Method and system for evaluating vehicle networking meteorological observation layout based on data assimilation

Abstract

This invention relates to the field of intelligent transportation and vehicle-to-everything (V2X) data processing technology, and discloses a method and system for evaluating the layout of V2X meteorological observation based on data assimilation. The key technical points are: a data cleaning module acquires multi-source hardware sensing data and performs physical interference stripping based on a nonlinear compensation model; a spatial mapping module constructs spatial Gaussian attenuation weights and performs coordinate mapping from discrete point trajectories to a static topological grid; a simulation error module schedules the hardware accelerator to run the simulation environment and deduces the dynamic observation error covariance based on an LSTM network; an assimilation evaluation module first applies the consistent Kalman filtering mechanism to perform physical state assimilation of the meteorological background field and sensor data, then quantifies the information gain scalar of the virtual deployment nodes, and finally configures a multi-objective constraint cost function to solve for the optimal sensor network hardware topology scheme; and an instruction execution module extracts local meteorological slices coupled with traffic flow features, generates physical prevention and control early warning instructions, and issues them for execution.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and vehicle-to-everything (V2X) data processing technology, and more specifically, to a method and system for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation. Background Technology

[0002] With the rapid development of intelligent transportation and vehicle networking technologies, using connected vehicles as mobile meteorological detection nodes to make up for the shortcomings of traditional fixed meteorological stations, such as high construction costs and low spatial resolution, has become an industry trend. However, in actual engineering applications, vehicle-mounted meteorological sensing components are inevitably affected by complex physical disturbances such as engine exhaust heat radiation and vehicle dynamics. Furthermore, there is a serious mismatch between the discrete vehicle movement trajectory points and the static topological mesh required by the fluid dynamics meteorological model in the spatial coordinate system.

[0003] More importantly, existing technologies struggle to accurately extrapolate high-frequency dynamic observation errors of vehicle-mounted hardware, lack mechanisms for assimilating and integrating heterogeneous sensor data from vehicle networks with meteorological background fields, and cannot quantitatively assess the information gain from deploying sensor hardware at different road network nodes in eliminating system uncertainties. This leads to a bottleneck in engineering where the optimal topological balance between the economic cost of hardware deployment and the effectiveness of meteorological sensing cannot be achieved, thus failing to provide high-precision data support for physical traffic control in severe weather on high-risk road sections (such as closed-loop control of throttle and braking systems).

[0004] Therefore, the present invention provides a method and system for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation, thereby improving the aforementioned technical problems. Summary of the Invention

[0005] This disclosure aims to address the shortcomings of existing technologies by providing a method and system for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation. The present invention employs nonlinear physical interference stripping and spatial topology mapping to process heterogeneous sensing data from the V2X network. It combines long short-term memory networks to extrapolate dynamic errors and applies consistent Kalman filtering for meteorological state assimilation. Furthermore, it utilizes information entropy to quantify node gains and constructs a multi-objective cost function to solve for the optimal hardware layout topology that balances economic cost and sensing efficiency. Finally, it extracts local meteorological slices and couples them with traffic flow to generate closed-loop control commands for driving the physical actuators of vehicles.

[0006] To achieve the above objectives, the present disclosure proposes the following technical solutions:

[0007] In a first aspect, this disclosure proposes a method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation, the method comprising the following steps: The system receives data packets reported by the mobile sensor network through the communication interface, extracts the mechanical dynamic characteristics recorded by the controller local area network bus on the node vehicle and the raw meteorological sensing signals collected by the on-board meteorological sensing components; calls the pre-stored nonlinear compensation model to perform compensation calculations and outputs the meteorological observation benchmark value. The electronic map database is read, and the physical road network is divided into multiple static topological grids. The real-time GPS coordinates of the vehicle at the node are extracted, the physical Euclidean distance between the coordinates and the center of the corresponding static topological grid is calculated, and a Gaussian decay weight function is constructed based on the principle of spatial heterogeneity to map the meteorological observation benchmark value to the corresponding static grid space, generating a grid fusion observation column vector. The cloud computing resources are scheduled to run the simulation test model of the observation system and generate a real atmospheric reference field. The long short-term memory network inference unit is called to use the mechanical dynamic features and the grid fused observation column vector as input to deduce the dynamic scaling weight. The dynamic observation error covariance matrix is ​​deduced by combining the quadratic transformation. Simulation physical noise is injected into the real atmospheric reference field to generate the state vector of the simulation observation field. Read the biased meteorological background field generated based on different parameterization schemes, substitute it into the dynamic observation error covariance matrix to solve the physical state gain matrix; use the physical state gain matrix to perform real-time correction on the meteorological background field, and output the global meteorological analysis field and the posterior error covariance matrix; In the road network topology model, a candidate deployment node is virtually configured. By comparing the change in the determinant of the posterior error covariance matrix before and after the introduction of the candidate deployment node, the information gain quantization scalar of the candidate deployment node is calculated. Under the multi-objective constraints of comprehensively considering the procurement cost of individual hardware and the benefits of meteorological sensing accuracy, a system cost function is constructed, and the genetic optimization engine is started to perform crossover and mutation iteration operations of binary sequences to output the optimal hardware topology layout vector. Based on the optimal hardware topology layout vector, the local meteorological feature column vector of the target risk road segment is extracted from the global meteorological analysis field; the local traffic flow physical state column vector extracted by the hardware supporting the target risk road segment is read, and a matrix concatenation operation is performed to generate a risk assessment feature tensor; the collision risk probability is evaluated using a Logit classifier, and a control message is generated when the collision risk probability is greater than a preset safety threshold to drive the on-board electronic control unit of the vehicle at that node to perform a physical closed-loop operation to adjust the throttle intake volume or adjust the braking intervention threshold.

[0008] As a preferred technical solution of the present invention, the calculation formula for performing compensation calculations and outputting meteorological observation benchmark values ​​is as follows:

[0009] ;

[0010] ;

[0011] in, This parameter represents the current sampling timestamp of the system. Indicates the first in the road network Hardware index of each node vehicle; This represents the first hardware calibration sparse diagonal matrix configured to compensate for thermal radiation bias. This represents the second hardware calibration matrix configured to compensate for speed-related disturbances; This represents the third hardware calibration matrix configured to compensate for longitudinal inertial disturbances; This represents the fourth hardware calibration matrix configured to compensate for lateral attitude disturbances; The column vector representing the dynamic features of vehicle i at time t; This indicates the real-time operating temperature of the engine coolant, obtained from the CAN bus. This represents the original meteorological perception column vector collected by the vehicle-mounted meteorological sensor component; This indicates the base reference temperature of the external environment; Representing vectors The real-time driving speed scalar in the data; Representing vectors The longitudinal acceleration scalar in; Representing vectors The scalar of the yaw angle in the middle; This represents the preset static system deviation compensation column vector; This represents the meteorological observation baseline column vector output after noise reduction and calibration. This represents the multidimensional system interference deviation column vector of the calculated output.

[0012] As a preferred embodiment of the present invention, the calculation formula for generating the grid-fused observation column vector is as follows:

[0013] ;

[0014] ;

[0015] in, This represents a static topological grid index with a total of M grids. Indicates node vehicle coordinates and grid The Euclidean distance scalar between the centers; L represents the physical length constant of the spatially dependent influences preset by the system; This represents the exponential operator with the natural constant e as its base; Indicates node vehicle Corresponding to grid The mapping weight scalar; This represents the grid fusion observation column vector generated after coordinate reconstruction.

[0016] As a preferred embodiment of the present invention, the calculation formulas for deriving the dynamic observation error covariance matrix and generating the simulated observation field state vector are as follows:

[0017] ;

[0018] ;

[0019] ;

[0020] in, This represents an algebraic operator that expands the elements of a vector into a diagonal matrix along the main diagonal. This represents the logistic regression activation operator, used to map the real number field to the interval (0,1); and These represent the pre-set weight matrix and bias vector parameters within the LSTM model, respectively. This represents the hidden state feature vector cached by the LSTM inference unit at the previous time step. Let represent the dynamic observation error covariance matrix of vehicle i at output node after quadratic transformation; This represents the dynamically scaled weight diagonal matrix of the output; This represents the static observation error covariance matrix preset by the meteorological hardware at the factory. Representation matrix The transpose of the matrix; This represents the state vector of the simulated observation field generated after noise fusion; This represents the observation operator matrix configured to perform interpolation transformation from grid state to observation space; This represents the column vector of the true atmospheric reference state generated by the system; This represents the column vector of simulated random physical noise generated by the system.

[0021] As a preferred embodiment of the present invention, the formulas for calculating the output global meteorological analysis field and the posterior error covariance matrix are as follows:

[0022] ;

[0023] ;

[0024] ;

[0025] in, This represents the physical state gain matrix obtained by the assimilation evaluation module. This represents the prior background error covariance matrix corresponding to the meteorological background field; Represents the observation operator matrix The transpose of the matrix; This represents the total number of entity nodes. The combined generated global dynamic observation error covariance diagonal matrix; This represents the global meteorological analysis field vector output after assimilation. This represents the generated initial fluid meteorological background field state vector; This indicates that the posterior error covariance matrix is ​​updated and written to memory after the assimilation calculation is completed. This indicates the identity matrix operator configured to participate in the operation.

[0026] As a preferred embodiment of the present invention, the calculation formula for the information gain quantization scalar is as follows:

[0027] ;

[0028] in, Represents physical nodes Quantize the information gain scalar generated by the configuration sensing hardware; This represents the index of a candidate hardware deployment node in the system's road network topology model; Represents the operator for the natural logarithm; This indicates an operator that performs matrix determinant calculations to obtain the scalar volume of the matrix; Indicates the assumption of adding a node The specific posterior error covariance matrix is ​​calculated after observing the data.

[0029] As a preferred embodiment of the present invention, the formula for calculating the optimal hardware topology layout vector is as follows:

[0030] ;

[0031] in, This represents the optimal hardware topology layout vector output after the optimization engine iteratively converges; This represents the deployment state vector of the hardware entities generated through system iteration; This represents an operator for optimizing the independent variable to minimize the value of the objective function. This represents the pre-configured root mean square error term penalty weight constant; This represents the root mean square error constant between the assimilated meteorological field and the real reference field under the current deployment scheme S; This represents the pre-configured information gain term penalty weight constant; A constant representing the total number of candidate deployment nodes; Represents the first state in the state vector S. Boolean elements; This represents the pre-configured penalty weight constant for the engineering economic cost item; Indicates at node The unit physical economic cost constant for installing hardware equipment.

[0032] As a preferred embodiment of the present invention, the formula for calculating the probability of collision risk is as follows:

[0033] ;

[0034] ;

[0035] ;

[0036] in, This indicates the preset target risk road segment index; Indicates based on road section Spatial feature extraction operator matrix for geographic boundary configuration; This represents the global meteorological analysis field vector output after assimilation based on the optimal network topology; Indicates the extracted target road segment The column vector of local microclimate characteristics; Indicates road segment Local traffic flow state column vector; This represents the risk assessment tensor generated by the cascading of meteorological and traffic flow characteristics along the travel dimension; Represents the weight row vector of the classification model; A scalar representing the bias offset of the classification model; This represents a scalar value indicating the probability of collision risk on the target road segment.

[0037] Secondly, this disclosure proposes a vehicle-to-everything (V2X) meteorological observation layout evaluation system based on data assimilation. The system includes: a cloud computing cluster, a roadside communication architecture, and a mobile sensor network. The cloud computing cluster integrates: The data cleaning module is used to receive data packets reported by the mobile sensor network through the communication interface, extract the mechanical dynamic characteristics recorded by the controller local area network bus of the node vehicle and the raw meteorological sensing signals collected by the vehicle-mounted meteorological sensor components; call the pre-stored nonlinear compensation model to perform compensation calculations and output meteorological observation benchmark values. The spatial mapping module is used to read the electronic map database, divide the physical road network into multiple static topological grids, extract the real-time GPS coordinates of the vehicle at the node, calculate the physical Euclidean distance between the coordinates and the center of the corresponding static topological grid, and construct a Gaussian decay weight function based on the principle of spatial heterogeneity to map the meteorological observation benchmark value to the corresponding static grid space, generating a grid fusion observation column vector. The simulation error module is used to schedule high-performance computing resources in the cloud to run the simulation test model of the observation system and generate a real atmospheric reference field. It calls the long short-term memory network inference unit, uses the mechanical dynamic features and the grid fused observation column vector as input to deduce the dynamic scaling weight, combines the quadratic transformation to deduce the dynamic observation error covariance matrix, and injects simulation physical noise into the real atmospheric reference field to generate the state vector of the simulation observation field. The assimilation evaluation module is used to read the biased meteorological background field generated based on different parameterization schemes, substitute it into the dynamic observation error covariance matrix to solve the physical state gain matrix, use the physical state gain matrix to perform real-time correction on the meteorological background field, and output the global meteorological analysis field and the posterior error covariance matrix; it is also used to virtually configure candidate deployment nodes in the road network topology model, and calculate the information gain quantization scalar of the candidate deployment node by comparing the change in the determinant of the posterior error covariance matrix before and after the introduction of the candidate deployment node; and it is used to construct the system cost function under the multi-objective constraints of comprehensively considering the unit hardware procurement cost and the meteorological sensing accuracy gain, start the genetic optimization engine to perform crossover and mutation iteration operations of binary sequences, and output the optimal hardware topology layout vector. The instruction execution module is used to extract the local meteorological feature column vector of the target risk road segment from the global meteorological analysis field based on the optimal hardware topology layout vector; read the local traffic flow physical state column vector extracted by the hardware supporting the target risk road segment, and perform matrix concatenation operation to generate a risk assessment feature tensor; use the Logit classifier to evaluate the collision risk probability, and generate a control message when the collision risk probability is greater than a preset safety threshold, so as to drive the on-board electronic control unit of the vehicle at the node to perform a physical closed-loop operation to adjust the throttle intake volume or adjust the braking intervention threshold.

[0038] Thirdly, embodiments of this disclosure propose a computer-readable storage medium storing a computer program thereon, which, when executed by a computing device, enables the computing device to implement a vehicle-to-everything (V2X) meteorological observation layout evaluation method based on data assimilation.

[0039] In summary, the present invention has the following beneficial effects:

[0040] Firstly, by constructing a multi-dimensional nonlinear compensation model, this invention matches corresponding sparse correction matrices for the thermal dimension and the dynamic disturbance dimension, effectively removing the systematic pollution of the vehicle's weather sensing components caused by engine waste heat radiation and mechanical movements such as vehicle acceleration, deceleration, and yaw. This significantly improves the baseline accuracy and reliability of heterogeneous sensing data in the Internet of Vehicles from the underlying physical signal level.

[0041] Secondly, this invention introduces the principle of spatial heterogeneity and the Gaussian decay weight function to smoothly map discrete moving vehicle trajectory points to the static topological grid required by the fluid dynamics model, effectively solving the problem of mismatch between the dynamic "point" observation vector of the Internet of Vehicles and the static "surface" grid field of the meteorological numerical model in terms of spatial coordinate system and physical scale.

[0042] Third, this invention breaks through the limitations of traditional meteorological assimilation relying on static constant errors. It relies on the Long Short-Term Memory (LSTM) network to extrapolate the high-frequency dynamic reliability of vehicle hardware, and combines quadratic transformation to ensure the symmetric positive semidefiniteness of the error covariance matrix. Then, it applies the consistent Kalman filtering mechanism to achieve high-fidelity fusion of vehicle network dynamic perception data and meteorological background field, which greatly improves the assimilation analysis accuracy of local complex microclimates.

[0043] Fourth, this invention uses the entropy reduction mechanism in information theory to quantitatively evaluate the information gain of virtual nodes and constructs a multi-objective constraint cost function that integrates hardware procurement costs, communication overhead and perception accuracy gains. The optimal hardware topology layout vector is accurately solved through a genetic iteration engine. Under the premise of maximizing the efficiency of road network meteorological perception, the invention effectively avoids the blind stacking of hardware and greatly reduces the economic cost of traffic engineering construction and equipment networking.

[0044] Fifth, this invention establishes a closed-loop physical execution mechanism from meteorological perception to traffic intervention. By extracting local meteorological features of the target road segment and coupling them with the actual traffic flow status, it uses a classification model to assess collision risks in real time. When the safety threshold is exceeded, it directly generates control commands conforming to the V2X protocol and sends them to the vehicle electronic control unit (ECU). This drives the vehicle to perform mechanical operations such as adjusting the throttle intake or the hydraulic threshold of the anti-lock braking system, thus realizing a high-level active traffic physical defense and control guidance for severe weather. Attached Figure Description

[0045] Figure 1 A framework diagram of a vehicle-to-everything (V2X) meteorological observation layout evaluation system based on data assimilation provided in an embodiment of the present invention;

[0046] Figure 2 A flowchart of a vehicle-to-everything (V2X) meteorological observation layout evaluation method based on data assimilation provided in an embodiment of the present invention. Detailed Implementation

[0047] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0050] Furthermore, the technical features involved in the various embodiments of this application described below can be combined with each other as long as they do not conflict with each other.

[0051] This disclosure aims to address the problems of high deployment cost and low spatial resolution of fixed meteorological stations in existing meteorological observation technologies, as well as severe dynamic interference of vehicle-mounted detection nodes and the mismatch between their movement trajectories and the spatial scale of the static meteorological grid, making it difficult to quantitatively evaluate the effectiveness of node deployment. In view of this, this disclosure proposes a vehicle-to-everything (V2X) meteorological observation deployment evaluation method and system based on data assimilation, for evaluating and optimizing the deployment scheme of meteorological detection nodes in a V2X environment. This method employs data assimilation and information entropy analysis techniques, using dynamic observation error modeling and differentiated information gain evaluation factors to address the physical heterogeneity and spatiotemporal complexity of the vehicle-mounted sensing environment. This achieves the goal of effectively reducing system deployment costs and enhancing traffic physical control effectiveness while improving the accuracy of local weather forecasts.

[0052] Please refer to Figure 1 , Figure 1 The framework of the vehicle-to-everything (V2X) meteorological observation layout evaluation system based on data assimilation, as described in this disclosure embodiment, is shown. In terms of physical hardware architecture, it includes: a cloud computing cluster, a roadside communication architecture, and a mobile sensor network.

[0053] The cloud computing cluster is equipped with high-performance processors and hardware acceleration units; the roadside communication architecture includes multiple roadside computing units (RSUs) distributed at key nodes of the road network; the mobile sensor network includes on-board units (OBUs), on-board weather sensors, and on-board electronic control units (ECUs) installed on multiple node vehicles.

[0054] At the functional logic level, the cloud computing cluster integrates a data cleaning module, a spatial mapping module, a simulation error module, an assimilation evaluation module, and an instruction execution module. The vehicle terminal uploads the underlying physical sensing signals to the cloud computing cluster via C-V2X or 5G links; the cloud computing cluster, based on the physical control messages generated after processing by each functional module, sends them to the vehicle ECUs of the corresponding road sections via the RSU.

[0055] Please refer to Figure 2 , Figure 2 A flowchart of the vehicle-to-everything (V2X) meteorological observation layout evaluation method based on data assimilation, as described in an embodiment of this disclosure, is shown. The overall process mainly includes the following seven steps:

[0056] S1: The data cleaning module acquires multi-source hardware-sensing data and performs physical interference stripping based on a nonlinear compensation model.

[0057] In practical implementation, the data cleaning module receives data packets reported by the mobile sensor network through the V2X communication interface and extracts the mechanical dynamic characteristics recorded by the vehicle controller local area network (CAN) bus and the raw meteorological sensing signals collected by the vehicle-mounted meteorological sensing components. Since the vehicle-mounted meteorological sensing components are exposed to the outside of the vehicle body, their measurement results are affected by the dynamic disturbances caused by engine exhaust heat radiation, vehicle acceleration and deceleration, and lateral sway. The data cleaning module calls the pre-stored nonlinear compensation model in memory, matches the corresponding correction matrices for the thermal dimension and the physical disturbance dimension respectively, performs compensation calculations, and outputs the calibrated meteorological observation benchmark value to eliminate the system deviation caused by the physical operation of the hardware to the environmental perception.

[0058] The calculation formulas for the above-mentioned multidimensional system interference bias and meteorological observation benchmark values ​​are as follows:

[0059] ;

[0060] ;

[0061] in, This parameter represents the current sampling timestamp of the system. Indicates the first in the road network Hardware index of each node vehicle; This represents the first hardware calibration sparse diagonal matrix configured to compensate for thermal radiation bias, with a dimension of 5×5, where the diagonal elements corresponding to the non-thermal meteorological dimension are preset to 0; This represents the second hardware calibration matrix configured to compensate for velocity-related disturbances, with a dimension of 5×5; This represents the third hardware calibration matrix configured to compensate for longitudinal inertial disturbances, with a dimension of 5×5; This represents the fourth hardware calibration matrix configured to compensate for lateral attitude disturbances, with a dimension of 5×5; The column vector representing the dynamic features of vehicle i at time t has a dimension of 3×1. This indicates the real-time operating temperature of the engine coolant, obtained from the CAN bus. This represents the column vector of raw meteorological sensing signals collected by the vehicle-mounted meteorological sensing components. It has a dimension of 5×1, and the elements correspond to the road surface water depth, ice thickness, snow thickness, visibility distance, and crosswind speed, respectively. This indicates the base reference temperature of the external environment; Representing vectors The real-time driving speed scalar in the data; Representing vectors The longitudinal acceleration scalar in; Representing vectors The scalar of the yaw angle in the middle; This represents a preset static system bias compensation column vector with a dimension of 5×1; This represents the meteorological observation baseline column vector output after noise reduction and calibration, with a dimension of 5×1; This represents the multidimensional system interference deviation column vector of the calculated output, with a dimension of 5×1.

[0062] S2: The spatial mapping module constructs spatial Gaussian decay weights and performs coordinate mapping from discrete point trajectories to static topological meshes.

[0063] The spatial mapping module reads the electronic map database and divides the physical road network into multiple static topological grids. This module extracts node vehicles. The real-time GPS coordinates are used to calculate the relationship between these coordinates and the static topology mesh. The physical Euclidean distance between centers; based on the principle of spatial heterogeneity, the spatial mapping module constructs a Gaussian decay weight function to map discrete moving point observations to the corresponding static grid space, realizing the topological reconstruction from dynamic point vectors to static grid field data, so as to meet the requirements of meteorological numerical models for gridded input.

[0064] The formulas for calculating the spatial Gaussian attenuation weights and the reconstructed grid-fused observation column vectors are as follows:

[0065] ;

[0066] ;

[0067] in, This represents a static topological grid index with a total of M grids. Indicates node vehicle coordinates and grid The physical Euclidean distance scalar between centers; L represents the system's preset spatial correlation physical length constant. This represents the exponential operator with the natural constant e as its base; Indicates node vehicle Corresponding to grid The mapping weight scalar; This represents the grid fusion observation column vector generated after coordinate reconstruction, with a dimension of 5×1.

[0068] S3: The simulation error module schedules high-performance computing resources in the cloud to run the simulation environment, and derives the dynamic observation error covariance based on the LSTM network.

[0069] The simulation error module schedules cloud-based high-performance computing resources to run the Observation System Experiment (OSSEs) model, generating a realistic atmospheric reference field. Considering the high-frequency dynamic noise generated by the hardware sensors under different vehicle driving conditions, the simulation error module activates the Long Short-Term Memory (LSTM) network inference unit. This inference unit reads the dynamic characteristics of the vehicle nodes. The system adaptively extrapolates the reliability of the hardware at the current moment using the grid observation vector. The simulation error module ensures the symmetric positive definite mathematical properties of the error matrix through quadratic transformation and injects simulated physical noise into the real atmospheric reference field to achieve physical simulation of the vehicle-mounted hardware perception process.

[0070] The formulas for calculating the dynamic scaling weight diagonal matrix, the dynamic observation error covariance matrix, and the simulated observation field state vector are as follows:

[0071] ;

[0072] ;

[0073] ;

[0074] in, This represents an algebraic operator that expands the elements of a vector into a diagonal matrix along the main diagonal. This represents the logistic regression activation operator, used to map the real number field to the interval (0,1); and These represent the pre-set weight matrix and bias vector parameters within the LSTM model, respectively. This represents the hidden state feature vector cached by the LSTM inference unit at the previous time step. Let represent the dynamic observation error covariance matrix of vehicle i at output node after quadratic transformation; This represents the output dynamically scaled weight diagonal matrix, with dimensions of 5×5; This represents the static observation error covariance matrix preset by the meteorological hardware at the factory, with a dimension of 5×5; Representation matrix The transpose of the matrix; This represents the state vector of the simulated observation field generated after noise fusion; This represents the observation operator matrix configured to perform interpolation transformation from grid state to observation space; This represents the column vector of the real atmospheric reference field generated by the system, with a dimension of 5M×1; This represents the column vector of simulated random physical noise generated by the system.

[0075] S4: The assimilation evaluation module applies the consistent Kalman filtering mechanism to perform physical state assimilation of the meteorological background field and sensor data.

[0076] The assimilation evaluation module reads the biased meteorological background field generated based on different parameterization schemes and substitutes it into the dynamic observation error matrix output in step S3. This module calculates the Kalman gain matrix to quantify the weight of the system's trust allocation between the hardware measured values ​​and the physical model prediction values. The assimilation evaluation module uses the gain to perform real-time correction on the background field, synthesizes the local microclimate features captured by the vehicle network into a high-precision meteorological analysis field, and updates the system's posterior error simultaneously.

[0077] The formulas for calculating the physical state gain matrix, the global meteorological analysis field vector, and the posterior error covariance matrix are as follows:

[0078] ;

[0079] ;

[0080] ;

[0081] in, This represents the physical state gain matrix obtained by the assimilation evaluation module. This represents the prior background error covariance matrix corresponding to the meteorological background field; Represents the observation operator matrix The transpose of the matrix; This represents the total number of entity nodes. The combined generated global dynamic observation error covariance diagonal matrix; This represents the global meteorological analysis field vector output after assimilation. This represents the generated initial fluid meteorological background field state vector; This indicates that the posterior error covariance matrix is ​​updated and written to memory after the assimilation calculation is completed. This indicates the identity matrix operator configured to participate in the operation.

[0082] S5: The assimilation evaluation module quantifies the information gain scalar of virtual deployment nodes.

[0083] The assimilation evaluation module virtually configures candidate deployment nodes in the road network topology model. Based on the entropy reduction principle in information theory, the assimilation evaluation module calculates the amount of uncertainty elimination of the forecast system by virtual nodes by comparing the changes in the determinant of the system's posterior error covariance matrix before and after the introduction of the node. This transforms the observation value of the geographical location into a quantified information gain scalar, guiding the selection of subsequent hardware installation schemes.

[0084] The formula for calculating the information gain quantization scalar of the aforementioned virtual node is as follows:

[0085] ;

[0086] in, Represents physical nodes Quantize the information gain scalar generated by the configuration sensing hardware; This represents the index of a candidate hardware deployment node in the system's road network topology model; Represents the operator for the natural logarithm; This indicates an operator that performs matrix determinant calculations to obtain the scalar volume of the matrix; Indicates the assumption of adding a node The specific posterior error covariance matrix is ​​calculated after observing the data.

[0087] S6: The assimilation evaluation module configures a multi-objective constraint cost function to solve for the optimal hardware topology layout vector.

[0088] The assimilation evaluation module initiates the genetic optimization engine, which performs crossover and mutation iteration operations on binary sequences under multi-objective constraints that comprehensively consider the procurement cost of individual hardware, communication overhead, and the benefits of meteorological sensing accuracy. The engine selects the individuals with the highest fitness and outputs the optimal hardware topology layout vector that balances economic cost and sensing efficiency, which is used to guide the physical construction layout of sensors and roadside base stations in actual engineering projects.

[0089] The formula for calculating the optimal sensor network hardware topology layout vector is as follows:

[0090] ;

[0091] in, This represents the optimal hardware topology layout vector output after the optimization engine iteratively converges; This represents the deployment state vector of the hardware entities generated through system iteration; This represents an operator for optimizing the independent variable to minimize the value of the objective function. This represents the pre-configured root mean square error term penalty weight constant; This represents the root mean square error constant between the assimilated meteorological field and the real reference field under the current deployment scheme S; This represents the pre-configured information gain term penalty weight constant; A constant representing the total number of candidate deployment nodes; Represents the first state in the state vector S. A Boolean element (1 represents configuring hardware, 0 represents empty). This represents the pre-configured penalty weight constant for the engineering economic cost item; Indicates at node The unit physical economic cost constant for installing hardware equipment.

[0092] S7: The instruction execution module extracts local meteorological slices coupled with traffic flow features, generates physical prevention and control early warning instructions, and issues them for execution.

[0093] Based on layout vectors After the physical construction is completed, the instruction execution module safeguards the road network security in real time; targeting specific road sections. The instruction execution module constructs a spatial feature extraction operator to extract a local meteorological feature column vector specific to the road segment from the global meteorological analysis field. Subsequently, the module reads the local traffic flow physical state column vector extracted by the supporting hardware of the road segment and performs matrix cascading operations to generate a risk assessment feature tensor. The instruction execution module uses a Logit classifier to assess the collision risk probability. Once the probability exceeds the safety threshold, the cloud cluster generates a control message that conforms to the protocol, driving the on-board ECU to perform a physical closed-loop operation to adjust the throttle intake volume or the braking intervention threshold.

[0094] The formulas for calculating the above-mentioned local microclimate characteristic column vector, risk assessment tensor, and collision risk probability value are as follows:

[0095] ;

[0096] ;

[0097] ;

[0098] in, This indicates the preset target risk road segment index; Indicates based on road section Spatial feature extraction operator matrix for geographic boundary configuration; This represents the global meteorological analysis field vector output after assimilation based on the optimal network topology; Indicates the extracted target road segment The local microclimate feature column vector has a dimension of 5×1; Indicates road segment The local traffic flow state column vector, with a dimension of 3×1, consists of the average vehicle speed, flow density, and average acceleration of the road segment. This represents a risk assessment tensor generated by the cascading of meteorological and traffic flow characteristics along the dimensional plane, with a dimension of 8×1. The weight row vector of the classification model has a dimension of 1×8; A scalar representing the bias offset of the classification model; This represents a scalar value indicating the probability of collision risk on the target road segment.

[0099] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation, characterized in that: The method includes the following steps: The system receives data packets reported by the mobile sensor network through the communication interface, extracts the mechanical dynamic characteristics recorded by the controller local area network bus on the node vehicle and the raw meteorological sensing signals collected by the on-board meteorological sensing components; calls the pre-stored nonlinear compensation model to perform compensation calculations and outputs the meteorological observation benchmark value. The electronic map database is read, and the physical road network is divided into multiple static topological grids. The real-time GPS coordinates of the vehicle at the node are extracted, the physical Euclidean distance between the coordinates and the center of the corresponding static topological grid is calculated, and a Gaussian decay weight function is constructed based on the principle of spatial heterogeneity to map the meteorological observation benchmark value to the corresponding static grid space, generating a grid fusion observation column vector. The cloud computing resources are scheduled to run the simulation test model of the observation system and generate a real atmospheric reference field. The long short-term memory network inference unit is called to use the mechanical dynamic features and the grid fused observation column vector as input to deduce the dynamic scaling weight. The dynamic observation error covariance matrix is ​​deduced by combining the quadratic transformation. Simulation physical noise is injected into the real atmospheric reference field to generate the state vector of the simulation observation field. Read the biased meteorological background field generated based on different parameterization schemes, substitute it into the dynamic observation error covariance matrix to solve the physical state gain matrix; use the physical state gain matrix to perform real-time correction on the meteorological background field, and output the global meteorological analysis field and the posterior error covariance matrix; In the road network topology model, a candidate deployment node is virtually configured. By comparing the change in the determinant of the posterior error covariance matrix before and after the introduction of the candidate deployment node, the information gain quantization scalar of the candidate deployment node is calculated. Under the multi-objective constraints of comprehensively considering the procurement cost of individual hardware and the benefits of meteorological sensing accuracy, a system cost function is constructed, and the genetic optimization engine is started to perform crossover and mutation iteration operations of binary sequences to output the optimal hardware topology layout vector. Based on the optimal hardware topology layout vector, the local meteorological feature column vector of the target risk road segment is extracted from the global meteorological analysis field; the local traffic flow physical state column vector extracted by the hardware supporting the target risk road segment is read, and a matrix concatenation operation is performed to generate a risk assessment feature tensor; the collision risk probability is evaluated using a Logit classifier, and a control message is generated when the collision risk probability is greater than a preset safety threshold to drive the on-board electronic control unit of the vehicle at that node to perform a physical closed-loop operation to adjust the throttle intake volume or adjust the braking intervention threshold.

2. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 1, characterized in that, The calculation formula for performing compensation operations and outputting meteorological observation baseline values ​​is as follows: ； ； in, This parameter represents the current sampling timestamp of the system. Indicates the first in the road network Hardware index of each node vehicle; This represents the first hardware calibration sparse diagonal matrix configured to compensate for thermal radiation bias. This represents the second hardware calibration matrix configured to compensate for speed-related disturbances; This represents the third hardware calibration matrix configured to compensate for longitudinal inertial disturbances; This represents the fourth hardware calibration matrix configured to compensate for lateral attitude disturbances; The column vector representing the dynamic features of vehicle i at time t; This indicates the real-time operating temperature of the engine coolant, obtained from the CAN bus. This represents the original meteorological perception column vector collected by the vehicle-mounted meteorological sensor component; This indicates the base reference temperature of the external environment; Representing vectors The real-time driving speed scalar in the data; Representing vectors The longitudinal acceleration scalar in; Representing vectors The scalar of the yaw angle in the text; This represents the preset static system deviation compensation column vector; This represents the meteorological observation baseline column vector output after noise reduction and calibration. This represents the multidimensional system interference deviation column vector of the calculated output.

3. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 1, characterized in that, The formula for generating the grid-fused observation column vector is as follows: ； ； in, This represents a static topological grid index with a total of M grids. Indicates node vehicle coordinates and grid The Euclidean distance scalar between the centers; L represents the physical length constant of the spatially dependent influences preset by the system; This represents the exponential operator with the natural constant e as its base; Indicates node vehicle Corresponding to grid The mapping weight scalar; This represents the grid fusion observation column vector generated after coordinate reconstruction.

4. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 3, characterized in that, The formulas for deriving the dynamic observation error covariance matrix and generating the state vector of the simulated observation field are as follows: ； ； ； in, This represents an algebraic operator that expands the elements of a vector into a diagonal matrix along the main diagonal. This represents the logistic regression activation operator, used to map the real number field to the interval (0,1); and These represent the pre-set weight matrix and bias vector parameters within the LSTM model, respectively. This represents the hidden state feature vector cached by the LSTM inference unit at the previous time step. Let represent the dynamic observation error covariance matrix of vehicle i at output node after quadratic transformation; This represents the dynamically scaled weight diagonal matrix of the output; This represents the static observation error covariance matrix preset by the meteorological hardware at the factory. Representation matrix The transpose of the matrix; This represents the state vector of the simulated observation field generated after noise fusion; This represents the observation operator matrix configured to perform interpolation transformation from grid state to observation space; This represents the column vector of the true atmospheric reference state generated by the system; This represents the column vector of simulated random physical noise generated by the system.

5. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 4, characterized in that, The formulas for calculating the global meteorological analysis field and the posterior error covariance matrix are as follows: ； ； ； in, This represents the physical state gain matrix obtained by the assimilation evaluation module. This represents the prior background error covariance matrix corresponding to the meteorological background field; Represents the observation operator matrix The transpose of the matrix; This represents the total number of entity nodes. The combined generated global dynamic observation error covariance diagonal matrix; This represents the global meteorological analysis field vector output after assimilation. This represents the generated initial fluid meteorological background field state vector; This indicates that the posterior error covariance matrix is ​​updated and written to memory after the assimilation calculation is completed. This indicates the identity matrix operator configured to participate in the operation.

6. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 5, characterized in that, The formula for calculating the information gain quantization scalar is as follows: ； in, Represents physical nodes Quantize the information gain scalar generated by the configuration sensing hardware; This represents the index of a candidate hardware deployment node in the system's road network topology model; Represents the operator for the natural logarithm; This indicates an operator that performs matrix determinant calculations to obtain the scalar volume of the matrix; Indicates the assumption of adding a node The specific posterior error covariance matrix is ​​calculated after observing the data.

7. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 6, characterized in that, The formula for calculating the optimal hardware topology layout vector is as follows: ； in, This represents the optimal hardware topology layout vector output after the optimization engine iteratively converges; This represents the deployment state vector of the hardware entities generated through system iteration; This represents an operator for optimizing the independent variable to minimize the value of the objective function. This represents the pre-configured root mean square error term penalty weight constant; This represents the root mean square error constant between the assimilated meteorological field and the real reference field under the current deployment scheme S; This represents the pre-configured information gain term penalty weight constant; A constant representing the total number of candidate deployment nodes; Represents the state vector S with the th Boolean elements; This represents the pre-configured penalty weight constant for the engineering economic cost item; Indicates at node The unit physical economic cost constant for installing hardware equipment.

8. The method for evaluating the layout of vehicle-to-everything (V2X) meteorological observation based on data assimilation according to claim 7, characterized in that, The formula for calculating the probability of a collision is as follows: ； ； ； in, This indicates the preset target risk road segment index; Indicates based on road section Spatial feature extraction operator matrix for geographic boundary configuration; This represents the global meteorological analysis field vector output after assimilation based on the optimal network topology; Indicates the extracted target road segment The column vector of local microclimate characteristics; Indicates road segment Local traffic flow state column vector; This represents the risk assessment tensor generated by the cascading of meteorological and traffic flow characteristics along the travel dimension. Represents the weight row vector of the classification model; A scalar representing the bias offset of the classification model; This represents a scalar value indicating the probability of collision risk on the target road segment.

9. A vehicle-to-everything (V2X) meteorological observation layout evaluation system based on data assimilation, characterized in that, The system is used to implement the vehicle-to-everything (V2X) meteorological observation layout evaluation method based on data assimilation as described in any one of claims 1 to 8. The system includes: a cloud computing cluster, a roadside communication architecture, and a mobile sensor network. The cloud computing cluster integrates: The data cleaning module is used to receive data packets reported by the mobile sensor network through the communication interface, extract the mechanical dynamic characteristics recorded by the controller local area network bus of the node vehicle and the raw meteorological sensing signals collected by the vehicle-mounted meteorological sensor components; call the pre-stored nonlinear compensation model to perform compensation calculations and output meteorological observation benchmark values. The spatial mapping module is used to read the electronic map database, divide the physical road network into multiple static topological grids, extract the real-time GPS coordinates of the vehicle at the node, calculate the physical Euclidean distance between the coordinates and the center of the corresponding static topological grid, and construct a Gaussian decay weight function based on the principle of spatial heterogeneity to map the meteorological observation benchmark value to the corresponding static grid space, generating a grid fusion observation column vector. The simulation error module is used to schedule high-performance computing resources in the cloud to run the simulation test model of the observation system and generate a real atmospheric reference field. It calls the long short-term memory network inference unit, uses the mechanical dynamic features and the grid fused observation column vector as input to deduce the dynamic scaling weight, combines the quadratic transformation to deduce the dynamic observation error covariance matrix, and injects simulation physical noise into the real atmospheric reference field to generate the state vector of the simulation observation field. The assimilation evaluation module is used to read the biased meteorological background field generated based on different parameterization schemes, substitute it into the dynamic observation error covariance matrix to solve the physical state gain matrix, use the physical state gain matrix to perform real-time correction on the meteorological background field, and output the global meteorological analysis field and the posterior error covariance matrix; it is also used to virtually configure candidate deployment nodes in the road network topology model, and calculate the information gain quantization scalar of the candidate deployment node by comparing the change in the determinant of the posterior error covariance matrix before and after the introduction of the candidate deployment node; and it is used to construct the system cost function under the multi-objective constraints of comprehensively considering the unit hardware procurement cost and the meteorological sensing accuracy gain, start the genetic optimization engine to perform crossover and mutation iteration operations of binary sequences, and output the optimal hardware topology layout vector. The instruction execution module is used to extract the local meteorological feature column vector of the target risk road segment from the global meteorological analysis field based on the optimal hardware topology layout vector; read the local traffic flow physical state column vector extracted by the hardware supporting the target risk road segment, and perform matrix concatenation operation to generate a risk assessment feature tensor; use the Logit classifier to evaluate the collision risk probability, and generate a control message when the collision risk probability is greater than a preset safety threshold, so as to drive the on-board electronic control unit of the vehicle at the node to perform a physical closed-loop operation to adjust the throttle intake volume or adjust the braking intervention threshold.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the computing device, the computing device implements the vehicle-to-everything (V2X) meteorological observation layout evaluation method based on data assimilation as described in any one of claims 1 to 8.

Images