Highway network operation state monitoring and service level intelligent evaluation method and system

Abstract

The invention discloses a highway network operation state monitoring and service level intelligent evaluation method and system, and belongs to the field of intelligent traffic, and the method comprises the steps: collecting video bayonet, ETC portal, mobile phone signaling and Internet map data to form multi-source heterogeneous traffic data, and collecting meteorological data; executing space-time registration, sparse completion and confidence fusion processing to generate road section traffic state data; a meteorological sensitivity coefficient matrix is constructed to analyze the correlation degree of meteorological conditions and service efficiency; calculating a basic service level index and superposing the basic service level index with the meteorological correction factor to generate a comprehensive service level index; according to the method, the state estimation problem that common road monitoring facilities are low in density and sparse in data is solved, and dynamic analysis of the correlation degree between the meteorological condition and the road network service efficiency is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method and system for monitoring the operational status and intelligent assessment of service levels of highway networks. Background Technology

[0002] With the continuous expansion of my country's highway network, the monitoring of the operational status and the assessment of the service level of ordinary highways, as important transportation infrastructure connecting urban and rural areas and serving people's livelihoods, have become core issues of concern for traffic management departments. Compared with expressways, ordinary highways are characterized by low monitoring facility density, complex road network structure, and significant influence from weather conditions. Traditional methods of traffic condition perception and assessment are insufficient to meet the needs of refined management.

[0003] In existing technologies, multi-source data fusion methods have made some progress in the field of traffic state estimation. For example, some studies have used Kalman filtering, neural networks, and other methods to fuse data from fixed detectors and floating car data to achieve rapid estimation of road segment travel time. However, these methods are mainly designed for scenarios with dense monitoring facilities, such as highways or urban expressways. For ordinary roads with sparse monitoring facilities and discontinuous data sampling, they lack effective sparse data completion mechanisms and dynamic confidence adjustment strategies.

[0004] In terms of service level assessment, existing methods are mostly based on static indicator systems provided by standards such as the "Highway Engineering Technical Standards," using single or combined indicators such as speed, density, and delay to classify service level levels. This assessment method has the following limitations: First, the assessment indicator system lacks quantitative consideration of the reliability of travel time, making it difficult to reflect the stability and predictability of traffic flow; second, the assessment results do not consider the dynamic impact of meteorological conditions on traffic capacity and service efficiency, failing to provide a scientific basis for traffic control under extreme weather conditions; third, the assessment cycle is usually at the daily or weekly level, which is insufficient to support real-time traffic management decisions.

[0005] In terms of meteorological impact analysis, existing studies have shown that meteorological conditions such as rainfall, fog, and snow cover have a significant impact on road capacity and driving speed. For example, road capacity can be reduced by 10% to 30% under heavy rainfall conditions, and a sudden drop in visibility under fog conditions may lead to a decrease in vehicle speed of more than 40%. However, current technology has not yet established a dynamic analysis model for the correlation between meteorological conditions and service efficiency, and lacks a quantitative assessment method for meteorological sensitivity specific to the characteristics of ordinary highways.

[0006] In summary, existing technologies for monitoring and evaluating ordinary highway networks suffer from technical problems such as sparse monitoring data that is difficult to integrate effectively, lack of dynamic quantitative indicators for service level assessment, and lack of correlation analysis of meteorological impacts. There is an urgent need for an intelligent evaluation method and system that can integrate multi-source heterogeneous data, accurately estimate traffic conditions in sparse perception scenarios, establish correlation analysis between meteorological conditions and service efficiency, and output a dynamic quantitative service level index.

[0007] It is important to note that there are fundamental differences in traffic flow characteristics between ordinary roads and expressways. Expressways operate in a fully enclosed manner, resulting in relatively stable traffic flow, dense monitoring facilities, and relatively comprehensive data collection. Ordinary roads, on the other hand, face challenges such as interference from at-grade intersections, frequent road openings, and complex traffic compositions, leading to greater traffic flow volatility and making their operational status more difficult to predict. Furthermore, ordinary roads often traverse areas with complex terrain, making them susceptible to local weather conditions, such as fog in mountainous areas and crosswinds in canyons. These factors make the level of service assessment (LSA) of ordinary roads more complex than that of expressways. Therefore, it is necessary to develop specialized monitoring and assessment technologies for ordinary roads, rather than simply applying mature solutions from the expressway sector. Summary of the Invention

[0008] To address the technical problems of existing technologies, such as low density of monitoring facilities on ordinary highways leading to data sparsity, lack of dynamic quantitative indicators for service level assessment, and lack of correlation analysis of meteorological impacts, this invention provides a method and system for monitoring the operational status and intelligently assessing the service level of highway networks.

[0009] To achieve the above objectives, this invention provides a method for monitoring the operational status and intelligently assessing the service level of a highway network, comprising the following steps:

[0010] The multi-source data acquisition process involves collecting video checkpoint data, ETC gantry data, mobile phone signaling data, and internet map data to form multi-source heterogeneous traffic data, and collecting meteorological station observation data, meteorological radar echo data, satellite remote sensing data, and road condition sensor data to form meteorological environmental data.

[0011] The data fusion processing steps involve performing spatiotemporal registration on multi-source heterogeneous traffic data to generate spatiotemporal registered data, performing sparse completion on road segments with insufficient monitoring facility coverage in the spatiotemporal registered data to generate sparse completed data, and calculating confidence weight values ​​for the spatiotemporal registered data and sparse completed data based on the sampling coverage, timeliness, and historical accuracy of each data source, and then performing weighted fusion processing to generate road segment traffic status data for each road segment.

[0012] The meteorological sensitivity analysis steps involve extracting the combined feature vectors of meteorological elements from meteorological environmental data, analyzing the service level attenuation patterns of each road segment under different meteorological conditions, and constructing a meteorological sensitivity coefficient matrix.

[0013] The service level assessment steps involve calculating the speed ratio, travel time reliability, and congestion duration percentage for each road segment based on traffic status data. A basic service level index is generated for each road segment based on these indicators. A meteorological correction factor is calculated based on the meteorological sensitivity coefficient matrix and current meteorological conditions. The meteorological correction factor is then superimposed on the basic service level index to generate a comprehensive service level index.

[0014] The road network analysis and prediction steps involve calculating the road network efficiency index based on the comprehensive service level index and the traffic volume weighting of each road segment, identifying key bottleneck road segments based on congestion propagation characteristics and queue length evolution patterns to generate bottleneck identification result data, and using a time-series deep learning model to integrate historical traffic flow patterns and meteorological forecast information to predict the service level change trend of each road segment to generate service level prediction data.

[0015] This invention also provides a highway network operation status monitoring and service level intelligent assessment system, including a data acquisition module, a data fusion module, a meteorological sensing module, a service level assessment module, and a road network analysis module. The data acquisition module collects multi-source heterogeneous traffic data and meteorological environmental data. The data fusion module performs spatiotemporal registration processing, sparse completion processing, and confidence-weighted fusion processing to generate road segment traffic status data. The meteorological sensing module extracts combined feature vectors of meteorological elements and constructs a meteorological sensitivity coefficient matrix. The service level assessment module calculates the basic service level index and meteorological correction factors to generate a comprehensive service level index. The road network analysis module calculates the road network efficiency index and generates bottleneck identification results and service level prediction data.

[0016] The present invention has the following beneficial effects:

[0017] First, this invention integrates four heterogeneous data sources—video checkpoints, ETC gantries, mobile signaling, and internet maps—and establishes a three-tiered data processing mechanism of spatiotemporal registration, sparse completion, and confidence fusion. This effectively solves the problem of traffic state estimation in scenarios with sparse monitoring facilities on ordinary highways, improving the accuracy of traffic state estimation for each road segment by 15% to 25% compared to single-data-source methods. This multi-source fusion method fully utilizes the complementary advantages of each data source: video checkpoints provide high-precision cross-sectional flow, ETC gantries provide travel time anchor points, mobile signaling supplements monitoring blind spots, and internet maps provide a road network-level reference benchmark. The four types of data mutually verify and complement each other, achieving comprehensive perception of road network traffic conditions.

[0018] Secondly, this invention innovatively introduces a meteorological sensitivity coefficient matrix to establish a nonlinear mapping quantitative model of precipitation intensity, visibility, and speed reduction. This enables dynamic analysis of the correlation between meteorological conditions and road network service efficiency, providing a quantitative basis for traffic control decisions under extreme weather conditions. Traditional service level assessment methods neglect the influence of meteorological factors, leading to significant discrepancies between assessment results and actual driving experiences under extreme weather conditions. This invention, through meteorological sensitivity analysis and a meteorological correction mechanism, enables the assessment results to accurately reflect the dynamic impact of changes in meteorological conditions on road network service efficiency, significantly improving the consistency between assessment results and drivers' subjective experiences.

[0019] Third, this invention employs a two-tiered assessment framework combining a basic service level index (LSI) with a meteorological correction factor. This framework organically integrates static service level indicators with dynamic meteorological influences, resulting in assessments that are significantly more timely and accurate than traditional single-tiered assessment methods. The two-tiered framework decomposes service level assessment into two levels: endogenous factor assessment and exogenous factor correction. The LSI reflects the traffic flow operation status of the road segment itself, while the meteorological correction factor reflects the degree of influence of external meteorological conditions. The two are superimposed to form a comprehensive assessment result. This layered approach makes the assessment mechanism clearer and more transparent, facilitating subsequent problem diagnosis and optimization.

[0020] Fourth, this invention integrates historical traffic flow patterns with weather forecast information using a time-series deep learning model to accurately predict the service level changes of various road sections over the next 2 to 24 hours, achieving a prediction accuracy of over 85%. This provides proactive decision-making support for traffic management departments. Predictive capability is the core value of intelligent transportation systems. Traditional traffic management is primarily reactive, taking control measures only after congestion occurs. This invention, through accurate trend prediction, enables traffic management departments to take preventative measures before congestion occurs, such as issuing traffic guidance information, adjusting signal timing, and activating emergency plans, significantly improving the initiative and effectiveness of traffic management. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method for monitoring the operation status and intelligently assessing the service level of highway networks provided in this embodiment of the invention.

[0022] Figure 2 This is an architecture diagram of the intelligent assessment system for monitoring the operation status and service level of highway networks provided in this embodiment of the invention. Detailed Implementation

[0023] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that the following embodiments are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.

[0024] This invention provides a method for monitoring the operational status and intelligently assessing the service level of highway networks, such as... Figure 1 As shown, the method includes a multi-source data acquisition step S1, a data fusion processing step S2, a meteorological sensitivity analysis step S3, a service level assessment step S4, and a road network analysis and prediction step S5. Each step is described in detail below.

[0025] Step S1: Multi-source data acquisition steps.

[0026] The core objective of this step is to construct a multi-source heterogeneous data acquisition system covering the ordinary highway network, providing fundamental data support for subsequent data fusion and service level assessment. Ordinary highways, as crucial transportation infrastructure connecting urban and rural areas, exhibit significantly different traffic flow characteristics compared to expressways, including more complex traffic composition, frequent intersection interference, and lower monitoring facility density. Therefore, establishing a multi-source heterogeneous data acquisition system is of great significance for achieving refined management of the ordinary highway network. Preferably, this invention divides data acquisition into two subsystems: traffic data acquisition and meteorological data acquisition. These two subsystems are independent yet complementary, jointly supporting subsequent data fusion and service level assessment.

[0027] From a data quality assurance perspective, this invention establishes a multi-level quality control mechanism during the data acquisition phase. At the data source level, all types of sensors are equipped with self-testing functions, enabling real-time monitoring of equipment operating status and automatic reporting of early warning information when a fault occurs. At the data transmission level, a redundant transmission channel design is adopted, with an automatic switching mechanism between primary and backup links ensuring the continuity and reliability of data transmission. At the data aggregation level, a data integrity verification step is set up to mark and supplement missing and abnormal data. This end-to-end quality control mechanism lays a solid foundation for the accuracy of subsequent data fusion and service level assessment.

[0028] In terms of traffic data collection, this invention integrates four heterogeneous data sources: video checkpoint data, ETC gantry data, mobile signaling data, and internet map data, to construct a dynamic perception system for traffic flow and speed on ordinary highway networks. These four data sources differ significantly in terms of collection granularity, coverage, and data characteristics, but they are also complementary. Through reasonable data fusion, a comprehensive perception of the road network's traffic status can be achieved. From the perspective of data source characteristics, video checkpoint data has high-precision cross-sectional flow collection capabilities but limited spatial coverage; ETC gantry data can provide accurate interval travel times but only covers vehicles equipped with ETC devices; mobile signaling data has broad spatial coverage but relatively low positioning accuracy; and internet map data can provide road network-level situational awareness but its timeliness and accuracy are uncertain. This invention fully utilizes the complementary characteristics of these four data sources, achieving complementary advantages and compensating for weaknesses through subsequent data fusion processing.

[0029] Video checkpoints collect vehicle passage records and vehicle type information, serving as a primary means of obtaining traffic flow data for road sections. In one embodiment of this invention, the deployment spacing of video checkpoints is differentiated according to the road section level. Considering the balance between construction costs and monitoring accuracy, the average deployment spacing for national highways is set to 8km to 15km, and for provincial highways, it is set to 12km to 20km. This differentiated deployment strategy ensures monitoring coverage of major routes while also considering economic efficiency. The video checkpoint equipment automatically extracts information such as vehicle passage time, license plate number, and vehicle type using image recognition technology. Vehicle types are categorized and statistically analyzed according to categories such as passenger vehicles, freight vehicles, and special-purpose vehicles, providing basic data for subsequent traffic flow calculation. Preferably, the video checkpoint data collection cycle is set to 1 minute, and the raw data is uploaded to the data center after noise reduction and outlier filtering. From the perspective of image recognition technology, this invention uses a deep learning object detection algorithm for vehicle recognition, achieving an accuracy rate of over 98%, maintaining an accuracy rate of over 95% even at night or in poor lighting conditions. The license plate recognition system uses an end-to-end recognition network, supporting the recognition of various license plate types such as new energy vehicle license plates and ordinary vehicle license plates, with a recognition accuracy rate of over 99%.

[0030] ETC gantry data reuse utilizes existing infrastructure along the route, collecting electronic tag transaction records to construct an anchor sequence for calculating interval travel time. This invention fully leverages the achievements of the nationwide ETC network construction project in recent years. By reading the passage records of on-board electronic tags, it can accurately obtain the travel time of vehicles between adjacent gantries. Unlike the cross-sectional detection of video checkpoints, ETC gantry data can provide travel information at the vehicle trajectory level, which is of great value for calculating interval average speed and travel time variability. In one embodiment of this invention, the processing of ETC gantry data includes three stages: electronic tag encoding matching, passage time calibration, and abnormal travel filtering. Abnormal travel filtering uses the median absolute deviation method of travel time to remove abnormal records that deviate from the median by more than three standard deviations. From the perspective of travel time calculation, this invention establishes logical connections between adjacent gantries by constructing a gantry sequence map, supporting the estimation of travel time between any two gantries. For vehicles without ETC devices, this invention uses virtual tag technology to associate video checkpoint data with ETC gantry data through license plate recognition, expanding the coverage of the travel time sample. Preferably, when the ETC penetration rate reaches 60% or more, statistically significant travel time distribution estimates can be obtained using only ETC data; when the ETC penetration rate is low, the present invention automatically enables a virtual tag association mechanism to supplement the sample size.

[0031] Mobile phone signaling data obtains anonymized location information through operator interfaces to supplement traffic state estimation in road sections with sparse monitoring facilities. Ordinary roads have significantly lower monitoring facility density than expressways, resulting in numerous monitoring blind spots. Mobile phone signaling data, as a ubiquitous sensing method, can estimate traffic state in these blind spots at a relatively low cost. In one embodiment of this invention, the positioning accuracy of the mobile phone signaling data ranges from 100m to 500m, and the sampling period is typically 5 to 15 minutes. It should be noted that the mobile phone signaling data undergoes anonymization processing by the operator, retaining only location and time information, and does not involve user privacy data. This invention estimates traffic flow speed and traffic density of road sections by reconstructing the trajectory of signaling positioning sequences from consecutive time points. Preferably, the mobile phone signaling data processing employs a Hidden Markov Model for map matching, matching the original positioning point sequence to the road network topology, achieving a matching accuracy of over 90%. Based on the technical principles of mobile signaling data processing, this invention first cleans and preprocesses the raw signaling data to remove abnormal data such as ping-pong handover and signal drift. Then, a map matching algorithm based on state transition probability is used to map the location point sequence onto the road network. Finally, traffic state parameters for road segments are generated through speed estimation and traffic flow estimation algorithms. To improve estimation accuracy, this invention introduces a multi-operator data fusion mechanism, which expands the sample coverage by integrating signaling data from multiple operators. Under typical conditions, the sample coverage can reach 8% to 15% of the target road traffic volume.

[0032] Internet map data obtains road segment congestion indices and estimated travel times through open interfaces, updated every 2 to 5 minutes, providing a road network-level traffic situation reference benchmark. Internet map platforms, by aggregating trajectory data from a large number of navigation users, can calculate the congestion status and estimated travel time of each road segment in real time. This invention uses internet map data as a road network-level traffic situation reference benchmark to verify and supplement the estimation results of other data sources. In one embodiment of this invention, the congestion index uses a scale of 0 to 10, where 0 to 2 represents free flow, 3 to 5 represents slow flow, 6 to 8 represents congestion, and 9 to 10 represents severe congestion.

[0033] In terms of meteorological data acquisition, this invention integrates meteorological station observation data, meteorological radar echo data, satellite remote sensing data, and road surface sensor data to construct a meteorological environment perception system covering the ordinary highway network. Compared with expressways, ordinary highways have lower alignment standards and less comprehensive protective facilities, making them more susceptible to extreme weather conditions. Therefore, establishing a refined meteorological environment perception system is of great significance for ensuring traffic safety. From the design concept of the meteorological perception system, this invention adopts a multi-scale, multi-layered perception architecture. Meteorological stations provide precise point-level observations, meteorological radar provides regional spatial coverage, satellite remote sensing provides large-scale situational monitoring, and road surface sensors provide real-time status of key points. These four types of data complement and verify each other, forming a three-dimensional meteorological environment perception network.

[0034] Meteorological station observation data, including ambient temperature, relative humidity, wind speed and direction, precipitation, and visibility, forms the foundational data source for meteorological sensing. In one embodiment of this invention, the collection cycle for conventional meteorological elements is 1 hour, while the collection cycle for real-time warning data automatically switches to 5 minutes when an extreme weather warning is triggered. Meteorological station data is acquired through the meteorological department's data sharing interface and, after quality control and interpolation processing, generates gridded meteorological field data covering the target road network. Preferably, spatial interpolation uses the Kriging interpolation method with an interpolation resolution of 1 km × 1 km. From the perspective of meteorological data processing, this invention addresses the problem of uneven distribution of meteorological stations by introducing a terrain correction factor. When performing spatial interpolation, it comprehensively considers the influence of factors such as altitude, slope aspect, and terrain shading, significantly improving the estimation accuracy of meteorological elements in complex terrain areas. Particularly for mountain roads, terrain has a significant modulating effect on elements such as temperature, wind speed, and visibility, making accurate estimation difficult with traditional simple interpolation methods. This invention introduces digital elevation model (DEM) data and terrain feature factors to construct a meteorological element estimation model that considers terrain influence, reducing the estimation error by more than 30% compared to traditional methods.

[0035] The spatial resolution of the meteorological radar echo data is 1 km, and the temporal resolution is 6 minutes. Short-term precipitation forecasts are achieved by calculating the precipitation intensity level based on the echo intensity values. Meteorological radar can detect the three-dimensional structure and movement trend of precipitation clouds, making it a key data source for short-term precipitation forecasting. In one embodiment of this invention, the conversion relationship between echo intensity and precipitation intensity adopts the ZR relationship, i.e. ,in Radar echo intensity factor, in dBZ. Precipitation intensity, in mm / h, coefficient and Localized calibration is performed based on regional climate characteristics, with typical values ​​taken as follows: , Based on radar echo data, this invention uses optical flow to track the movement trajectory of precipitation clouds, enabling short-term precipitation forecasts for the next 0 to 2 hours.

[0036] Satellite remote sensing data is used to acquire information on large-scale cloud distribution and surface temperature, compensating for coverage blind spots in ground-based observation networks. In one embodiment of this invention, the satellite remote sensing data primarily acquires visible light cloud images and infrared cloud images. Visible light cloud images are used to identify cloud types and coverage areas, while infrared cloud images are used to retrieve cloud top temperatures and estimate precipitation probabilities. The spatial resolution of the satellite data is 1 km to 4 km, and the temporal resolution is 15 min to 30 min.

[0037] Road surface condition sensors are deployed at key locations such as bridge and tunnel entrances / exits to collect parameters including road surface temperature, humidity, friction coefficient, and water depth. These key locations are most prone to traffic accidents under extreme weather conditions and require focused monitoring. In one embodiment of this invention, the road surface condition sensor uses non-contact infrared thermometry to collect road surface temperature with an accuracy of ±0.5℃; it uses a capacitive sensor to detect road surface humidity, distinguishing between four states: dry, wet, waterlogged, and icy; and it uses laser ranging technology to measure water depth, with a measurement range of 0 to 100 mm and a resolution of 1 mm. The data acquisition cycle of the road surface condition sensor is 1 minute, and it automatically triggers an early warning report when an abnormal condition is detected.

[0038] Step S2: Data fusion processing step.

[0039] The core objective of this step is to fuse multi-source heterogeneous traffic data to generate a unified traffic state estimation result for each road segment. This invention employs a three-level processing architecture of spatiotemporal registration, sparse completion, and confidence fusion, effectively solving the technical challenges of heterogeneous, sparse, and asynchronous ordinary highway monitoring data. The design concept of this three-level processing architecture stems from an in-depth analysis of the problem of multi-source heterogeneous data fusion. The first level, spatiotemporal registration, addresses the issue of inconsistent data coordinates; the second level, sparse completion, addresses the issue of incomplete data coverage; and the third level, confidence fusion, addresses the issue of uneven data quality. These three levels progress sequentially, with the output of each level serving as the input for the next, forming a complete data processing pipeline.

[0040] From the perspective of data fusion theory, the core challenges of multi-source heterogeneous data fusion include spatiotemporal asynchrony, differences in sampling granularity, and uneven data quality. Spatiotemporal asynchrony manifests as inconsistencies in the collection time and spatial location of different data sources, requiring unification through spatiotemporal registration. Differences in sampling granularity manifest as variations in the sampling frequency and spatial coverage of different data sources, requiring homogenization through sparse completion. Uneven data quality manifests as differences in the reliability of different data sources under different spatiotemporal conditions, requiring adaptive weighting through confidence fusion. This invention designs corresponding processing modules to address these three core challenges, achieving effective fusion of multi-source heterogeneous data.

[0041] The spatiotemporal registration unit converts the timestamps of various data sources into a standard time base and projects the spatial location onto the road network linear reference system. Road segment division is based on adjacent intersections as the basic unit. Spatiotemporal registration is a prerequisite for multi-source data fusion; only by unifying the spatiotemporal coordinates of different data sources to the same reference system can meaningful fusion processing be performed. In one embodiment of this invention, the time base is Coordinated Universal Time (UTC). The timestamps of all data sources are first converted to UTC time, and then aggregated and statistically analyzed according to a set fusion time window (usually 5 minutes or 15 minutes). Spatial registration uses the road network linear reference system, that is, with the route starting point as the origin, the cumulative mileage measured along the route direction is used as the location coordinates. Road segment division is based on adjacent intersections as the basic unit. For excessively long road segments (over 10 km), subdivision is performed according to the principle of equal spacing to ensure the uniformity of the evaluation granularity. From the perspective of the technical implementation of spatiotemporal registration, this invention designs a three-level spatiotemporal registration architecture: the first level is coarse registration, which converts the original coordinates of each data source into a unified geographic coordinate system (WGS84) and timestamps into a unified time zone; the second level is fine registration, which projects the geographic coordinates onto the road network topology space and aggregates the timestamps into a set fusion window; the third level is verification registration, which checks the consistency of different data sources at the same spatiotemporal location through cross-validation, and identifies and marks abnormal data. This three-level registration architecture can effectively handle various abnormal situations and ensure the accuracy and reliability of the registration results.

[0042] The sparse sensing data completion unit addresses data gaps caused by insufficient monitoring facility coverage by employing a combined approach of temporal interpolation and spatial inference for state estimation. The monitoring facility density on ordinary roads is far lower than on expressways, resulting in numerous road sections lacking direct observation data, necessitating indirect estimation through data completion methods. In one embodiment of this invention, cubic spline interpolation is used for temporal interpolation to estimate traffic parameters at missing times. This method ensures the continuity and smoothness of the interpolation curve and is suitable for the periodic variations in traffic flow. Spatial inference is based on traffic flow propagation theory, utilizing the correlation between traffic states of adjacent road sections. From the theoretical foundation of data completion, traffic flow exhibits spatiotemporal continuity and correlation, providing a theoretical basis for data completion. Temporally, traffic flow typically exhibits periodic variations, with the traffic state at the same moment correlated with those at previous moments. Spatially, the traffic states of adjacent road sections are influenced by traffic flow propagation, exhibiting causal relationships. This invention fully utilizes these spatiotemporal characteristics of traffic flow, designing a spatiotemporal joint completion algorithm. By comprehensively utilizing information from both temporal and spatial dimensions, it significantly improves the accuracy of data completion. Preferably, for points where there are missing temporal and spatial neighborhood observation data, the present invention adopts a weighted fusion method to integrate the estimation results of the two dimensions, and the fusion weight is dynamically determined according to the quality and relevance of the neighborhood data.

[0043] In one embodiment of the present invention, the spatial dimension inference is specifically implemented using a weighted inference method for the target road segment. Traffic conditions Based on its set of adjacent road segments Estimate based on observational data:

[0044] ,

[0045] in, For the target road section Traffic condition estimates can be parameters such as speed or flow rate, with the unit determined based on the specific parameter. Adjacent road sections Observed traffic status values; For the target road section The set of adjacent road segments, including upstream and downstream road segments; For road section For the section The inference weight, ranging from 0 to 1, is calculated based on the distance decay between road segments and the similarity of road segment attributes. Preferably, the formula for calculating the inference weight is: ,in For road section and The topological distance between them, in km. This is the distance attenuation factor, typically taken as 5km. The similarity of road segment attributes is calculated by comprehensively considering the similarity of attributes such as road grade, number of lanes, and speed limit, with a value ranging from 0 to 1. The technical advantage of this formula is that the closer the road segments are and the more similar their attributes are, the greater the reference value of their traffic conditions for the target road segment. The exponential decay function achieves a smooth transition, avoiding unreasonable influences from distant road segments on the estimation results.

[0046] The multi-source fusion weights are dynamically adjusted based on sampling coverage, timeliness, and historical accuracy. Different data sources exhibit varying reliability under different spatiotemporal conditions. This invention employs a dynamic confidence weight mechanism to adaptively adjust the fusion weights based on the actual performance of each data source, thereby achieving the optimal fusion result. In one embodiment of this invention, the confidence weight is calculated using the following formula:

[0047] ,

[0048] in, For the first The fusion weight of each data source ranges from 0 to 1, and the sum of the weights of all data sources is 1. For the first The sampling coverage of a data source is defined as the ratio of the number of valid samples of that data source within the target road segment and time window to the theoretical maximum number of samples, with a value ranging from 0 to 1. For the first The timeliness factor of a data source is defined as a negative exponential function of the difference between the data acquisition time and the fusion calculation time, with a value range of 0 to 1. The timeliness factor reflects the freshness of the data, and the newer the data, the closer the timeliness factor is to 1. For the first The historical accuracy of a data source is calculated based on the error statistics between the historical estimate of that data source and the true value (such as through manual observation or verification by a more reliable data source), and the value ranges from 0 to 1; , , The weighting coefficients for the three indicators are, in this invention, preferably set to a value of [value to be filled in]. , , This weighting configuration comprehensively considers the importance of sampling coverage to data representativeness, as well as the contribution of timeliness and historical accuracy to data reliability. The total number of data sources, in this invention The technical effect of this formula is that data sources with high sampling coverage, good timeliness, and high historical accuracy receive greater fusion weights, while the weights of data sources with poor quality are automatically reduced, thereby achieving adaptive fusion optimization of data quality.

[0049] The final output of the fusion processing is the traffic status data for each road segment, including core parameters such as average speed, traffic flow, and occupancy. In one embodiment of the present invention, the fusion calculation of the average speed of road segments adopts a weighted average method:

[0050] ,

[0051] in, For road section The average speed after fusion, in km / h; For the first Road segments provided by data sources Speed ​​estimate, in km / h; For the first The weights of each data source are merged. When a data source has no valid data in a specific road segment, its weight is automatically set to zero and the weights of other data sources are renormalized.

[0052] Step S3: Meteorological sensitivity analysis steps.

[0053] The core objective of this step is to establish a correlation analysis model between meteorological conditions and road network service efficiency, providing a quantitative basis for subsequent meteorological adjustments to service levels. This invention innovatively proposes the concept and calculation method of a meteorological sensitivity coefficient matrix, addressing the vulnerability of ordinary highways to extreme weather events such as heavy rainfall and fog. The meteorological sensitivity coefficient matrix is ​​one of the core innovations of this invention; it transforms traditional qualitative analysis of meteorological impacts into quantitative assessment, enabling precise quantification of the impact of meteorological conditions on traffic operations and their application in service level assessment.

[0054] From the perspective of traffic meteorology theory, the impact mechanism of meteorological conditions on road traffic can be analyzed from two dimensions: driver behavior and road conditions. In terms of driver behavior, drivers will proactively reduce their speed to ensure safety under adverse weather conditions. This behavioral response exhibits significant non-linear characteristics; that is, drivers will significantly adjust their driving behavior after weather conditions deteriorate to a certain extent. In terms of road conditions, precipitation leads to a decrease in road surface friction coefficient and an increase in the risk of water accumulation; fog significantly reduces visibility; and icing severely affects driving stability. These factors directly impact the actual traffic capacity of roads. The meteorological sensitivity coefficient matrix of this invention comprehensively considers the impact mechanisms of the above two dimensions, and can fully reflect the combined impact of meteorological conditions on traffic operations.

[0055] Preferably, the update cycle of the meteorological sensitivity coefficient matrix is ​​set to a quarterly level to adapt to changes in meteorological characteristics across different seasons. Simultaneously, this invention also establishes a rapid response mechanism for extreme weather events. When a historically rare extreme weather event occurs, the system can collect traffic status data during the event in real time and quickly update the sensitivity coefficients of the affected road sections, ensuring that the assessment model can adapt to new meteorological risks.

[0056] The meteorological impact factor extraction unit constructs combined feature vectors of meteorological elements for four typical extreme weather types: heavy rainfall, dense fog, road snow and ice accumulation, and crosswinds. Different types of extreme weather affect road traffic through different mechanisms, requiring different feature vectors for representation. In one embodiment of this invention, the feature vectors for various extreme weather types are constructed as follows:

[0057] Feature vectors of heavy rainfall type It includes three components: precipitation intensity The unit is mm / h, reflecting the intensity of the current precipitation; the duration of precipitation. The unit is h, reflecting the impact of cumulative precipitation on road surface condition; road surface water depth. The unit is mm, reflecting the road surface drainage capacity and driving safety risks. Preferably, the heavy rainfall warning levels are divided into: light rain (…). mm / h), moderate rain ( mm / h), heavy rain ( mm / h), heavy rain ( mm / h).

[0058] Feature vectors of fog type It includes three components: visibility The unit is meters (m), and it is the most critical factor affecting driving safety; relative humidity. The unit is %, reflecting the meteorological conditions for fog formation and maintenance; temperature-dew point difference. The unit is °C, reflecting the degree of water vapor saturation near the ground. Preferably, the fog warning level is divided into: light fog ( m), dense fog ( m), dense fog ( m), dense fog ( m).

[0059] Feature vectors of road snow and ice accumulation types It includes three components: road surface temperature The unit is °C, which is the core indicator for judging the possibility of freezing; ambient temperature. The unit is °C, which affects the trend of road surface temperature change; precipitation type , which are categorical variables, including four categories: no precipitation, precipitation, sleet, and snowfall. Different precipitation types have different effects on road surface conditions under low temperature conditions.

[0060] crosswind type feature vector Includes two components: wind speed The unit is m / s, reflecting the wind force; the angle between the wind direction and the road direction. The unit is degrees. Crosswinds (angles close to 90°) have the greatest impact on vehicle stability. Preferably, crosswind warning levels are divided into: mild (…). m / s), moderate ( m / s), severe ( m / s and ).

[0061] The weather sensitivity calculation unit analyzes the service level attenuation patterns of various road sections under different meteorological conditions and establishes a meteorological sensitivity coefficient matrix. The meteorological sensitivity coefficient reflects the degree of service level attenuation of a specific road section under specific meteorological conditions. Due to differences in road geometry, drainage capacity, and protective facilities, different road sections may exhibit significantly different sensitivities to the same meteorological conditions. In one embodiment of this invention, the meteorological sensitivity coefficient matrix... elements Defined as:

[0062] ,

[0063] in, For road section In the The sensitivity coefficient under similar weather conditions ranges from 0 to 1. The larger the value, the more sensitive the road section is to such weather conditions, and the more obvious the service level degradation. For road section The baseline service level index under normal weather conditions ranges from 0 to 100. For road section In the The actual service level index under similar meteorological conditions ranges from 0 to 100; subscript The value range is 1 to , The total number of road segments in the road network; subscript The value range is 1 to , The number of meteorological condition types in this invention These correspond to four categories: heavy rainfall, dense fog, road snow and ice accumulation, and crosswinds.

[0064] The meteorological sensitivity coefficient matrix is ​​obtained through statistical analysis of historical data. In one embodiment of the present invention, traffic status data of each road segment under different meteorological conditions are first extracted from a historical database. Then, the average service level under various meteorological conditions is statistically analyzed. Finally, the proportion of the difference between the service level and the service level under normal weather conditions is calculated. Preferably, the historical data statistical period is not less than one year to cover typical samples of various meteorological conditions. For newly built road segments or road segments with insufficient historical data, the meteorological sensitivity coefficient can be estimated using the analogy method of similar road segments.

[0065] This invention specifically addresses the significant impact of heavy rainfall and fog on ordinary highways, establishing a quantitative model of the relationship between precipitation intensity and visibility on speed reduction. Precipitation and visibility are two of the most critical meteorological factors affecting the safety and efficiency of ordinary highway traffic. This invention uses a nonlinear mapping model to accurately describe their speed reduction effect.

[0066] The mapping model between precipitation intensity and velocity reduction factor is as follows:

[0067] ,

[0068] in, The precipitation speed reduction factor is dimensionless and ranges from 0.50 to 1.0. It represents the ratio of the speed a vehicle can maintain under a specific precipitation intensity to the speed under normal weather conditions. The value represents precipitation intensity, expressed in mm / h. The technical advantages of this piecewise linear model are as follows: precipitation intensity below 5 mm / h has virtually no impact on vehicle speed; speed reduction is relatively gradual under light rain conditions of 5 to 10 mm / h; speed reduction accelerates under moderate rain conditions of 10 to 25 mm / h; speed reduction is significant under heavy rain conditions of 25 to 50 mm / h; and speed is reduced to 50% of normal under torrential rain conditions. The model parameters are calibrated based on actual traffic flow observation data, reflecting the speed selection behavior patterns of drivers under different precipitation conditions.

[0069] The mapping model between visibility and speed reduction factor is as follows:

[0070] ,

[0071] in, Visibility speed reduction factor, dimensionless, with a value range of 0.40 to 1.0; Visibility is measured in meters (m). The technical advantages of this piecewise linear model are as follows: visibility above 1000m has virtually no impact on driving speed; speed reduction is relatively gradual in light fog conditions (200-1000m); speed reduction is significant in dense fog conditions (50-200m); and speed is reduced to 40% of normal in very dense fog conditions (visibility below 50m). Visibility has a more dramatic impact on speed than precipitation because it directly affects the driver's safe perception distance and reaction time.

[0072] Step S4: Service Level Assessment Steps.

[0073] The core objective of this step is to calculate the comprehensive service level index for each road segment based on the fused road segment traffic status data and meteorological sensitivity analysis results. This invention employs a two-tiered evaluation framework that combines a basic service level index with a meteorological correction factor. This framework retains the core indicator system of traditional service level assessments while innovatively introducing quantitative corrections for meteorological impacts, thus achieving an organic integration of static evaluation and dynamic correction.

[0074] From the development history of road service level assessment (LSA) theory, traditional LSAs are mainly based on the speed-density-flow relationship in traffic flow theory, defining service level grades by delineating different traffic flow state intervals. While this assessment method is simple and clear, it has three limitations: first, it only considers the immediate traffic flow state, neglecting the important service quality dimension of travel time reliability; second, it uses static grading standards, failing to reflect the dynamic impact of external factors such as weather conditions on service levels; and third, the assessment cycle is usually long, making it difficult to support real-time traffic management decisions. This invention systematically improves upon these limitations by introducing travel time reliability and congestion duration percentage indicators to enrich the assessment dimensions, implementing dynamic assessment through a weather correction mechanism, and supporting minute-level real-time assessment through optimized computing architecture.

[0075] The core design concept of the two-tiered assessment architecture is to classify the factors influencing service levels into two categories: endogenous and exogenous factors. Endogenous factors refer to service level influencing factors determined by the characteristics of traffic flow itself, including traffic volume, speed, and density. These factors are characterized by a basic service level index. Exogenous factors refer to service level influencing factors determined by external environmental conditions, primarily meteorological conditions. These factors are characterized by a meteorological correction factor. The two types of factors are superimposed through a multiplicative coupling method to generate a comprehensive service level index. The advantages of this tiered design are: it facilitates the analysis of the causes of service level changes, i.e., it can clearly determine whether a decline in service level is due to worsening traffic flow or deteriorating meteorological conditions; and it facilitates the development of targeted improvement measures, such as traffic control measures for worsening endogenous factors and the issuance of meteorological warnings for the impact of exogenous factors.

[0076] The basic service level calculation subunit calculates three core indicators based on road segment traffic status data. The speed ratio indicator reflects the degree of matching between the actual operating speed and the design capacity of a road segment, and is the most intuitive indicator of service level. In one embodiment of this invention, the speed ratio... The calculation formula is:

[0077] ,

[0078] in, This is a speed ratio, dimensionless, and typically ranges from 0.1 to 1.2. A value greater than 1.0 indicates that the actual speed exceeds the design speed, potentially indicating speeding. The actual average speed of the road segment, in km / h, is output from the data fusion processing step. The design speed for a road section is expressed in km / h. It is determined by referring to a table based on the road's technical grade. For national highways, it is generally 60 to 80 km / h, and for provincial highways, it is generally 40 to 60 km / h.

[0079] Travel time reliability reflects the stability and predictability of traffic flow on a road segment and is an important dimension for evaluating service quality. In one embodiment of the present invention, travel time reliability... The calculation formula is:

[0080] ,

[0081] in, It is a reliability index for travel time, dimensionless, and its value ranges from 1.0 to 3.0. The closer the value is to 1.0, the more stable and reliable the travel time is. The 95th percentile value of the travel time distribution of road segments within the evaluation period, in minutes, reflects the travel time under the most unfavorable conditions; To assess the median travel time distribution of road segments within a given period, the unit is minutes, reflecting travel time under typical conditions. The technical advantage of this indicator is that when... When the value is close to 1.0, it indicates that the travel time is highly stable, and travelers can accurately estimate their arrival time; when... A value significantly greater than 1.0 indicates that travel time fluctuates greatly, and travelers need to allow more buffer time to cope with possible delays.

[0082] The congestion duration percentage indicator reflects the proportion of time a road segment is congested within the assessment period, and is a time-dimensional indicator for measuring service levels. In one embodiment of the present invention, the congestion duration percentage... The calculation formula is:

[0083] ,

[0084] in, This represents the percentage of time spent in traffic congestion, expressed as a percentage, ranging from 0% to 100%. For a moment The average speed of the road segment, in km / h; The design speed for the road section is expressed in km / h. This is an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. To assess the total cycle time, the unit is consistent with the cumulative time step in the denominator. The congestion threshold is set at 40% of the design speed because when the actual speed is lower than 40% of the design speed, the traffic flow has clearly entered a congested state, and the traveler experience is significantly reduced.

[0085] Based on the above three core indicators, this invention calculates the basic service level index for each road segment. In one embodiment of this invention, the basic service level index... The calculation formula is:

[0086] ,

[0087] in, It is a basic service level index, dimensionless, with a value range from 0 to 100. The higher the value, the better the service level. , , The weighting coefficients for the three indicators are, in this invention, preferably set to a value of [value to be filled in]. , , This weighting configuration reflects the dominant role of speed indicators as well as the supplementary roles of reliability and congestion duration; , , This is a scoring mapping function for each indicator, which maps the original indicator values ​​to a standardized scoring range of 0 to 1.

[0088] Preferably, the scoring mapping function for each indicator is defined as follows: Speed ​​Ratio Scoring Function Use piecewise linear mapping: when hour, ;when hour, ;when hour, ;when hour, ;when hour, The mapping function is designed so that a perfect score is obtained when the speed ratio is above 0.9, and a score of zero is obtained when it is below 0.3, with a smooth transition in the middle range.

[0089] Travel time reliability scoring function Using exponential decay mapping: This mapping function ensures a perfect score when the reliability metric is 1.0, with the score dropping rapidly as the reliability metric increases, demonstrating the importance of reliability to the service experience.

[0090] Congestion duration percentage scoring function Using linear mapping: This mapping function ensures that a perfect score is awarded when the congestion duration percentage is 0%, and a score of zero is awarded when the percentage is 100%, with a linear transition in between.

[0091] The basic service level index is divided into six levels, from A to F. In one embodiment of the present invention, the index range corresponding to each level is as follows: Level A corresponds to... Level B indicates smooth traffic flow and a good travel experience; Level B corresponds to... Level C indicates a generally smooth traffic flow and a good travel experience; Level C corresponds to... Level D indicates mild congestion, with a reasonably acceptable travel experience; Level D corresponds to... Level E indicates moderate congestion and a generally poor travel experience; Level E corresponds to... Level F indicates severe congestion and a poor travel experience; Level F corresponds to... This indicates extreme congestion and a very poor travel experience.

[0092] The meteorological correction calculation subunit calculates the meteorological correction factor based on the meteorological sensitivity coefficient matrix and current meteorological conditions. In one embodiment of the present invention, the meteorological correction factor... The calculation formula is:

[0093] ,

[0094] in, This is a meteorological correction factor, dimensionless, with a value range of 0.3 to 1.0. The smaller the value, the greater the negative impact of meteorological conditions on service levels. The precipitation velocity reduction factor is calculated by formula (5); The visibility speed reduction factor is calculated using formula (6); The maximum sensitivity coefficient for a road segment under current meteorological conditions is calculated by taking the coefficient value of the current road segment in the meteorological sensitivity coefficient matrix under the current meteorological type, with a value ranging from 0 to 1. The technical advantage of this formula is that it first takes the smaller value of the reduction coefficients for precipitation and visibility, reflecting the dominant role of the most unfavorable meteorological factor; then it multiplies by a sensitivity adjustment term. This makes weather corrections more significant for highly sensitive road sections.

[0095] The integrated service level calculation subunit superimposes the meteorological correction factor and the basic service level index to generate the integrated service level index. In one embodiment of the present invention, the integrated service level index... The calculation formula is:

[0096] ,

[0097] in, The comprehensive service level index is dimensionless and ranges from 0 to 100. Basic service level index; This is a meteorological correction factor. The technical advantage of this multiplicative superposition method is that when meteorological conditions are favorable ( (Approaching 1), the comprehensive service level is basically consistent with the basic service level; when meteorological conditions are severe ( (Significantly less than 1), the overall service level will decrease accordingly, and the magnitude of the decrease will be proportional to the basic service level and the degree of meteorological impact.

[0098] Step S5: Road network analysis and prediction steps.

[0099] The core objective of this step is to conduct network-level analysis and trend prediction based on the comprehensive service level index of each road segment, providing decision support for traffic management departments. This invention achieves a cross-scale extension from segment-level assessment to network-level analysis through three sub-units: network efficiency assessment, bottleneck identification, and trend prediction.

[0100] From a traffic systems engineering perspective, while service level assessments of individual road segments can reflect local traffic conditions, they struggle to capture the overall operational status and development trends of the road network. Traffic management decisions often require a holistic perspective, comprehensively considering the relationships and resource constraints between different road segments. Therefore, network-level analysis has significant practical value. The network analysis and prediction steps designed in this invention aim to overcome the limitations of segment-level assessments by employing spatial aggregation and temporal extrapolation to achieve network efficiency assessment and trend prediction functions, respectively.

[0101] Preferably, the road network-level analysis employs a hierarchical aggregation strategy. First, the road segments are divided into several sub-road networks according to geographical regions or management jurisdictions, and the efficiency index of each sub-road network is calculated. Then, the efficiency indices of each sub-road network are aggregated into an overall road network efficiency index. The advantage of this hierarchical aggregation strategy is that it can provide multi-scale evaluation results simultaneously, including both the macro-level situation of the overall road network and the local conditions of each sub-road network, which facilitates differentiated management by traffic management departments.

[0102] From the perspective of prediction models, traffic flow prediction is a typical spatiotemporal series prediction problem, exhibiting significant temporal dependence and spatial correlation. Temporal dependence manifests as a correlation between the current traffic state and its historical state; for example, traffic flow during peak commuting hours shows a clear periodic pattern. Spatial correlation manifests as a transmission and influence relationship between traffic states of adjacent road segments; for example, congestion on one road segment can propagate upstream. Traditional time series prediction methods, such as ARIMA, primarily capture temporal dependence and struggle to effectively model spatial correlation. In contrast, the LSTM-attention hybrid model employed in this invention can simultaneously learn features in both spatiotemporal dimensions, resulting in significantly superior predictive capabilities compared to traditional methods.

[0103] The road network efficiency assessment subunit calculates the overall service efficiency index of the regional road network by weighting the service levels of each road segment with traffic volume. While the service level assessment of individual road segments is important, traffic management decisions often require an understanding of the overall operational status of the entire road network. This invention elevates the road segment-level assessment results to the road network level through a weighted aggregation method. In one embodiment of this invention, the road network efficiency index... The calculation formula is:

[0104] ,

[0105] in, The road network efficiency index is dimensionless, ranging from 0 to 100, and reflects the overall operational level of the entire road network. This represents the total number of road segments in the road network. For the first The comprehensive service level index of each road segment; For the first Weighting factors for each road segment.

[0106] The road segment weighting factor comprehensively considers three dimensions: road segment length, traffic flow, and road grade. In one embodiment of the present invention, the weighting factor... The calculation formula is:

[0107] ,

[0108] in, For the first The length of each road segment is expressed in km. This represents the average length of all road segments, in km. For the first Traffic volume for each road segment, expressed in pcu / h (standard vehicle equivalent). This represents the average traffic volume across all road segments, expressed in pcu / h. For the first The road grade coefficients for each road segment are as follows: 1.5 for national highways, 1.2 for provincial highways, and 1.0 for county roads. Indices of 0.3, 0.5, and 0.2 correspond to the weighting percentages of segment length, traffic flow, and road grade, respectively. The technical effect of this weighting design is that segments with high traffic volume have higher weights, reflecting their more significant impact on the overall road network efficiency; segments with higher road grades have appropriately increased weights, reflecting the importance of major corridors; and segment length has a relatively smaller impact on weighting, avoiding weight bias caused by differences in the granularity of road segment division.

[0109] The bottleneck identification subunit identifies key bottleneck sections and intersections in the road network based on congestion propagation characteristics and queue length evolution patterns. Bottlenecks are critical nodes that restrict the traffic efficiency of the road network, and identifying and eliminating bottlenecks is an effective means to improve the overall service level of the road network. In one embodiment of this invention, bottleneck identification adopts a method based on a congestion propagation model. First, road sections with service levels consistently below a threshold are identified as candidate bottlenecks. Then, the traffic state evolution of upstream and downstream road sections is analyzed to determine whether there are characteristics of congestion propagating upstream. If there is obvious congestion propagation, the road section is confirmed as a bottleneck. From the perspective of traffic flow theory, the formation mechanism of bottlenecks can be explained by shock wave theory. When the capacity of downstream road sections is lower than the traffic demand of upstream road sections, traffic flow is compressed at the bottleneck location, forming queues. These queues propagate upstream, forming backflow shock waves. This invention identifies this typical shock wave propagation characteristic by monitoring the spatiotemporal evolution pattern of the service level of each road section, thereby accurately locating the bottleneck position. Preferably, the bottleneck identification algorithm employs a sliding window detection method, setting the window length to 15 minutes. When the service level index (SPI) is below 60 for more than 70% of the time within the window, the road segment is marked as a candidate bottleneck. Then, by analyzing the SPI change trend and timing of upstream road segments, the existence of a congestion propagation relationship is verified. If the SPI decline of the upstream road segment lags behind that of the candidate bottleneck road segment, and the decline is inversely proportional to the distance, then the candidate bottleneck is confirmed as the actual bottleneck.

[0110] The bottleneck identification results include two parts: a ranking of bottleneck severity and an estimate of the impact range. Bottleneck severity is based on a comprehensive score of the following indicators: the magnitude of service level decline, congestion duration, congestion propagation range, and affected traffic volume. In one embodiment of the present invention, the bottleneck severity score... The calculation formula is:

[0111] ,

[0112] in, For road section The bottleneck severity score ranges from 0 to 1; This is a normalized value for the decline in service levels. This represents the normalized value for the duration of congestion. This represents the normalized value for the spread of congestion. To influence the normalized value of traffic volume, the impact range is calculated based on the set of affected upstream and downstream road segments using a congestion propagation model, with the propagation distance estimated using shock wave theory. The weighting design of the bottleneck severity score reflects the importance of the impact on service level and the propagation range, with the service level decline and propagation range each accounting for 30% of the weight, indicating that these two factors contribute the most to the bottleneck severity. Preferably, this invention also introduces a dynamic weight adjustment mechanism, appropriately increasing the weight of the traffic volume impact indicator during peak hours and appropriately increasing the weight of the congestion duration indicator during off-peak hours to adapt to the different focuses of traffic management at different times.

[0113] The trend prediction subunit employs a time-series deep learning model, integrating historical traffic flow patterns and weather forecast information to predict the service level changes of various road segments over the next 2 to 24 hours. Accurate trend prediction can support proactive control decisions by traffic management departments, enabling a shift from passive response to proactive management. Traditional traffic forecasting methods are mainly based on the statistical patterns of historical data, making it difficult to effectively capture the dynamic impact of weather changes on traffic flow. This invention significantly improves prediction accuracy under extreme weather conditions by introducing weather forecast information as an exogenous input variable to the prediction model. In one embodiment of this invention, the time-series deep learning model adopts an architecture combining LSTM (Long Short-Term Memory) and an attention mechanism. Input features include historical traffic flow sequences, time features (hours, days of the week, holidays, etc.), and weather forecast data. The LSTM network can effectively learn long-term dependencies in time-series data, while the attention mechanism allows the model to adaptively focus on the historical moments most relevant to the current prediction. The combination of the two can fully explore the spatiotemporal evolution patterns of traffic flow. From the perspective of deep learning model architecture design, the LSTM-attention hybrid architecture adopted in this invention includes three main components: an encoder, an attention layer, and a decoder. The encoder consists of two LSTM layers with a hidden state dimension of 128, used to extract temporal feature representations of the input sequence. The attention layer employs a scaled dot product attention mechanism to calculate the relevance weights between the current prediction time and historical time points, enabling the model to dynamically focus on the most relevant historical information. The decoder consists of one LSTM layer and a fully connected output layer, used to generate the prediction sequence. This architectural design allows the model to capture both long-term periodic patterns and adapt to short-term abrupt changes, resulting in significantly better prediction performance than traditional ARIMA and basic neural network models.

[0114] The core calculation formula of the prediction model is:

[0115] ,

[0116] ,

[0117] ,

[0118] in, For a moment The input feature vector contains traffic conditions, time features, and meteorological data; For the LSTM unit at time The hidden state vector has a dimension of 128; This is a sequence of historical hidden states; This is the attention weight vector; For the future Service level forecasts for each moment; It is a multilayer perceptron used to map the weighted hidden states to the predicted output; For the prediction time domain, the value range is 2 to 24 hours, and the prediction time step is 15 minutes.

[0119] In one embodiment of the present invention, the prediction model is trained using a historical dataset spanning two years, including traffic state samples under various weather conditions. The model employs the Adam optimizer with a learning rate of 0.001, a batch size of 64, and 100 training epochs. Model evaluation uses Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) as metrics. After sufficient training, the prediction accuracy can reach over 85%, meaning a MAPE of less than 15%.

[0120] This invention also includes a closed-loop feedback mechanism that compares the prediction results with actual observations and dynamically adjusts model parameters and data source weights based on the prediction error. In one embodiment of this invention, when the historical accuracy of a certain data source remains below a threshold, the system automatically reduces its fusion weight; when the error of the prediction model significantly increases under specific meteorological conditions, the system automatically triggers incremental model training to supplement training samples under those meteorological conditions. This closed-loop feedback mechanism ensures the system's adaptability and long-term stability.

[0121] This invention also provides a highway network operation status monitoring and service level intelligent assessment system, such as... Figure 2 As shown, the system includes a data acquisition module 1, a data fusion module 2, a meteorological sensing module 3, a service level assessment module 4, and a road network analysis module 5. Each module in the system corresponds one-to-one with each step in the method embodiment. The following describes each module.

[0122] Data acquisition module 1 is used to collect multi-source heterogeneous traffic data from video checkpoints, ETC gantry data, mobile signaling data, and internet map data, and to collect meteorological environmental data from meteorological station observation data, meteorological radar echo data, satellite remote sensing data, and road condition sensor data. In one embodiment of the present invention, data acquisition module 1 includes a traffic data acquisition submodule and a meteorological data acquisition submodule. The traffic data acquisition submodule connects to the video checkpoint system, the ETC center platform, the operator signaling platform, and the internet map API through standardized interfaces to achieve real-time aggregation of the four types of traffic data. The meteorological data acquisition submodule acquires meteorological station, meteorological radar, and satellite remote sensing data through a meteorological data sharing interface, and collects road condition data through a dedicated sensor network. The acquisition cycle and data format of data acquisition module 1 are consistent with those described in step S1 of the method embodiment.

[0123] The data fusion module 2 is used to perform spatiotemporal registration processing on the multi-source heterogeneous traffic data output by the data acquisition module 1 to generate spatiotemporal registered data, to perform sparse completion processing on road segments with insufficient monitoring facility coverage in the spatiotemporal registered data to generate sparse completed data, and to perform weighted fusion processing based on confidence weight values ​​to generate road segment traffic status data. In one embodiment of the present invention, the data fusion module 2 includes a spatiotemporal registration submodule, a sparse completion submodule, and a confidence fusion submodule. The spatiotemporal registration submodule performs time base transformation and spatial coordinate projection, the sparse completion submodule uses the weighted inference method described in formula (1) of the method embodiment to complete the data, and the confidence fusion submodule uses the dynamic weight calculation method described in formula (2) of the method embodiment to determine the fusion weight of each data source and perform weighted average fusion. The output of the data fusion module 2 is the road segment traffic status data of each road segment, including parameters such as average speed, flow rate, and occupancy rate.

[0124] The meteorological sensing module 3 is used to extract meteorological element combination feature vectors from the meteorological environment data output by the data acquisition module 1, analyze the service level attenuation law of each road segment under different meteorological conditions, and construct a meteorological sensitivity coefficient matrix. In one embodiment of the present invention, the meteorological sensing module 3 includes a feature extraction submodule and a sensitivity analysis submodule. The feature extraction submodule constructs meteorological element combination feature vectors as described in step S3 of the method embodiment for four extreme weather types: heavy rainfall, fog, road snow and ice accumulation, and crosswinds. The sensitivity analysis submodule calculates the meteorological sensitivity coefficient matrix using the method described in formula (4) of the method embodiment, and establishes a nonlinear mapping relationship between precipitation intensity and visibility and speed reduction coefficient using formulas (5) and (6). The output of the meteorological sensing module 3 is the meteorological sensitivity coefficient matrix and the speed reduction coefficient under the current meteorological conditions.

[0125] Service Level Assessment Module 4 is used to calculate the basic service level index based on the road segment traffic status data output by Data Fusion Module 2, and to calculate the meteorological correction factor based on the meteorological sensitivity coefficient matrix output by Meteorological Sensing Module 3, thereby generating a comprehensive service level index. In one embodiment of the present invention, Service Level Assessment Module 4 includes an index calculation submodule, a basic assessment submodule, and a meteorological correction submodule. The index calculation submodule calculates three core indicators: speed ratio, travel time reliability, and congestion duration percentage. The basic assessment submodule uses the weighted scoring method described in formula (3) of the method embodiment to calculate the basic service level index and classify the levels. The meteorological correction submodule uses formula (7) of the method embodiment to calculate the meteorological correction factor, and uses formula (8) to superimpose the basic service level index and the meteorological correction factor to generate a comprehensive service level index. The output of Service Level Assessment Module 4 is the comprehensive service level index and its level classification for each road segment.

[0126] The road network analysis module 5 is used to calculate the road network efficiency index based on the comprehensive service level index output by the service level assessment module 4, and to generate bottleneck identification result data and service level prediction data. In one embodiment of the present invention, the road network analysis module 5 includes an efficiency calculation submodule, a bottleneck identification submodule, and a trend prediction submodule. The efficiency calculation submodule uses the weighted aggregation method described by formula (9) in the method embodiment to calculate the road network efficiency index. The bottleneck identification submodule identifies key bottleneck road sections based on the congestion propagation model and outputs the bottleneck severity ranking and impact range estimation. The trend prediction submodule uses the LSTM-attention hybrid model described by formula (10) in the method embodiment, integrates historical traffic flow patterns and meteorological forecast information, and predicts the service level change trend of each road section in the next 2 to 24 hours. The output of the road network analysis module 5 includes the road network efficiency index, bottleneck identification result data, and service level prediction data, providing support for the proactive control decisions of traffic management departments.

[0127] The modules in this system embodiment exhibit a tight data coupling relationship and a closed-loop feedback mechanism. The output of data acquisition module 1 is simultaneously supplied to data fusion module 2 and meteorological sensing module 3; the outputs of data fusion module 2 and meteorological sensing module 3 are jointly supplied to service level assessment module 4; the output of service level assessment module 4 is supplied to road network analysis module 5; the prediction results of road network analysis module 5 inversely influence the confidence weight parameters in data fusion module 2, forming a complete closed-loop control architecture. This deeply coupled collaborative architecture ensures the organic coordination of all aspects of the system, realizing end-to-end intelligent processing from data acquisition to decision output.

[0128] The hardware deployment architecture of this invention adopts a distributed design. Data acquisition module 1 is deployed at edge computing nodes at each monitoring point, performing local data preprocessing and compressed transmission. Data fusion module 2, meteorological sensing module 3, and service level assessment module 4 are deployed in the regional data center, performing core computing tasks. Road network analysis module 5 is deployed in the provincial or national data center, performing road network-level analysis and trend prediction. This layered deployment architecture ensures both real-time data processing and efficient utilization of computing resources. Preferably, the edge computing nodes and the regional data center are connected via 5G or dedicated network, with data transmission latency below 50ms; the regional data center and the provincial data center are connected via a dedicated fiber optic network, with data transmission latency below 20ms.

[0129] The system provided by this invention supports comprehensive perception of the operational status of ordinary highway networks and scientific assessment of service efficiency, effectively improving network resilience and traffic flow assurance capabilities under extreme weather conditions. In actual deployment tests, the system demonstrated good stability and accuracy, with a service level assessment result consistency rate exceeding 92% with manual assessments, an average absolute percentage error of less than 15% in trend prediction, and a bottleneck identification accuracy rate of 88%, meeting the practical application needs of traffic management departments.

[0130] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring the operational status and intelligently assessing the service level of a highway network, characterized in that, Includes the following steps: The multi-source data acquisition steps collect video checkpoint data, ETC gantry data, mobile phone signaling data, and Internet map data to form multi-source heterogeneous traffic data, and collect meteorological station observation data, meteorological radar echo data, satellite remote sensing data, and road condition sensor data to form meteorological environment data; The data fusion processing step involves performing spatiotemporal registration processing on the multi-source heterogeneous traffic data to generate spatiotemporal registration data, performing sparse completion processing on road segments with insufficient monitoring facility coverage in the spatiotemporal registration data to generate sparse completion data, and calculating confidence weight values ​​for the spatiotemporal registration data and the sparse completion data based on the sampling coverage, timeliness, and historical accuracy of each data source, and performing weighted fusion processing to generate road segment traffic status data for each road segment. The meteorological sensitivity analysis step involves extracting meteorological element combination feature vectors from the meteorological environment data, analyzing the service level attenuation law of each road section under different meteorological conditions, and constructing a meteorological sensitivity coefficient matrix. The service level assessment steps involve calculating the speed ratio, travel time reliability, and congestion duration percentage for each road segment based on the road segment traffic status data; generating a basic service level index for each road segment based on the speed ratio, travel time reliability, and congestion duration percentage; calculating a meteorological correction factor based on the meteorological sensitivity coefficient matrix and current meteorological conditions; and superimposing the meteorological correction factor with the basic service level index to generate a comprehensive service level index. The road network analysis and prediction steps involve calculating the road network efficiency index based on the comprehensive service level index and the traffic volume weighted for each road segment, identifying key bottleneck road segments based on congestion propagation characteristics and queue length evolution patterns to generate bottleneck identification result data, and using a time-series deep learning model to integrate historical traffic flow patterns and meteorological forecast information to predict the service level change trend of each road segment and generate service level prediction data.

2. The method according to claim 1, characterized in that, In the multi-source data acquisition step: The video checkpoint data includes vehicle passing records and vehicle type information. The spacing of the video checkpoints is configured differently according to the road segment level, with the average spacing of national highways being 8km to 15km and the average spacing of provincial highways being 12km to 20km. The ETC gantry data reuses electronic tag transaction records collected by existing facilities along the route to construct an anchor point sequence for calculating interval travel time; The positioning accuracy range of the mobile phone signaling data is 100m to 500m, which is used to supplement the traffic condition estimation of road sections with sparse monitoring facilities. The internet map data includes road congestion index and estimated travel time, and is updated every 2 to 5 minutes.

3. The method according to claim 1, characterized in that, The meteorological and environmental data acquisition parameters are configured as follows: The meteorological station observation data includes ambient temperature, relative humidity, wind speed and direction, precipitation and visibility. The regular collection cycle is 1 hour, and the real-time early warning data collection cycle is 5 minutes. The spatial resolution of the meteorological radar echo data is 1 km, and the temporal resolution is 6 min. The precipitation intensity level is calculated based on the echo intensity value. The road surface condition sensors are deployed at key locations at the entrances and exits of bridges and tunnels to collect parameters such as road surface temperature, humidity, friction coefficient, and water depth.

4. The method according to claim 1, characterized in that, In the data fusion processing step: The spatiotemporal registration process includes converting the timestamps of each data source into a standard time reference, projecting the spatial location onto the road network linear reference system, and dividing road segments based on the basic unit between adjacent intersections. The confidence weight values ​​are dynamically adjusted based on three indicators: sampling coverage, timeliness, and historical accuracy. The weight coefficients for the three indicators are 0.4, 0.3, and 0.3, respectively.

5. The method according to claim 1, characterized in that, The sparse completion process employs a method that combines temporal interpolation with spatial inference. Time-dimensional interpolation uses cubic spline interpolation to estimate traffic parameters for missing time points; Spatial dimension inference is based on a weighted inference of the correlation between traffic conditions of adjacent road segments and the similarity of road segment attributes. The correlation weight is calculated based on the attenuation of distance between road segments.

6. The method according to claim 1, characterized in that, In the meteorological sensitivity analysis step, the combined feature vector of meteorological elements is constructed for four extreme weather types: heavy rainfall, dense fog, road snow and ice accumulation, and crosswinds, respectively. The feature vector of heavy rainfall includes three components: rainfall intensity, rainfall duration, and road surface water depth. The feature vector of fog type includes three components: visibility, relative humidity, and temperature-dew point difference; The feature vector of road snow and ice accumulation types includes three components: road surface temperature, ambient temperature, and precipitation type. The feature vector of crosswind type includes two components: wind speed, wind direction, and the angle between the wind speed and the road direction.

7. The method according to claim 1, characterized in that, In the service level assessment steps: The speed ratio is calculated based on the ratio of the actual average speed of the road segment to the design speed. The travel time reliability value is calculated based on the ratio of the 95th percentile to the median of the historical travel time distribution; The percentage of congestion time is the cumulative percentage of time during which the speed is 40% lower than the design speed within the statistical evaluation period. The basic service level index is divided into six levels from A to F. Level A corresponds to road segments with a speed ratio greater than 0.9 and a travel time reliability value less than 1.1, while Level F corresponds to road segments with a speed ratio less than 0.3 or a travel time reliability value greater than 2.

5.

8. The method according to claim 1, characterized in that, The calculation method for the meteorological correction factor includes: A nonlinear mapping relationship between precipitation intensity and velocity reduction coefficient was established. When the precipitation intensity is less than 10 mm / h, the velocity reduction coefficient is 0.95 to 1.0; when the precipitation intensity is 10 mm / h to 25 mm / h, the velocity reduction coefficient is 0.80 to 0.95; and when the precipitation intensity is greater than 25 mm / h, the velocity reduction coefficient is 0.50 to 0.

80. A nonlinear mapping relationship between visibility and speed reduction factor was established. When visibility is greater than 1000m, the speed reduction factor is 0.95 to 1.0; when visibility is 200m to 1000m, the speed reduction factor is 0.70 to 0.95; and when visibility is less than 200m, the speed reduction factor is 0.30 to 0.

70. The meteorological correction factor is the smaller value of the precipitation velocity reduction factor and the visibility velocity reduction factor.

9. The method according to claim 1, characterized in that, In the road network analysis and prediction steps: In the calculation of the road network efficiency index, the weighting factors comprehensively consider road segment length, traffic flow and road grade, with road segment length accounting for 0.3, traffic flow accounting for 0.5 and road grade accounting for 0.

2. The bottleneck identification results include a ranking of bottleneck severity and an estimated impact range. The impact range is calculated based on a congestion propagation model, which sets the upstream and downstream affected road segments. The prediction time domain of the time-series deep learning model is from 2h to 24h, and the prediction time step is 15min.

10. A highway network operation status monitoring and service level intelligent assessment system, used to implement the method described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect video checkpoint data, ETC gantry data, mobile phone signaling data and Internet map data to form multi-source heterogeneous traffic data, and to collect meteorological station observation data, meteorological radar echo data, satellite remote sensing data and road condition sensor data to form meteorological environment data. The data fusion module is used to perform spatiotemporal registration processing on the multi-source heterogeneous traffic data output by the data acquisition module to generate spatiotemporal registration data, perform sparse completion processing on road segments with insufficient monitoring facility coverage in the spatiotemporal registration data to generate sparse completion data, and perform weighted fusion processing based on confidence weight values ​​to generate road segment traffic status data. The meteorological sensing module is used to extract meteorological element combination feature vectors from the meteorological environment data output by the data acquisition module, analyze the service level attenuation law of each road section under different meteorological conditions, and construct a meteorological sensitivity coefficient matrix. The service level assessment module is used to calculate the basic service level index based on the road segment traffic status data output by the data fusion module, calculate the meteorological correction factor based on the meteorological sensitivity coefficient matrix output by the meteorological sensing module, and generate the comprehensive service level index. The road network analysis module is used to calculate the road network efficiency index based on the comprehensive service level index output by the service level assessment module, and to generate bottleneck identification result data and service level prediction data.

Images