disaster impact heat map, generate an optimized rescue path, and combine it with communication energy consumption data and backed terrain parameters to generate updated communication power and redundancy configuration parameters.

2. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the standardized spatiotemporal matrix are as follows: The dynamic time warping algorithm is used to synchronize the time of multimodal heterogeneous data, outputting a time-aligned data sequence. Combined with RTK positioning information, the time-aligned data sequence is converted into spatial coordinates to generate spatiotemporally aligned data. The spatiotemporal aligned data is format-standardized to form a standardized spatiotemporal matrix.

3. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for the output energy control command are as follows: Based on the time and location synchronization information of the standardized spatiotemporal matrix, the output power and supercapacitor charge are collected in real time by energy monitoring sensors and preprocessed to form energy state parameters. Based on the energy state parameters, the current energy sufficiency and task computation requirements are determined. The mode selection logic is used to determine the matching computation mode, and the selected computation mode is encoded into a computation mode signal. The computation mode signal is converted into a corresponding energy control command, which is sent to the physical neural operator accelerator to drive its operation and output the energy control command.

4. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the disaster risk probability map are as follows: The standardized spatiotemporal matrix and energy control commands are analyzed to extract spatiotemporal field variables and solution configurations, map and update the three-dimensional computational grid, and generate a discrete grid model through spatial interpolation. Initial mechanical and boundary conditions are applied to the discrete mesh model, and the coupling equations, time step and solver parameters are set according to the solution configuration to form a discretized set of equations. The discretized equations are loaded onto a physical neural operator accelerator for numerical solution to extract the mechanical field distribution. The probability of instability of the mechanical field distribution is calculated and spatially mapped to generate a disaster risk probability map.

5. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 4, characterized in that: The standardized spatiotemporal matrix and energy control commands are analyzed to extract spatiotemporal field variables and solution configurations. The three-dimensional computational grid is mapped and updated, and a discrete grid model is generated through spatial interpolation. The specific steps are as follows. The standardized spatiotemporal matrix and energy control commands are analyzed to extract spatiotemporal field variables and calculation requirements parameters, and the spatiotemporal field variables and solution configuration are output. An initial mesh structure is generated based on the 3D computational domain and the solution configuration, and the 3D computational mesh is output. The spatiotemporal field variables are mapped to a three-dimensional computational grid, and the grid density is adjusted based on the disaster risk probability map, mechanical field distribution and energy state parameters of the previous calculation cycle, and the updated three-dimensional computational grid is output. Spatial interpolation is performed on the updated 3D computational grid to assign spatiotemporal field variables to the grid nodes, generating a discrete grid model.

6. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the beamforming matrix are as follows: A terrain risk dataset is formed by using terrain elevation data from existing digital terrain model databases and combining it with disaster risk probability maps; Using the terrain risk dataset, the spatial coordinate differences and height differences between communication nodes are calculated to obtain the distance and height difference matrix between nodes; The phase gradient of the communication node is calculated using the distance and height difference matrix between nodes, and then mapped to the array elements to generate phase modulation values, which are combined to form a beamforming matrix.

7. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the encrypted data packet are as follows: Data transmission status is obtained by using beamforming matrix and disaster risk probability map; Based on the transmission status, assess the quality of the communication link and calculate the communication energy consumption to generate communication energy consumption data; By using communication energy consumption data, the disaster risk probability map is encrypted to generate encrypted data packets.

8. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the disaster impact heat map are as follows. Decrypt the encrypted data packet and output the decrypted disaster risk probability map; The decrypted disaster risk probability map is mapped to the nodes of the discrete grid model, the initial values ​​of the mechanical field and the spatial boundary constraints are initialized, and a three-dimensional disaster scenario is simulated to generate preliminary disaster simulation data. The preliminary disaster simulation data is fused with meteorological and hydrological observation data collected through on-site sensors, and a three-dimensional disaster simulation is performed again to generate a disaster impact heat map.

9. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the optimized rescue route are as follows: Risk levels and obstacle information for the route planning area are extracted from the disaster impact heat map to generate route planning input data; Based on the path planning input data, initialize the initial parameters and heuristic function of the improved path planning algorithm, perform path search and evaluate the cost of candidate paths, optimize path selection, and generate a preliminary rescue path; The initial rescue route is smoothed and risk-adjusted to generate an optimized rescue route.

10. The multimodal data-driven intelligent perception and emergency data engineering system for potential disasters as described in claim 1, characterized in that: The specific steps for generating the updated communication power and redundancy configuration parameters are as follows: The coordinates of high-risk areas are extracted from the optimized rescue route, triggering the correction of the beamforming matrix control parameters and generating the corrected communication parameters. Monitor and collect communication link quality indicators, calculate link power consumption and energy consumption by combining the corrected communication parameters, generate energy consumption feedback information, and perform fusion processing to generate comprehensive energy status data; An adaptive optimization algorithm is used to analyze the comprehensive energy state data, dynamically adjust the communication power allocation ratio and the redundancy setting of the beamforming matrix, and generate updated communication power and redundancy configuration parameters.

Images