Road property adaptive topology awareness and application system and method

Abstract

The invention relates to the technical field of road condition intelligent perception and application, in particular to a road condition adaptive topology perception and application system and method. A condition cluster perception module monitors road condition parameters in real time; the data transmission communication module realizes differential data transmission strategies; a road section dynamic risk grade evaluation module calculates a road section risk index and grade according to the road structural state parameters and a risk evaluation model; the node adjustment and resource allocation module adjusts the activation state and sampling frequency of each sensor according to the road section risk level, simulates and verifies the implementation effect of a resource allocation scheme under different road section risk levels by using a constructed road system digital twin model, and optimizes the resource allocation scheme by using a multi-objective optimization algorithm; the application service module converts the risk level into a specific service action; the cooperative operation of the modules realizes the flexible allocation of road condition perception, communication and computing resources, and improves the utilization efficiency of the resources and the service efficiency of the system.

Description

Technical Field

[0001] This application relates to the technical field of intelligent perception and application of road conditions, and particularly to a road condition adaptive topology perception and application system and method. Background Art

[0002] At present, with the continuous development of intelligent roads, the road condition perception system, as an important support for the monitoring and service of road traffic infrastructure that integrates advanced information and digital technologies such as the Internet of Things, big data, and artificial intelligence, collects real-time data on the structural state parameters of the road body by deploying various sensors and monitoring devices inside and outside the road, such as fiber Bragg grating strain sensors, earth pressure cells, cameras, radars, infrared detectors, etc., and realizes the real-time perception and prediction of the road body state by combining high-performance communication transmission technology and advanced data processing and analysis algorithms, which is of great significance for ensuring road traffic safety and improving road service efficiency.

[0003] Traditional road condition perception systems usually adopt a fixed sensor network deployment. For example, a method and device for monitoring the state of a power pipe group in a key section of an intelligent road disclosed in a patent application with the publication number CN118933068A realizes the refined monitoring of the state of the power pipe group in the key section through a pipe pillow pressure sensor network. However, this fixed sensor network layout is extremely likely to lead to redundant perception resources in conventional sections and insufficient monitoring capabilities for abnormal events in key sections, and it is unable to dynamically adjust the activation logic of perception devices and the resource allocation scheme according to road risks, restricting the efficiency of road safety early warning and emergency response, and ultimately resulting in the current road condition perception system being unable to adapt to different section scenarios and the dynamic risks in the actual service process of the road, with low monitoring efficiency and service efficiency. Summary of the Invention

[0004] Aiming at the problem of low monitoring efficiency and service efficiency of the current road condition perception system, a road condition adaptive topology perception and application system and method are provided.

[0005] This application provides a road condition adaptive topology perception and application system. The system includes:

[0006] A state cluster perception module for real-time monitoring of the road structure state parameters through a heterogeneous perception network composed of multi-modal sensors;

[0007] A section dynamic risk level assessment module for calculating a section risk index according to the road structure state parameters and a risk assessment model, and obtaining a section risk level according to the section risk index;

[0008] A data transmission and communication module is used to perform low-latency transmission of abnormal events and low-power consumption transmission of regular events using a hierarchical communication protocol, and perform spatio-temporal alignment on the multi-source heterogeneous data generated by the heterogeneous sensing network based on three-dimensional geographic grid coding and precise time protocol.

[0009] A node adjustment and resource allocation module is used to adjust the activation status and sampling frequency of each sensor in the heterogeneous sensing network according to the road section risk level; it is also used to simulate and verify the implementation effects of various resource allocation schemes under different road section risk levels using the constructed digital twin model of the road system, and obtain the optimal resource allocation scheme using a multi-objective optimization algorithm.

[0010] Furthermore, the road section dynamic risk level assessment module is also used to dynamically adjust the weights in the risk assessment model according to historical monitoring data and the LSTM-GRU hybrid network model; wherein, the LSTM-GRU hybrid network model includes an input layer, an LSTM hidden layer, a fully connected layer, and an output layer. The input layer is used to receive the historical monitoring data. The LSTM hidden layer includes three layers of LSTM units, and each layer of LSTM units includes 128 neurons. Each neuron has a forget gate, an input gate, and an output gate structure. The fully connected layer includes a first fully connected layer and a second fully connected layer. The first fully connected layer includes 64 neurons, and the second fully connected layer includes 32 neurons. The output layer is used to output the environmental prediction parameters and weight adjustment coefficients for a future period of time.

[0011] Furthermore, during the training stage of the LSTM-GRU hybrid network model, the road section dynamic risk level assessment module uses the AdamW optimizer to achieve adaptive learning rate adjustment and weight decay, and configures an early stopping mechanism to prevent overfitting. The loss function of the LSTM-GRU hybrid network is the Focal Loss function.

[0012] Furthermore, the LSTM-GRU hybrid network model assigns different weights to the prediction results at different time scales through an attention mechanism for short-term, medium-term, and long-term predictions, and the output of the LSTM-GRU hybrid network model also includes the confidence interval of the prediction results.

[0013] Furthermore, the node adjustment and resource allocation module is also used to calculate the contribution degree of each sensor in the heterogeneous sensing network to risk assessment based on historical monitoring data, and use an ensemble learning method to calculate the information gain of different sensor combinations to obtain a set of high-value sensing nodes, and activate each node in the set of high-value sensing nodes according to the real-time road section risk level.

[0014] Further, the data transmission and communication module is also used to obtain the timestamp messages sent by each node in the heterogeneous sensing network, calculate the transmission delay of each node based on the system main clock source, perform time compensation on the timestamp messages of each node according to the transmission delay, and use the compensated time as the data acquisition time tag of each node.

[0015] Further, the road condition adaptive topology perception and application system further includes:

[0016] An application service module, which is used to map the road section risk level to specific service actions and perform cross-departmental service closed-loop linkage.

[0017] In addition, the present application also provides a road condition adaptive topology perception and application method. The method includes:

[0018] Step S1, obtaining the road structural condition parameters monitored in real time by a heterogeneous sensing network composed of multimodal sensors;

[0019] Step S2, calculating the road section risk index according to the road structural condition parameters and the risk assessment model, and obtaining the road section risk level according to the road section risk index;

[0020] Step S3, using a hierarchical communication protocol to perform low-latency transmission of abnormal events and low-power consumption transmission of regular events, and performing spatio-temporal alignment on the multi-source heterogeneous data generated by the heterogeneous sensing network based on three-dimensional geographic grid coding and precise time protocol;

[0021] Step S4, adjusting the activation state and sampling frequency of each sensor in the heterogeneous sensing network according to the road section risk level, using the constructed digital twin model of the road system to simulate and verify the implementation effects of various resource allocation schemes under different road section risk levels, and using a multi-objective optimization algorithm to obtain the optimal resource allocation scheme.

[0022] Further, the expression of the risk assessment model is:

[0023] In the formula, RI is the comprehensive risk index of the road section; G is the subgrade deformation rate, and the value range is 0-1; A is the asphalt layer fatigue damage index, and the value range is 0-1; B is the base layer void risk index, and the value range is 0-1; α, β, and γ are all weight coefficients; among them, the calculation formula of the subgrade deformation rate is:

[0024] In the formula, is the cumulative subgrade settlement amount (mm) during continuous monitoring, is the monitoring time interval (days), is the critical deformation rate threshold (mm / day);

[0025] The calculation formula for the fatigue damage index of the asphalt layer is:

[0026] In the formula, is the strain level The actual number of load applications at is the bottom tensile strain of the asphalt layer (με), is the elastic modulus of the asphalt mixture (MPa), is the experimental parameter related to the characteristics of the asphalt material;

[0027] The calculation formula for the base void risk index is:

[0028] In the formula, is the pressure at the bottom of the base monitored by the pressure sensor array exceeding the void pressure threshold The area (kPa), is the total monitoring area (m 2 ).

[0029] Furthermore, the steps of three-dimensional geographic grid coding include:

[0030] Step S301, divide the road and its surrounding environment into three-dimensional grids to form basic spatial units;

[0031] Step S302, adopt a hierarchical nested coding strategy, apply the Geohash or H3 algorithm in the horizontal direction to generate plane coordinate coding, and superimpose the Z-axis hierarchical coding in the vertical direction to obtain the unique geographic coding of each basic spatial unit;

[0032] Step S303, map the physical positions of the sensors in the heterogeneous perception network to the corresponding three-dimensional grid cells, and use their geographic coding as the spatial label of the data collected by the sensors.

[0033] The above road condition adaptive topology perception and application system and method provide a perception resource elastic scheduling mechanism based on real-time risk levels, specifically including a condition cluster perception module for real-time monitoring of road structural condition parameters; a section dynamic risk level assessment module for calculating a section risk index based on road structural condition parameters and a risk assessment model, and obtaining a section risk level based on the section risk index; a node adjustment and resource allocation module for adjusting the activation status and sampling frequency of each sensor in a heterogeneous perception network composed of multi-modal sensors according to the section risk level, and using the constructed digital twin model of the road system to simulate and verify the implementation effects of various resource allocation schemes under different section risk levels, and obtaining an optimal resource allocation scheme using a multi-objective optimization algorithm. This application realizes the flexible allocation of perception and computing resources according to the dynamic risk levels of sections, avoids resource waste on conventional sections while ensuring sufficient perception of key sections and abnormal events, and significantly improves the efficiency of the road condition perception and application system; at the same time, the spatio-temporal alignment of multi-source heterogeneous data through the data transmission and communication module further improves the reliability of monitoring data. This system can provide more accurate and efficient decision-making and services for traffic management, road maintenance, public travel, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 FIG. is a schematic structural diagram of a road condition adaptive topology perception and application system in an embodiment;

[0035] Figure 2 FIG. is a schematic diagram comparing the training process accuracies of an LSTM-GRU hybrid model, an LSTM model, and a traditional time series model in an embodiment;

[0036] Figure 3 FIG. is a schematic diagram comparing the overall model performances of an LSTM-GRU hybrid model, an LSTM model, and a traditional time series model in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] In order to make the objectives, technical solutions, and advantages of this application clearer, the following further details this application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0038] Embodiment 1

[0039] As Figure 1 shown, this embodiment provides a road condition adaptive topology perception and application system, including a condition cluster perception module, a section dynamic risk level assessment module, a data transmission and communication module, and a node adjustment and resource allocation module.

[0040] The behavior cluster perception module is used to monitor road structural behavior parameters in real time through a heterogeneous perception network composed of multi-modal sensors.

[0041] Among them, road structural behavior parameters include subgrade vertical deformation rate and collapse rate, mechanical response of pavement structure layer, environmental parameters, pavement condition, etc. The heterogeneous perception network composed of multi-modal sensors includes different types of perception terminals, and each perception terminal monitors different road structural behavior parameters in real time. Exemplarily, the perception terminals can include distributed fiber Bragg grating sensors, piezoelectric sensor arrays, humidity and temperature sensors, and high-resolution cameras to form a monitoring system covering the road body and its surrounding environment. The distributed fiber Bragg grating sensors are used to collect the subgrade vertical deformation rate, settlement amount, and bottom tensile strain of the asphalt layer in real time. The piezoelectric sensor array is used to collect the mechanical response of the pavement structure layer in real time, including the bottom tensile strain of the asphalt layer and the pressure at the bottom of the base layer. The humidity and temperature sensors are used to monitor environmental parameters such as temperature and humidity inside and outside the road. The high-resolution cameras are used for visual monitoring of the pavement condition.

[0042] Furthermore, the deployment scheme of the perception terminals has flexibility and can be adjusted according to factors such as road grade, traffic volume, and historical risks. Exemplarily, the distributed fiber Bragg grating sensors can specifically adopt a grid layout scheme, forming a continuous monitoring baseline at intervals of 50 meters along the longitudinal direction of the road, and achieving millimeter-level deformation resolution through the wavelength drift characteristics of Bragg gratings to accurately capture the subgrade vertical settlement rate and potential collapse signs. The piezoelectric sensor array can lay multiple sensing lines transversely along the asphalt layer to collect the interlayer stress distribution and dynamic strain modulus change under the action of vehicle loads in real time. The humidity and temperature sensor network is deployed at intervals of 10 meters on the road shoulder and the central isolation belt to continuously monitor the influence of temperature and humidity gradients on the performance of asphalt materials. The high-resolution cameras adopt a hierarchical three-dimensional layout strategy, setting a monitoring point every 1-2 kilometers along the longitudinal direction of the road, and adding a camera group with different vertical angles at key nodes to form an all-round visual monitoring network covering the road behavior and its surrounding environment. Among them, the key nodes are the nodes included in the sections with a higher risk level, such as long and steep longitudinal slope sections, filled sections, cut sections, half-filled and half-excavated sections, special soil subgrade sections, curved sections, bridgehead transition sections, entrance and exit ramps, etc. The camera group with different vertical angles includes lenses with different pitch angles, such as cameras with pitch angles of 30°, 60°, and 90° respectively. The 30° lens is a low pitch angle lens that can focus on local details of the pavement, such as microscopic features like crack width. The 60° lens is a medium pitch angle lens that can cover the entire pavement within the range of a single lane. The 90° lens is a vertical overhead lens that can achieve full-width pavement image acquisition. This multi-angle combination can eliminate monitoring blind spots through the superposition of perspectives, thereby improving the monitoring accuracy.

[0043] The section dynamic risk level assessment module is used to calculate the section risk index according to the road structural state parameters and the risk assessment model, and obtain the section risk level based on the section risk index.

[0044] Among them, the section risk index can be the weighted average sum of the road structural state parameters. In a preferred embodiment, the specific mathematical expression of the risk assessment model is shown in Equation (1): (1)

[0045] In the formula, RI is the comprehensive risk index of the section; G is the subgrade deformation rate, and the value range is 0 - 1; A is the asphalt layer fatigue damage index, and the value range is 0 - 1; B is the base course debonding risk index, and the value range is 0 - 1; α, β, and γ are all weight coefficients.

[0046] The subgrade deformation rate reflects the deformation speed of the subgrade under the action of load and is an important index for evaluating the stability of the subgrade. The subgrade deformation rate can be calculated based on the cumulative settlement of the subgrade during the monitoring period, the monitoring time interval, and the critical deformation rate threshold. The specific calculation formula is shown in Equation (2): (2)

[0047] In the formula, is the cumulative settlement of the subgrade (mm) during the continuous monitoring period, is the monitoring time interval (days), is the critical deformation rate threshold (mm / day), usually taking 0.05 - 0.2 mm / day.

[0048] The asphalt layer fatigue damage index reflects the cumulative damage degree of the asphalt mixture under the action of cyclic load and is calculated based on the actual number of load applications at the strain level, the bottom tensile strain of the asphalt layer, and the elastic modulus of the asphalt mixture. The specific calculation formula is shown in Equation (3): (3)

[0049] In the formula, is the strain level is the actual number of load applications at the strain level, is the bottom tensile strain of the asphalt layer ( ), is the elastic modulus of the asphalt mixture (MPa),

[0050] is the experimental parameter related to the asphalt material properties.

[0050] The base course debonding risk index reflects the separation degree between the base course and the underlying structural layer and is mainly calculated by monitoring the interlayer stress distribution characteristics. The specific calculation formula is shown in Equation (4): (4)

[0051] In the formula, is the pressure at the bottom of the base layer monitored by the pressure sensor array exceeding the void pressure threshold area (kPa), is the total monitoring area (m 2 ).

[0052] The risk level of the road section can be divided into low risk, medium risk and high risk. In a preferred embodiment, when the comprehensive risk index of the road section is less than 0.3, the risk level of the road section is low risk. When the comprehensive risk index of the road section is greater than or equal to 0.3 and less than 0.6, the risk level of the road section is medium risk. When the comprehensive risk index of the road section is greater than or equal to 0.6, the risk level of the road section is high risk.

[0053] Furthermore, the road section dynamic risk level evaluation module of this embodiment is also used to dynamically adjust the weight coefficients in the risk assessment model according to historical monitoring data and the LSTM-GRU hybrid network model. Specifically, the LSTM-GRU hybrid network model includes an input layer, an LSTM hidden layer, a fully connected layer and an output layer. The input layer is used to receive historical monitoring data. Specifically, the historical monitoring data can be multi-dimensional time series data including historical temperature and humidity data, historical rainfall and historical traffic load data. The LSTM hidden layer includes three layers of LSTM cells, and each layer of LSTM cells includes 128 neurons. Each neuron has a forget gate, an input gate and an output gate structure. The fully connected layer includes a first fully connected layer and a second fully connected layer. The first fully connected layer includes 64 neurons, and the second fully connected layer includes 32 neurons. Both the first fully connected layer and the second fully connected layer use the ReLU activation function. The output layer is used to output environmental prediction parameters and weight adjustment coefficients for a future period of time, such as outputting environmental prediction parameters and weight adjustment coefficients for the next 24 hours.

[0054] Furthermore, the LSTM-GRU hybrid network model of this embodiment assigns different weights to the prediction results of different time scales through an attention mechanism to perform short-term, medium-term and long-term predictions. Specifically, high-frequency dynamic features are captured through a spatio-temporal attention mechanism to achieve short-term prediction (1-3 hours); a periodic pattern encoder is constructed using GRU units to achieve medium-term prediction (24 hours), and long-term prediction (7 days) is performed through the LSTM hidden layer. The outputs of each branch are dynamically weighted through an adaptive fusion gate, and the weight coefficients are generated in real time by the covariance matrix of historical prediction errors, forming a collaborative optimization mechanism for short-medium-long term prediction results.

[0055] In terms of training optimization, the LSTM-GRU hybrid network model uses Focal Loss to improve the cross-entropy loss function, enhances the attention to abnormal traffic events through a modulation factor, and realizes adaptive parameter update in cooperation with the AdamW optimizer (weight decay coefficient λ = 1e-5). To prevent overfitting, the model is also configured with an early stopping strategy, that is, training stops if the validation set loss does not improve for 10 rounds, and the model parameters are updated regularly to ensure the model's continuous adaptability to the dynamic traffic environment. Aiming at the sparse problem of traffic scene data, the LSTM-GRU hybrid network model adopts a progressive transfer learning strategy. Specifically, first, parameter initialization is performed on historical section data, and the spatio-temporal domain differences are eliminated through the feature alignment layer. Subsequently, progressive transfer is implemented for the new section data, and the model parameters are fine-tuned. Figure 2 Figure 2 shows the comparison of the accuracy of the LSTM-GRU hybrid network model during training with that of traditional time series models and LSTM models after adopting this training optimization strategy. It can be seen from the figure that the accuracy of the LSTM-GRU hybrid network model during training is relatively high. In addition, Figure 3 Figure 3 shows the overall comparison of the model performance of the LSTM-GRU hybrid network model with that of traditional time series models and LSTM models. It can be seen from the figure that the overall performance of the LSTM-GRU hybrid network model is better than that of other models, and the prediction accuracy of the LSTM-GRU hybrid network model is about 18% higher than that of traditional time series models.

[0056] In addition, the output of the LSTM-GRU hybrid network model also includes the confidence interval of the prediction result, which provides an uncertainty measure for risk assessment and further improves the reliability of the prediction result.

[0057] The data transmission and communication module is used to perform low-latency transmission of abnormal events and low-power consumption transmission of normal events using a hierarchical communication protocol. Among them, the hierarchical communication protocol includes a high-speed communication protocol and a low-power wide area network protocol. The high-speed communication protocol includes, but is not limited to, the V2X or 5G-Uu protocol, and the low-power wide area network protocol includes, but is not limited to, the LoRa or NB-IoT protocol. Specifically, different transmission methods are selected according to the importance and timeliness requirements of the data. For example, emergency data such as abnormal events (such as severe deformation alarms) are transmitted with low latency through the high-speed communication protocol, and normal events (such as the structural state parameters of the road monitored daily) are transmitted with low power consumption through the low-power wide area network protocol to reduce energy consumption.

[0058] Further, to solve the spatio-temporal alignment problem of multi-source heterogeneous data, the data transmission and communication module is also used to perform spatio-temporal alignment on the multi-source heterogeneous data generated by the heterogeneous sensing network composed of multi-modal sensors based on three-dimensional geographical grid coding and Precision Time Protocol. Among them, the multi-source heterogeneous data are different types of data generated after preprocessing the road structural state parameters collected by various sensors. The preprocessing can specifically include data deduplication, format conversion, outlier processing, etc. The specific steps of three-dimensional geographical grid coding are as follows: First, divide the road and its surrounding environment into three-dimensional grids to form basic spatial units (such as cubes or hexagonal columns); then, adopt a hierarchical nested coding strategy, apply the Geohash or H3 algorithm in the horizontal direction to generate plane coordinate coding, and stack the Z-axis hierarchical coding in the vertical direction to obtain the unique geographical coding of each basic spatial unit; then, map the physical location of the sensor to the corresponding three-dimensional grid unit, and use the geographical coding of the three-dimensional grid unit as the spatial label of the data collected by the sensor. Through the hierarchical structure characteristics of the coding, the three-dimensional geographical grid coding realizes the rapid retrieval and association of spatial data with different ranges and granularities, and based on the same or adjacent coding areas, performs spatial aggregation and fusion analysis on multi-source sensor data, effectively solving the spatial alignment problem of multi-source heterogeneous data and realizing the spatial association of multi-source sensor data.

[0059] The data transmission and communication module is also used to obtain the timestamp messages sent by each node in the heterogeneous sensing network composed of multi-modal sensors, calculate the transmission delay of each node based on the system master clock source, perform time compensation on the timestamp messages of each node according to the transmission delay, and use the compensated time as the data acquisition time label of each node.

[0060] Specifically, the system constructs a master-slave clock architecture based on the Precision Time Protocol (PTP), deploys a master clock source (Master Clock) as the global time reference, and each sensing terminal serves as a slave clock (Slave Clock) node to achieve nanosecond-level synchronization accuracy through a two-way timestamp exchange mechanism. The specific working process is as follows: Each node in the heterogeneous sensing network composed of multi-modal sensors continuously sends timestamp packets carrying local clock information to the data transmission and communication module. The data transmission and communication module calculates the transmission delay of each node based on the system master clock source, and performs offset compensation on the timestamps of each node according to the measured transmission delay in real time to generate a corrected unified time reference, which is used as the time label of the multi-source heterogeneous data to ensure the temporal consistency of the data. Further, the system also supports an automatic recovery mechanism in the network environment to ensure the stability of long-term operation. This precise time synchronization mechanism ensures the accuracy of the correlation analysis of multi-source heterogeneous data in the time dimension and is an important basis for multi-source data fusion.

[0061] The data transmission and communication module also supports an edge-cloud collaborative computing architecture, and flexibly adjusts the computing task allocation according to data processing requirements. Specifically, when processing data with high timeliness, such as alarming when the road structural state parameters exceed the corresponding thresholds, data can be processed at the sensor end; when processing data with high computational complexity, the road structural state parameters monitored by the sensor are uploaded to the cloud for relevant calculations. For example, the data transmission and communication module can send the monitored road structural state parameters to the cloud for calculating the risk level of the road section. A risk assessment model is deployed in the cloud to further improve the monitoring efficiency.

[0062] The node adjustment and resource allocation module is used to adjust the activation status and sampling frequency of each sensor in the heterogeneous sensing network composed of multi-modal sensors according to the road section risk level. Exemplarily, when the road section risk level is low risk, some sensors (about 20%) of the road section are activated, and the sampling frequency is reduced; when the road section risk level is medium risk, most sensors (about 70%) of the road section are activated, and the sampling frequency is appropriately adjusted; when the road section risk level is high risk, all sensors (about 100%) of the road section are activated, the sampling frequency is increased, and more edge computing resources are allocated.

[0063] The node adjustment and resource allocation module is also used to simulate and verify the implementation effects of various resource allocation schemes under different road section risk levels by using the constructed digital twin model of the road system, and obtain the optimal resource allocation scheme by using a multi-objective optimization algorithm.

[0064] Among them, the digital twin model of the road system includes a geometric model, a physical model, and a behavior model. The geometric model is used to accurately reproduce the road structure. The physical model is used to simulate the material properties of the road. The behavior model is used to simulate traffic flow and environmental effects. The model parameters of the digital twin model of the road system are inverted by measured data, and the mapping relationship between the real world and the digital model is established, while supporting macroscopic road network-level and microscopic structure-level simulations to achieve cross-scale analysis.

[0065] Specifically, the effects of various resource allocation schemes under different road section risk levels are simulated based on the digital twin model, and a multi-objective optimization algorithm is used to solve the optimal resource allocation scheme that meets multiple constraints. The constraints include but are not limited to monitoring coverage, data accuracy, energy consumption, and response time. By establishing a comprehensive evaluation index including monitoring coverage, data accuracy, energy consumption, and response time, the best balance point is found between monitoring quality and resource consumption.

[0066] The node adjustment and resource allocation module is also used to obtain the actual effect data after implementing the optimal resource allocation plan, and optimize the parameters of the digital twin model of the road system according to the actual effect data. The specific implementation process is as follows: The node adjustment and resource allocation module evaluates the current risk status every 5 minutes, calculates the road section risk index of each road section, executes the node activation / sleep instruction of each road section according to the road section risk index, and generates the optimal resource allocation plan based on the digital twin model of the road system and the multi-objective optimization algorithm. After that, collect the actual effect data and update and optimize the model parameters.

[0067] Furthermore, the node adjustment and resource allocation module is also used to evaluate the sensitivity of the impact of resource allocation on the accuracy of risk assessment and determine the key resource points.

[0068] Among them, the resource allocation includes the activation and sleep schemes of different sensors in the heterogeneous perception network composed of multi-modal sensors. By controlling the activation and sleep of different sensors and observing the change range of the road section risk index, when the change range exceeds a certain threshold, it is judged that the sensors activated in this resource allocation scheme have a greater impact on the accuracy of risk assessment, and they are used as key resource points.

[0069] Furthermore, the activation logic of the sensor is optimized in this embodiment. Specifically, the node adjustment and resource allocation module is also used to calculate the contribution degree of each sensor in the heterogeneous perception network composed of multi-modal sensors to risk assessment based on historical monitoring data, and calculate the information gain of different sensor combinations using the ensemble learning method to obtain a set of high-value perception nodes, and activate each node in the set of high-value perception nodes according to the real-time road section risk level.

[0070] Among them, the historical monitoring data includes the monitoring data of various sensors under different environmental conditions and road section risk level scenarios. The node adjustment and resource allocation module deeply mines the historical monitoring data by using the sensitivity analysis method, identifies the contribution degree of different sensor data to the calculation of the road section risk level, and then evaluates the uniqueness and importance of the information provided by each sensor in the risk assessment process. Further, the node adjustment and resource allocation module calculates the information gain of different sensor combinations through the random forest algorithm. The random forest algorithm improves the overall classification or regression accuracy by constructing multiple decision tree models and integrating their prediction results. Specifically, the random forest algorithm takes different sensor combinations as input features and the calculation result of the road section risk level as the target variable, and simulates the impact of different sensor combinations on the calculation of the road section risk level by training multiple decision tree models. During the training process, the algorithm will automatically learn the weights of each sensor in different combinations and the degree of improvement of each combination on the risk assessment accuracy, that is, the information gain. This calculation method based on ensemble learning can make full use of the complementarity of multiple models, effectively avoid the overfitting or underfitting problems that may exist in a single model, and thus obtain a more robust and accurate evaluation result of the information gain of the sensor combination.

[0071] Based on the above contribution degree evaluation and information gain calculation results, the node adjustment and resource allocation module can further discover the distribution law of the contribution degree of different sensor combinations to the risk assessment. By setting reasonable thresholds or optimization objectives (such as maximizing the information gain, minimizing the sensing cost, etc.), the node adjustment and resource allocation module can screen out those high-value sensing node sets that have a significant contribution to the risk assessment from all possible sensor combinations. These high-value sensing nodes not only have high data quality and accuracy, but also can provide unique and complementary information under different risk scenarios, which helps to comprehensively improve the comprehensiveness and reliability of the risk assessment.

[0072] Finally, according to the dynamic changes of the real-time road section risk level, the node adjustment and resource allocation module can flexibly activate each node in the high-value sensing node set. For example, when it is detected that the risk level of a certain road section rises, the module will immediately activate the corresponding sensing node to increase the sensing density and accuracy in this area to obtain more detailed and timely risk information. On the contrary, when the risk level decreases, the module will also timely turn off some non-high-value sensing nodes to save sensing resources and reduce network energy consumption. This dynamic node activation strategy based on the risk level not only improves the response speed and adaptability of the sensing network, but also effectively extends the overall service life of the network, providing a strong guarantee for the stable operation of the heterogeneous sensing network composed of multi-modal sensors in a complex environment.

[0073] Further, the road condition adaptive topology perception and application system further includes:

[0074] The application business module is used to map the road section risk level to specific business actions and conduct cross-departmental business closed-loop linkages. Specifically, different levels of lane speed limits or closure measures are implemented according to the road section risk level. Dynamic route planning and risk warnings are pushed through the navigation application, a maintenance priority list including risk types, location coordinates, and urgency levels is generated, and it is automatically synchronized to the municipal maintenance and management system and work orders are dispatched. When the road section risk level breaks through the threshold, a cross-departmental collaborative response mechanism is automatically triggered. In addition, the application business module provides a standardized API interface to support data intercommunication with heterogeneous systems such as traffic management and municipal management, ensuring the timely flow of information and the efficient linkage of operations.

[0075] The road state adaptive topology perception and application system of this embodiment adopts: first, a dynamic risk-driven resource allocation mechanism that dynamically adjusts the sensor activation state and acquisition frequency according to the risk level to achieve "perception on demand", which not only ensures the monitoring quality of key road sections but also avoids resource waste on conventional road sections; second, a multi-source heterogeneous data fusion method that solves the spatio-temporal alignment problem of multi-source heterogeneous data through three-dimensional geographic grid coding and precise time protocol to improve the accuracy of the risk assessment model; third, an intelligent weight adjustment algorithm that introduces an LSTM network to dynamically adjust the weights of the risk assessment model, effectively adapting to changes in environmental conditions and improving the risk prediction accuracy; fourth, a cross-departmental business linkage framework that establishes a mapping relationship between the risk level and business actions to achieve a rapid response from abnormal events to emergency disposal, significantly improving the road safety management level. Through the collaborative design and innovative operation of each module, the system can flexibly allocate perception and computing resources according to the road section risk level, avoid resource waste on conventional road sections while ensuring that key road sections and abnormal events are fully perceived, significantly improving the overall performance of the intelligent road state perception and application system, and providing more accurate and efficient support and services for traffic management, road maintenance, public travel, etc.

[0076] Embodiment 2

[0077] This embodiment provides a road state adaptive topology perception and application method, which specifically includes the following steps:

[0078] Step S1, obtain the road structural state parameters real-time monitored by the heterogeneous perception network composed of multi-modal sensors.

[0079] Step S2, calculate the road section risk index according to the road structural state parameters and the risk assessment model, and obtain the road section risk level according to the road section risk index.

[0080] Step S3, use a hierarchical communication protocol to perform low-latency transmission of abnormal events and low-power consumption transmission of regular events, and perform spatio-temporal alignment on the multi-source heterogeneous data generated by the heterogeneous perception network based on three-dimensional geographic grid coding and precise time protocol.

[0081] Step S4: Adjust the activation status and sampling frequency of each sensor in the heterogeneous perception network according to the road section risk level, use the constructed digital twin model of the road system to simulate and verify the implementation effects of various resource allocation schemes under different road section risk levels, and use a multi-objective optimization algorithm to obtain the optimal resource allocation scheme.

[0082] Furthermore, the road behavior adaptive topology perception and application method further includes:

[0083] Step S5: Dynamically adjust the weights in the risk assessment model according to historical monitoring data and the LSTM-GRU hybrid network model.

[0084] Among them, the LSTM-GRU hybrid network model includes an input layer, an LSTM hidden layer, a fully connected layer, and an output layer. The input layer is used to receive historical monitoring data. The LSTM hidden layer includes three layers of LSTM units, and each layer of LSTM units includes 128 neurons. Each neuron has a forget gate, an input gate, and an output gate structure. The fully connected layer includes a first fully connected layer and a second fully connected layer. The first fully connected layer includes 64 neurons, and the second fully connected layer includes 32 neurons. The output layer is used to output environmental prediction parameters and weight adjustment coefficients for a future period of time. In addition, during the training stage of the LSTM-GRU hybrid network model, this method uses the AdamW optimizer to achieve adaptive learning rate adjustment and weight decay, and configures an early stopping mechanism to prevent overfitting. The loss function of the LSTM-GRU hybrid network is the Focal Loss function. Specifically, the LSTM-GRU hybrid network model can also assign different weights to the prediction results at different time scales through an attention mechanism to perform short-term, medium-term, and long-term predictions, and the output of the LSTM-GRU hybrid network model also includes the confidence interval of the prediction results.

[0085] Furthermore, the road behavior adaptive topology perception and application method further includes:

[0086] Step S6: Calculate the contribution degree of each sensor in the heterogeneous perception network to risk assessment based on historical monitoring data, and use an ensemble learning method to calculate the information gain of different sensor combinations to obtain a set of high-value perception nodes, and activate each node in the set of high-value perception nodes according to the real-time road section risk level.

[0087] Furthermore, the road behavior adaptive topology perception and application method further includes:

[0088] Step S7: Evaluate the sensitivity of the impact of resource configuration on risk assessment accuracy and determine key resource points.

[0089] Furthermore, the road behavior adaptive topology perception and application method further includes:

[0090] Step S8: Obtain the timestamp messages sent by each node in the heterogeneous perception network, calculate the transmission delay of each node based on the system's master clock source, perform time compensation on the timestamp messages of each node according to the transmission delay, and use the compensated time as the data acquisition time tag for each node.

[0091] Further, the road condition adaptive topology perception and application method further includes:

[0092] Step S9: Map the road section risk level to specific business actions for cross-departmental business closed-loop linkage.

[0093] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0094] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A road condition adaptive topology perception and application system, characterized in that The system includes: A behavior cluster perception module for real-time monitoring of road structural behavior parameters through a heterogeneous perception network composed of multi-modal sensors; A section dynamic risk level assessment module for calculating a section risk index according to the road structural behavior parameters and a risk assessment model, and obtaining a section risk level according to the section risk index; A data transmission and communication module for low-latency transmission of abnormal events and low-power consumption transmission of regular events using a hierarchical communication protocol, and for spatio-temporal alignment of multi-source heterogeneous data generated by the heterogeneous perception network based on three-dimensional geographic grid coding and precise time protocol; A node adjustment and resource allocation module for adjusting the activation status and sampling frequency of each sensor in the heterogeneous perception network according to the section risk level; and for simulating and verifying the implementation effects of various resource allocation schemes under different section risk levels through a constructed digital twin model of the road system, and then obtaining an optimal resource allocation scheme using a multi-objective optimization algorithm.

2. The road condition adaptive topology perception and application system according to claim 1, characterized in that The section dynamic risk level assessment module is also used for dynamically adjusting the weights in the risk assessment model according to historical monitoring data and an LSTM-GRU hybrid network model; wherein, the LSTM-GRU hybrid network model includes an input layer, an LSTM hidden layer, a fully connected layer, and an output layer, the input layer is used for receiving the historical monitoring data, the LSTM hidden layer includes three layers of LSTM cells, each layer of LSTM cells includes 128 neurons, each neuron has a forget gate, an input gate, and an output gate structure, the fully connected layer includes a first fully connected layer and a second fully connected layer, the first fully connected layer includes 64 neurons, the second fully connected layer includes 32 neurons, and the output layer is used for outputting environmental prediction parameters and weight adjustment coefficients for a future period of time.

3. The road condition adaptive topology perception and application system according to claim 2, characterized in that, During the training stage of the LSTM-GRU hybrid network model, the section dynamic risk level assessment module uses an AdamW optimizer to achieve adaptive learning rate adjustment and weight decay, and configures an early stopping mechanism to prevent overfitting. The loss function of the LSTM-GRU hybrid network is a Focal Loss function.

4. The road condition adaptive topology perception and application system according to claim 2 or 3, characterized in that The LSTM-GRU hybrid network model assigns different weights to prediction results at different time scales through an attention mechanism for short-term, medium-term, and long-term predictions, and the output of the LSTM-GRU hybrid network model also includes the confidence interval of the prediction results.

5. The road condition adaptive topology perception and application system according to claim 1, characterized in that, The node adjustment and resource allocation module is also used for calculating the contribution degree of each sensor in the heterogeneous perception network to risk assessment based on historical monitoring data, and using an ensemble learning method to calculate the information gain of different sensor combinations to obtain a set of high-value perception nodes, and activating each node in the set of high-value perception nodes according to the real-time section risk level.

6. The road condition adaptive topology perception and application system according to claim 1, characterized in that The data transmission and communication module is also used for acquiring timestamp messages sent by each node in the heterogeneous perception network, calculating the transmission delay of each node based on the system main clock source, compensating the timestamp messages of each node according to the transmission delay, and using the compensated time as the data acquisition time tag of each node.

7. The road condition adaptive topology perception and application system according to claim 1, characterized in that The system further includes: An application service module, which is used to map the road section risk level to specific service actions and perform cross-departmental service closed-loop linkage.

8. A method for road condition adaptive topology perception and application, characterized in that, The method includes: Step S1: Obtain the road structural state parameters monitored in real time by a heterogeneous perception network composed of multi-modal sensors; Step S2: Calculate the road section risk index according to the road structural state parameters and the risk assessment model, and obtain the road section risk level according to the road section risk index; Step S3: Use a hierarchical communication protocol to perform low-latency transmission of abnormal events and low-power consumption transmission of regular events, and perform spatio-temporal alignment on the multi-source heterogeneous data generated by the heterogeneous perception network based on three-dimensional geographic grid coding and precise time protocol; Step S4: Adjust the activation state and sampling frequency of each sensor in the heterogeneous perception network according to the road section risk level, use the constructed digital twin model of the road system to simulate and verify the implementation effects of various resource allocation schemes under different road section risk levels, and use a multi-objective optimization algorithm to obtain the optimal resource allocation scheme.

9. The road condition adaptive topology perception and application method according to claim 8, wherein The expression of the risk assessment model is: ； In the formula, RI is the comprehensive risk index of the road section; G is the subgrade deformation rate, and its value range is 0-1; A is the asphalt layer fatigue damage index, and its value range is 0-1; B is the base layer void risk index, and its value range is 0-1; α, β, and γ are all weight coefficients; among them, the calculation formula of the subgrade deformation rate is: ； In the formula, is the cumulative subgrade settlement during continuous monitoring (mm), is the monitoring time interval (days), is the critical deformation rate threshold (mm / day); The calculation formula of the asphalt layer fatigue damage index is: ； In the formula, is the actual number of load applications at the strain level, is the bottom tensile strain of the asphalt layer (με), is the elastic modulus of the asphalt mixture (MPa), is an experimental parameter related to the characteristics of the asphalt material; The calculation formula of the base layer void risk index is: ； In the formula, is the base bottom pressure monitored by the pressure sensor array exceeding the void pressure threshold area (kPa), is the total monitoring area (m 2 ).

10. The method for road condition adaptive topology perception and application according to claim 8, characterized in that The steps of three-dimensional geographic grid coding include: Step S301: Perform three-dimensional grid division on the road and its surrounding environment to form basic spatial units; Step S302: Adopt a hierarchical nested coding strategy, apply the Geohash or H3 algorithm in the horizontal direction to generate a plane coordinate code, and superimpose the Z-axis hierarchical code in the vertical direction to obtain the unique geographic code of each basic spatial unit; Step S303: Map the physical positions of the sensors in the heterogeneous perception network to the corresponding three-dimensional grid cells, and use the geographic code of the three-dimensional grid cells as the spatial label of the data collected by the sensors.

Images