A highway traffic anomaly detection and emergency method
By combining data collected from vehicle-mounted terminals and roadside equipment, and using spatiotemporal clustering and digital twin models to analyze highway traffic anomalies, the problems of inaccurate identification and inflexible emergency decision-making in existing technologies have been solved. This has enabled accurate identification of abnormal vehicles and optimization of emergency strategies, thereby improving the efficiency and safety of traffic safety management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YUNNAN COMM INVESTMENT & CONSTR GRP CO LTD
- Filing Date
- 2026-04-01
- Publication Date
- 2026-06-16
AI Technical Summary
Existing technologies are not precise enough in detecting traffic anomalies and responding to emergencies on highways. They also have limited timeliness and accuracy in predicting accidents, and lack dynamism and flexibility in emergency decision-making, making it difficult to adjust and optimize them in a timely manner according to complex traffic conditions.
By combining vehicle-mounted terminals and roadside equipment to collect real-time vehicle data, and combining this with drivers' historical behavior to predict driving trajectories, the behavior patterns of abnormal vehicles are analyzed using spatiotemporal clustering algorithms to assess accident risk levels. Furthermore, digital twin models are used to simulate traffic conditions, formulate multiple emergency strategies, and select the optimal emergency measures.
It enables accurate real-time identification of abnormal vehicles and prediction of potential accident types, improves the timeliness of traffic anomaly detection and emergency response accuracy, optimizes the emergency decision-making process, ensures the flexibility and efficiency of traffic safety management, and minimizes the impact of traffic accidents.
Smart Images
Figure CN122223965A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of road traffic safety detection, and in particular to a method for detecting and responding to traffic anomalies on highways. Background Technology
[0002] With the continuous development of my country's economy and the acceleration of urbanization, the construction of expressways has entered a stage of rapid expansion, gradually forming an important transportation network with wide coverage. At the same time, with the continuous growth of urban population and the number of vehicles, the operation and management of expressways have become significantly more difficult, traffic congestion and accidents have become increasingly prominent, and the safety hazards of expressways have become more and more serious. Traffic accidents not only put enormous pressure on road traffic safety, but also cause serious property damage and casualties.
[0003] Therefore, timely detection and early warning of abnormal vehicle behavior is of great significance. By using advanced technologies to improve the intelligence level of highways, efficient collection and accurate processing of road information can be achieved, thereby quickly identifying abnormal driving behavior and taking timely countermeasures. To a certain extent, this helps roads to resume normal traffic as soon as possible and reduces the occurrence of traffic accidents.
[0004] Existing technologies for traffic anomaly detection and emergency response mainly rely on a single data source or traditional monitoring methods, resulting in insufficient accuracy in identifying abnormal vehicles and limited timeliness and accuracy in accident prediction. In addition, existing emergency decision-making often lacks dynamism and flexibility, making it difficult to adjust and optimize in a timely manner according to complex traffic conditions. Summary of the Invention
[0005] To address the problem that existing emergency decision-making lacks dynamism and flexibility, and is difficult to adjust and optimize in a timely manner according to complex traffic conditions, this application provides a method for detecting and responding to traffic anomalies on highways.
[0006] A method for detecting and responding to traffic anomalies on highways, comprising the following steps: S1. Collect real-time vehicle data using pre-configured vehicle terminals and roadside equipment, and predict driving trajectories by combining driver history; compare the actual path with the predicted driving trajectory to identify abnormal vehicles whose deviation exceeds the set deviation threshold. S2. Analyze the behavior patterns of abnormal vehicles, combine spatiotemporal clustering algorithms to analyze the spatial distribution and temporal characteristics of accident types, infer the types of accidents that occurred, assess the risk level of accidents, and infer the points where anomalies occurred. S3. Based on the accident risk level, lightweight tasks are issued using edge nodes, and the speed, trajectory deviation and driving behavior of abnormal vehicles are verified according to emergency response standards. Based on the verification results, various emergency strategies are formulated. S4. Use a digital twin model to simulate the traffic conditions in the target area, evaluate the effectiveness of various emergency strategies, and then select the optimal emergency strategy to issue to the abnormal vehicle and other vehicles around it.
[0007] Optionally, real-time vehicle data is collected using pre-configured on-board terminals and roadside equipment, and the driving trajectory is predicted in conjunction with the driver's historical behavior. The process of comparing the actual path with the predicted trajectory and identifying abnormal vehicles whose deviation exceeds a set deviation threshold includes the following steps: S11. Real-time vehicle driving data, including speed, position, acceleration, and steering angle, of each vehicle in the target area is obtained through pre-configured vehicle terminals and roadside equipment. S12. Retrieve historical behavioral data from the driver, including common driving habits, historical trajectories, and records of abnormal behaviors; S13. Based on vehicle driving data and historical behavior data, use a trajectory prediction model to generate the driving trajectory of each vehicle within a preset time period in the future. S14. Compare the actual driving path of the vehicle with the driving trajectory within a preset time period in the future, and calculate the deviation between the two. S15. Compare the deviation with a preset deviation threshold, identify vehicles whose deviation exceeds the preset deviation threshold, and mark them as abnormal vehicles.
[0008] Optionally, analyzing the behavioral patterns of abnormal vehicles, combining spatiotemporal clustering algorithms to analyze the spatial distribution and temporal characteristics of accident types, inferring the accident type, assessing the accident risk level, and inferring the point of anomaly occurrence includes the following steps: S21. The predicted driving trajectory of the abnormal vehicle is spatiotemporally segmented, and the trajectory is divided into multiple trajectory segments based on the time window. The spatial features, temporal features and kinematic features of the predicted trajectory segments are extracted. S22. Use a spatiotemporal clustering algorithm to cluster the segmented trajectory segments, perform similarity analysis based on the spatial, temporal and kinematic features of the predicted trajectory segments, and mark abnormal trajectory segments during the clustering process. S23. Based on the abnormal trajectory segments, infer the type of accident that occurred and conduct a risk level assessment of the abnormal vehicles; S24. Based on the risk level assessment results, mark the location of the abnormal vehicle and infer the point of occurrence of the anomaly.
[0009] Optionally, the segmented trajectory segments are clustered using a spatiotemporal clustering algorithm. Similarity analysis is performed based on the spatial, temporal, and kinematic features of the predicted trajectory segments, and abnormal trajectory segments are marked during the clustering process, including the following steps: S221. Standardize the spatial, temporal, and kinematic features of the acquired predicted trajectory segment; S222. Set the weights of the standardized spatial features, temporal features, and kinematic features, and define the spatiotemporal distance metric between predicted trajectory segments; S223. Set the spatial neighborhood threshold based on typical vehicle speed and road speed limit, set the spatial neighborhood threshold based on specific time windows, and set the minimum number of points by observing the expected number of trajectory segments in the same lane. S224. Based on the spatiotemporal distance metric, spatial neighborhood threshold, temporal neighborhood threshold, and minimum number of points, perform spatiotemporal clustering on all trajectory segments to obtain the cluster label and noise label for each trajectory segment. S225. Identify the trajectory segment marked as noise as an abnormal trajectory segment.
[0010] Optionally, the expression for the spatiotemporal distance metric between predicted trajectory segments is defined as follows: ; In the formula, Represents trajectory points With trajectory points The spatiotemporal distance between them; Weights representing spatial features; Weights representing time features; Weights representing kinematic characteristics; Represents trajectory points With trajectory points Spatial characteristics and The Euclidean distance between them; Represents trajectory points With trajectory points In terms of time characteristics and The time difference between them; Represents trajectory points With trajectory points Kinematic characteristics between and The kinematic differences between them.
[0011] Optionally, based on spatiotemporal distance metrics, spatial neighborhood thresholds, temporal neighborhood thresholds, and minimum number of points, spatiotemporal clustering is performed on all trajectory segments to obtain cluster labels and noise labels for each trajectory segment, including the following steps: S2231. Use spatiotemporal distance metrics to obtain all neighboring trajectory segments in the spatial and temporal neighborhoods of each trajectory segment; S2232. Check whether the neighborhood of the current trajectory segment contains at least a minimum number of trajectory segments. If it contains at least a minimum number of trajectory segments, determine the current trajectory segment as the core point and create a new cluster based on the core point. S2233. Starting from the core point, add all trajectory segments within the neighborhood of the core point. If the trajectory segments within the neighborhood meet the condition of minimum number of points, continue to expand the current cluster. S2234. If the current trajectory segment cannot be classified into any cluster or is not the core point of any cluster and does not have enough neighborhood points, then the current trajectory segment is marked as noise. S2235. Assign each trajectory segment to a cluster, or label it as noise, and generate a cluster label and a noise label for each trajectory segment.
[0012] Optionally, inferring the type of accident based on abnormal trajectory segments and assessing the risk level of abnormal vehicles includes the following steps: S231. Based on abnormal trajectory segments, analyze the kinematic characteristics of sudden speed changes, rapid acceleration or deceleration, and sharp turns to make a preliminary inference about the current accident type. S232. Based on the spatial characteristics of abnormal trajectory segments, determine whether they occur in specific road sections and whether there is persistent lane departure, and refine the prediction of accident types. S233. Calculate the trajectory deviation of the abnormal trajectory segment according to the accident type; assess the speed change characteristics of the abnormal trajectory segment, and count the number of vehicles involved in the abnormal trajectory segment. S234. Weighted fusion of trajectory deviation, speed change, and number of vehicles to generate a comprehensive risk score; S235. Compare the comprehensive risk score with the preset risk level classification standard, and classify the abnormal vehicles into low-risk, medium-risk and high-risk categories based on the comparison results.
[0013] Optionally, based on the accident risk level, lightweight tasks are issued using edge nodes, and the speed, trajectory deviation, and driving behavior of abnormal vehicles are verified according to emergency response standards. Based on the verification results, various emergency strategies are formulated, including the following steps: S31. Based on the accident risk level, distribute the lightweight task to the edge node; S32. At edge nodes, verify the speed, trajectory deviation, and driving behavior of abnormal vehicles based on emergency response standards; S33. Based on the verification results, assess the dangers of abnormal vehicle speed, trajectory deviation, and driving behavior, and formulate corresponding emergency strategies based on the assessment results.
[0014] Optionally, the traffic conditions in the target area are simulated using a digital twin model to evaluate the effectiveness of various emergency strategies, and then the optimal emergency strategy is selected and issued to the abnormal vehicle and other vehicles in its vicinity, including the following steps: S41. Obtain static environmental information and dynamic vehicle information of the target area; S42. Construct a static twin model of the scene based on static environment information, and then construct a dynamic twin model of the vehicle based on dynamic vehicle information; S43. Construct a digital twin model of the target area based on the static twin model of the scene and the dynamic twin model of the vehicle, and simulate the traffic conditions of the target area; S44. Input multiple emergency strategies into the digital twin model of the target area and perform simulation calculations to predict the impact and evolution of emergency strategies on traffic flow. S45. Evaluate the simulation results of multiple emergency strategies, select the optimal emergency strategy based on the evaluation results, and distribute the optimal emergency strategy to the abnormal vehicle and other relevant vehicles around it.
[0015] In summary, this application includes at least one of the following beneficial technical effects: By combining vehicle-mounted terminals, roadside equipment, driver historical behavior, and spatiotemporal clustering analysis, this application can accurately identify abnormal vehicles and predict potential accident types in real time; through multi-dimensional data collection and analysis, it improves the accuracy of timely detection and emergency response to traffic anomalies; at the same time, by utilizing digital twin model simulation and evaluation of multiple emergency strategies, it effectively optimizes the emergency decision-making process, ensuring the flexibility and efficiency of traffic safety management; and through comprehensive risk assessment and precise strategy issuance, it minimizes the impact of traffic accidents and improves the overall safety and emergency response capabilities of highways. Attached Figure Description
[0016] Figure 1 This is a flowchart of the method in this application. Detailed Implementation
[0017] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.
[0018] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0019] This application discloses a method for detecting and responding to traffic anomalies on highways, referring to... Figure 1 The highway traffic anomaly detection and emergency response methods include the following steps: S1. Collect real-time vehicle data using pre-configured vehicle terminals and roadside equipment, and predict driving trajectories by combining driver history; compare the actual path with the predicted driving trajectory to identify abnormal vehicles whose deviation exceeds the set deviation threshold.
[0020] Preferably, the process of collecting real-time vehicle data using pre-configured onboard terminals and roadside equipment, and predicting driving trajectories by combining this data with the driver's historical behavior, and then comparing the actual path with the predicted trajectory to identify abnormal vehicles whose deviation exceeds a set deviation threshold, includes the following steps: S11. Real-time vehicle driving data, including speed, position, acceleration, and steering angle, of each vehicle in the target area is obtained through pre-configured vehicle terminals and roadside equipment.
[0021] S12. Retrieve historical behavioral data of the driver, including common driving habits, historical trajectories, and records of abnormal behaviors.
[0022] S13. Based on vehicle driving data and historical behavior data, use a trajectory prediction model to generate the driving trajectory of each vehicle within a preset time period in the future.
[0023] Based on the vehicle driving data and the historical behavior data, generating the driving trajectory of each vehicle within a preset future time using a trajectory prediction model includes the following steps: Extract speed, position, acceleration, and steering angle from vehicle driving data, and common driving habits, historical trajectories, and abnormal behavior records from historical behavior data; Speed, position, acceleration, steering angle, driving habits, historical trajectory, and abnormal behavior records are assigned corresponding weights, and vehicle state information is synthesized according to these weights. Input the state information of the synthetic vehicle into the trajectory prediction model; The trajectory prediction model is used to generate the vehicle's driving trajectory within a preset time period in the future, and the predicted value of the future path is output.
[0024] It needs to be explained that, firstly, the vehicle-mounted terminal (such as GPS, inertial measurement unit) and roadside equipment (such as camera, radar) collect dynamic data of vehicles in the target area in real time, including key parameters such as speed, position, acceleration and steering angle; at the same time, the driver's historical behavior database is called up to integrate their driving habits (such as following distance, lane change frequency), historical trajectory (such as usual driving route) and abnormal behavior records (such as sudden braking, frequent lane changes, etc.).
[0025] By integrating and weighting data, real-time driving data and historical behavior data are combined. Specifically, dynamic parameters such as speed and position are assigned real-time weights, while long-term behavior data such as driving habits and historical trajectories are assigned stability weights. Through weighted synthesis, a "state information" that comprehensively reflects the vehicle's current state and potential behavioral tendencies is generated (for example, an aggressive driver may be assigned a higher steering angle weight in lane change prediction). This state information is input into a trajectory prediction model (such as the Transformer in existing technologies or a hybrid model based on physical rules). By analyzing time series dependencies and driving behavior patterns, the model outputs a predicted value of the vehicle's driving trajectory within a preset future time. This prediction not only considers the vehicle's physical kinematics but also incorporates the driver's personalized characteristics, thereby improving the accuracy of trajectory prediction in complex traffic scenarios (such as congestion and sudden obstacles).
[0026] S14. Compare the actual driving path of the vehicle with the driving trajectory within a preset time period in the future, and calculate the deviation between the two.
[0027] S15. Compare the deviation with a preset deviation threshold, identify vehicles whose deviation exceeds the preset deviation threshold, and mark them as abnormal vehicles.
[0028] Specific examples are illustrated below: During the morning rush hour, a city has deployed intelligent monitoring terminals based on vehicle-mounted terminals and roadside equipment on its expressways; these terminals are required to detect abnormal driving behaviors of vehicles caused by fatigue driving or sudden illness in real time (such as sudden braking or serpentine driving).
[0029] The specific implementation steps are as follows: (1) Data acquisition stage Vehicle terminal: Vehicle A's GPS reports its real-time location (longitude 116.404, latitude 39.915) at a frequency of 10Hz. The inertial measurement unit detects an abnormal lateral acceleration of 0.5g (normal value <0.3g). Roadside equipment: Camera No. 3 uses computer vision to identify that vehicle A's turn signal is not activated, but the vehicle body is clearly deviating from the center line of the lane; Historical data: This driver has two records of sudden braking in the past 3 months, and his daily commute route is fixed on the expressway from east to west.
[0030] (2) Trajectory prediction stage Data weighting: 0.7 weight is assigned to real-time lateral acceleration (high real-time performance), and 0.3 weight is assigned to historical emergency braking records (low stability). Model prediction: Based on the current state (abnormal lateral acceleration + no steering signal) and historical behavior, the Transformer model predicts that the vehicle should maintain a straight line in the next 10 seconds with a confidence level of 85%.
[0031] (3) Deviation detection stage Actual path: 2 seconds later, the roadside radar detected that vehicle A suddenly veered 1.5 meters to the right (exceeding 50% of the lane width); Threshold comparison: The preset lateral deviation threshold is 0.8 meters (set according to the lane width of 3.5 meters), and the actual deviation triggers an alarm.
[0032] (4) Anomaly determination: Mark vehicle A as a "high-risk abnormal vehicle".
[0033] S2. Analyze the behavior patterns of abnormal vehicles, combine spatiotemporal clustering algorithms to analyze the spatial distribution and temporal characteristics of accident types, infer the types of accidents that occurred, assess the risk level of accidents, and infer the location of anomalies.
[0034] Preferably, analyzing the behavioral patterns of abnormal vehicles, combining spatiotemporal clustering algorithms to analyze the spatial distribution and temporal characteristics of accident types, inferring the accident type, assessing the accident risk level, and inferring the point of anomaly occurrence includes the following steps: S21. The predicted driving trajectory of the abnormal vehicle is spatiotemporally segmented. Based on the time window, the trajectory is divided into multiple trajectory segments, and the spatial features, temporal features and kinematic features of the predicted trajectory segments are extracted.
[0035] It needs to be explained that the predicted trajectory needs to be divided into time windows, so that each trajectory segment represents the vehicle's driving state within a certain time range. Each trajectory segment includes spatial features (such as position, lane departure, steering angle, etc.), temporal features (such as rate of change of speed, travel time, dwell time, etc.), and kinematic features (such as acceleration, angular velocity, braking state, etc.). Among them, spatial features can reveal the spatial distribution of the vehicle within a specific time period and the degree of deviation from the normal driving route, while temporal features reflect the temporal changes and trends during the vehicle's driving process, and kinematic features describe the vehicle's dynamic characteristics, such as whether there are abnormal sudden braking, excessive acceleration, or other behaviors. For example, if the vehicle's steering angle changes sharply and its speed drops abruptly within a certain trajectory segment, it may indicate that it has engaged in abnormal driving behavior. Through spatiotemporal segmentation and feature extraction, we can better understand and respond to complex traffic situations.
[0036] S22. Use a spatiotemporal clustering algorithm to cluster the segmented trajectory segments, perform similarity analysis based on the spatial, temporal and kinematic features of the predicted trajectory segments, and mark abnormal trajectory segments during the clustering process.
[0037] Preferably, the process of clustering the segmented trajectory segments using a spatiotemporal clustering algorithm, performing similarity analysis based on the spatial, temporal, and kinematic features of the predicted trajectory segments, and marking abnormal trajectory segments during the clustering process includes the following steps: S221. Standardize the spatial, temporal, and kinematic features of the obtained predicted trajectory segments.
[0038] S222. Set the weights of the standardized spatial features, temporal features, and kinematic features, and define the spatiotemporal distance metric between predicted trajectory segments.
[0039] Preferably, the expression for the spatiotemporal distance metric between predicted trajectory segments is defined as follows: ; In the formula, Represents trajectory points With trajectory points The spatiotemporal distance between them; Weights representing spatial features; Weights representing time features; Weights representing kinematic characteristics; Represents trajectory points With trajectory points Spatial characteristics and The Euclidean distance between them; Represents trajectory points With trajectory points In terms of time characteristics and The time difference between them; Represents trajectory points With trajectory points Kinematic characteristics between and The kinematic differences between them.
[0040] It is important to explain that, firstly, different features are standardized to eliminate the impact of differences in dimensions, ensuring that spatial location, temporal distribution, and kinematic parameters are comparable on a uniform scale. Then, weights are assigned to various features based on the importance of the actual scenario; for example, the weight of spatial features can be increased in lane-keeping scenarios, while the weight of temporal features can be appropriately increased during congested or special traffic periods. Next, a comprehensive spatiotemporal distance metric is defined to integrate spatial offsets (such as the Euclidean distance of lane deviation), temporal differences (such as the difference in rate of change of speed or dwell time), and kinematic differences (such as changes in acceleration or angular velocity) between trajectory segments into a single distance function, thereby characterizing the overall similarity between different trajectory segments. During clustering, trajectory segments with high similarity are grouped into the same category, while trajectory segments that significantly deviate from the mainstream pattern are identified and marked as anomalies, dynamically reflecting the behavioral patterns of vehicles in the spatiotemporal dimensions and effectively highlighting potentially risky trajectories.
[0041] S223. Set the spatial neighborhood threshold based on typical vehicle speed and road speed limit, set the spatial neighborhood threshold based on specific time windows, and set the minimum number of points by observing the expected number of trajectory segments in the same lane.
[0042] Preferably, based on spatiotemporal distance metrics, spatial neighborhood thresholds, temporal neighborhood thresholds, and minimum number of points, spatiotemporal clustering is performed on all trajectory segments to obtain cluster labels and noise labels for each trajectory segment, including the following steps: S2231. Use spatiotemporal distance metrics to obtain all neighboring trajectory segments in the spatial and temporal neighborhoods of each trajectory segment; S2232. Check whether the neighborhood of the current trajectory segment contains at least a minimum number of trajectory segments. If it contains at least a minimum number of trajectory segments, determine the current trajectory segment as the core point and create a new cluster based on the core point. S2233. Starting from the core point, add all trajectory segments within the neighborhood of the core point. If the trajectory segments within the neighborhood meet the condition of minimum number of points, continue to expand the current cluster. S2234. If the current trajectory segment cannot be classified into any cluster or is not the core point of any cluster and does not have enough neighborhood points, then the current trajectory segment is marked as noise. S2235. Assign each trajectory segment to a cluster, or label it as noise, and generate a cluster label and a noise label for each trajectory segment.
[0043] It needs to be explained that the process utilizes spatiotemporal distance metrics to calculate all neighboring trajectory segments within the spatial and temporal neighborhoods of each trajectory segment, ensuring accurate capture of the relationships between them. Next, it checks whether each trajectory segment's neighborhood contains at least a preset minimum number of points. If this condition is met, the trajectory segment is considered a core point, and a new cluster can be formed based on this core point. Then, the cluster is expanded by adding all trajectory segments within the core point's neighborhood. If these trajectory segments meet the minimum number of points condition, they are further expanded and added to the current cluster. Trajectory segments that do not meet the core point condition or have insufficient neighboring trajectory segments are classified as noise, indicating low similarity to other trajectory segments and not belonging to any cluster. Finally, by classifying each trajectory segment, cluster labels and noise labels are generated, thereby achieving trajectory segment clustering and anomaly trajectory segment identification. This process can discover group patterns of trajectory segments in the spatiotemporal dimension and identify abnormal trajectories that do not conform to mainstream behavior patterns, ultimately providing important evidence for the monitoring and early warning of abnormal events.
[0044] S224. Based on the spatiotemporal distance metric, spatial neighborhood threshold, temporal neighborhood threshold, and minimum number of points, perform spatiotemporal clustering on all trajectory segments to obtain the cluster label and noise label for each trajectory segment.
[0045] S225. Identify the trajectory segment marked as noise as an abnormal trajectory segment.
[0046] It should be explained that, based on spatiotemporal distance metrics and combined with thresholds for spatial and temporal neighborhoods, the clustering algorithm can determine which trajectory segments are similar in both spatial and temporal dimensions, and assign these trajectory segments to different clusters based on these similarities. Simultaneously, the set minimum number of points helps ensure that each cluster consists of a sufficient number of trajectory segments, avoiding isolated clusters caused by a single anomalous behavior. After clustering, each trajectory segment is assigned a cluster label indicating which cluster it belongs to. Furthermore, if a trajectory segment cannot be classified into any cluster, or if the number of trajectory segments in its neighborhood is insufficient and does not meet the core point condition, it will be marked as a noise label.
[0047] S23. Based on the abnormal trajectory segment, infer the type of accident that occurred and conduct a risk level assessment of the abnormal vehicle.
[0048] Preferably, inferring the type of accident based on abnormal trajectory segments and assessing the risk level of abnormal vehicles includes the following steps: S231. Based on abnormal trajectory segments, analyze the kinematic characteristics of sudden speed changes, rapid acceleration or deceleration, and sharp turns to make a preliminary inference about the current accident type. S232. Based on the spatial characteristics of abnormal trajectory segments, determine whether they occur in specific road sections and whether there is persistent lane departure, and refine the prediction of accident types. S233. Calculate the trajectory deviation of the abnormal trajectory segment according to the accident type; assess the speed change characteristics of the abnormal trajectory segment, and count the number of vehicles involved in the abnormal trajectory segment. S234. Weighted fusion of trajectory deviation, speed change, and number of vehicles to generate a comprehensive risk score; S235. Compare the comprehensive risk score with the preset risk level classification standard, and classify the abnormal vehicles into low-risk, medium-risk and high-risk categories based on the comparison results.
[0049] S24. Based on the risk level assessment results, mark the location of the abnormal vehicle and infer the point of occurrence of the anomaly.
[0050] It should be explained that sudden speed change refers to a rapid change in vehicle speed, which usually means sudden stopping, hard braking, or an accident; rapid acceleration reflects a sudden acceleration of the vehicle, which may indicate that the driver is trying to react quickly or avoid an obstacle; rapid deceleration reflects a rapid deceleration of the vehicle, which is usually an emergency braking response and may indicate that a collision is imminent; and sharp turns and rapid changes of direction may mean that the vehicle is out of control or a collision has occurred.
[0051] Spatial features mainly include: (1) Whether it occurs on a specific road section: such as highways, curves, intersections, etc. Since these road sections usually have a higher risk of accidents, if the abnormal trajectory occurs in these places, the possibility of the accident type is higher.
[0052] (2) Persistent lane departure: Whether the vehicle continuously deviates from the lane usually reflects the driver's control problem or error, which may be caused by factors such as fatigue driving, loss of control, or drowsiness.
[0053] In addition, trajectory deviation refers to the degree of deviation of a vehicle in an abnormal trajectory segment, that is, the difference between the vehicle's driving trajectory and the normal driving trajectory; trajectories with large deviations are usually associated with accidents, indicating that the vehicle did not follow the predetermined route and may have experienced a collision, loss of control or other abnormal behavior.
[0054] Speed change refers to assessing the magnitude of speed changes in abnormal trajectory segments; sudden speed changes, such as rapid acceleration or deceleration, are usually accompanied by accidents; these changes reflect the driver's rapid response to the surrounding environment or emergency braking.
[0055] The number of vehicles involved refers to the number of vehicles that may be involved in an abnormal trajectory segment in complex traffic environments such as multi-lane roads or intersections. Situations involving a large number of vehicles are usually related to more serious accidents or traffic congestion and require special attention.
[0056] In the weighted fusion formula, the comprehensive risk score = w1 × trajectory deviation + w2 × speed change + w3 × number of vehicles, where w1, w2, and w3 are the weights of trajectory deviation, speed change, and number of vehicles, respectively. The weight values reflect the importance of different features to accident risk. For example, sharp turns and lane departures have a greater impact on accident types and may be given higher weights.
[0057] The comprehensive risk score is obtained by weighted fusion of different features (such as trajectory deviation, speed change, number of vehicles, etc.); assuming the score range is between 0 and 100, for simplification, the comprehensive risk score is divided into three levels: low risk, medium risk and high risk.
[0058] The following is an example of risk level classification criteria: Based on specific business needs and historical data, the following risk score thresholds are set: Among them, the overall risk score for low-level risks is 0-30; Note: A lower score in this category indicates that the vehicle's abnormal behavior is minor and the likelihood of an accident is low; such vehicles may have slight deviations or mild speed fluctuations, but overall do not pose a serious safety threat.
[0059] Overall risk score for intermediate risk: 31-60; Note: This score indicates a moderate risk level for the abnormal trajectory segment, posing a certain safety hazard. The vehicle may have experienced a sharp turn, significant speed fluctuations, or slight lane departure. Although the probability of an accident is low, further monitoring and attention are required.
[0060] Overall risk score for advanced risk: 61-100; Note: A higher score in this category indicates a greater risk of abnormal trajectory segments; the vehicle may have experienced rapid acceleration, sudden braking, significant lane departure, or sudden speed changes, indicating a higher probability of potential accidents and that it may be approaching the critical point of an accident.
[0061] The specific scenarios are as follows: A city's traffic management platform detected an abnormality in a truck during the evening rush hour. It needs to determine whether this indicates a potential accident risk and issue a warning. The vehicle's 30-minute trajectory is divided into several small segments, each segment containing the vehicle's position, time, and motion status; At an abnormal moment, it was found that the vehicle suddenly decelerated (from 60 km / h to 15 km / h) on the elevated curve, accompanied by obvious lateral deviation and a sudden lane change. Grouping together normal trajectory segments forms a "normal pattern"; Abnormal segments are identified as "isolated points" because they differ too much from other segments in terms of location, time, and motion. These isolated points precisely correspond to the time periods when the vehicle braked suddenly and yawed. Based on the vehicle's motion characteristics (sudden braking + lateral sway) and the road environment (curves, high speed), it is speculated that the anomaly is likely due to "sideslip or rollover risk". A comprehensive score is given based on factors such as the degree of deviation, the magnitude of speed change, and the number of surrounding vehicles affected. The results showed a high risk score, exceeding the threshold for "high risk"; The truck was at high risk of overturning at around 17:45 on a curve of an elevated road. The platform automatically sends warnings to traffic police and reminds vehicles behind to slow down through road information screens, while also dispatching emergency rescue resources.
[0062] S3. Based on the accident risk level, lightweight tasks are issued using edge nodes, and the speed, trajectory deviation and driving behavior of abnormal vehicles are verified according to emergency response standards. Based on the verification results, various emergency strategies are formulated.
[0063] Preferably, based on the accident risk level, lightweight tasks are issued using edge nodes, and the speed, trajectory deviation, and driving behavior of abnormal vehicles are verified according to emergency response standards. Based on the verification results, various emergency strategies are formulated, including the following steps: S31. Based on the accident risk level, distribute the lightweight task to the edge node; S32. At edge nodes, verify the speed, trajectory deviation, and driving behavior of abnormal vehicles based on emergency response standards; S33. Based on the verification results, assess the dangers of abnormal vehicle speed, trajectory deviation, and driving behavior, and formulate corresponding emergency strategies based on the assessment results.
[0064] It should be explained that, based on the risk level of abnormal vehicles, lightweight emergency tasks are distributed to edge nodes close to the data source. These tasks typically include vehicle monitoring, behavior analysis, and emergency response instructions. Next, the edge nodes verify in real time whether the vehicle's speed is abnormal, whether there is obvious trajectory deviation, and whether there are dangerous signs of driving behavior, such as sudden braking or lane departure, based on preset emergency response standards. (Lightweight tasks refer to those tasks that have low computational requirements, high processing speed requirements, and can be executed efficiently in resource-constrained environments.)
[0065] S4. Use a digital twin model to simulate the traffic conditions in the target area, evaluate the effectiveness of various emergency strategies, and then select the optimal emergency strategy to issue to the abnormal vehicle and other vehicles around it.
[0066] Preferably, the process of simulating traffic conditions in the target area using a digital twin model, evaluating the effectiveness of various emergency strategies, and then selecting the optimal emergency strategy to issue to the abnormal vehicle and other vehicles in its vicinity includes the following steps: S41. Obtain static environmental information and dynamic vehicle information of the target area; S42. Construct a static twin model of the scene based on static environment information, and then construct a dynamic twin model of the vehicle based on dynamic vehicle information; S43. Construct a digital twin model of the target area based on the static twin model of the scene and the dynamic twin model of the vehicle, and simulate the traffic conditions of the target area; S44. Input multiple emergency strategies into the digital twin model of the target area and perform simulation calculations to predict the impact and evolution of emergency strategies on traffic flow. S45. Evaluate the simulation results of various emergency strategies, select the optimal emergency strategy based on the evaluation results, and distribute the optimal emergency strategy to the abnormal vehicle and other relevant vehicles in its vicinity. A specific example is illustrated below: A truck suddenly brakes while traveling at high speed and deviates significantly from its lane, causing it to leave the normal driving lane; the vehicle does not immediately resume normal driving and remains speeding for a period of time; the anomaly is detected by cameras and sensors, quickly identifying a potential accident risk for the vehicle.
[0067] After analyzing the behavior of abnormal vehicles, the monitoring terminal assesses their risk level; based on factors such as sudden braking, speeding, and lane departure, the risk level of the vehicle is determined to be "high".
[0068] Lightweight tasks (such as monitoring the vehicle, analyzing its speed, trajectory deviation, and driving behavior) are distributed to edge nodes. The edge nodes are responsible for executing these tasks, including acquiring the vehicle's location, speed, and driving status in real time and transmitting this data back to the center for analysis. The edge nodes monitor and perform simple analysis of the vehicle's behavior to provide rapid feedback from the edge nodes and reduce data transmission latency.
[0069] Edge nodes analyze whether the detected vehicle speed exceeds the speed limit of the road; for example, the speed of a truck is detected as exceeding the speed limit by 30%; the edge nodes further check whether the vehicle trajectory deviates from the prescribed lane; the vehicle's trajectory is significantly deviated from the center of the lane in the monitoring data and is close to the edge line of the lane; check for sudden braking and other possible dangerous driving behaviors, such as lane departure, sudden acceleration, etc., and the results show that the vehicle has engaged in sudden braking and trajectory deviation behaviors.
[0070] Based on the data provided by the edge nodes, the center assesses that the vehicle poses a high risk and may cause traffic accidents or threaten other vehicles; it sends a deceleration command to the vehicle to force it to slow down, ensuring that the vehicle can resume normal driving and maintain a safe distance from other vehicles; and it sends a warning to other vehicles via the vehicle's onboard system, informing them that there is an abnormally driven vehicle ahead and that they need to remain vigilant.
[0071] Based on the location and behavior of the abnormal vehicle, traffic lights may be activated to adjust traffic flow in the surrounding area in advance and prevent other vehicles from approaching the abnormal vehicle.
[0072] Acquire static environmental information (such as traffic signs, road conditions, lane distribution, etc.) and dynamic vehicle information (such as traffic flow, real-time vehicle location, etc.) for the area; based on the collected information, construct a static twin model and a dynamic digital twin model for the target area; the static twin model includes the road network, while the dynamic model includes real-time vehicle location and speed information; simulate the traffic conditions in the area and the effects of various emergency strategies using the digital twin model; for example, simulate what kind of traffic congestion and safety risks would be caused if the vehicle continued to travel at high speed, and what potential risks would be reduced if the vehicle slowed down and traffic was diverted.
[0073] The simulation aims to assess the effects of different emergency strategies, such as speed reduction, speed limits, and early traffic diversion, on the overall traffic flow and predict the evolution of traffic flow after the implementation of these strategies. Based on the simulation results, the effectiveness of different strategies is evaluated, such as whether speed reduction can effectively mitigate danger and whether traffic diversion can reduce congestion. Finally, the optimal strategy is selected. If the simulation results indicate that speed reduction and early traffic diversion are the most effective measures, then the current strategy is adopted.
[0074] It should be noted that the calculation formulas and all parameters involved in the calculations in this application have been dimensionless beforehand. The process of dimensionless processing is well known in the industry and will not be described here.
[0075] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.
Claims
1. A method for detecting and responding to traffic anomalies on highways, characterized in that, The highway traffic anomaly detection and emergency response methods include the following steps: S1. Collect real-time vehicle data using pre-configured vehicle terminals and roadside equipment, and predict driving trajectories by combining driver history; compare the actual path with the predicted driving trajectory to identify abnormal vehicles whose deviation exceeds the set deviation threshold. S2. Analyze the behavior patterns of abnormal vehicles, combine spatiotemporal clustering algorithms to analyze the spatial distribution and temporal characteristics of accident types, infer the types of accidents that occurred, assess the risk level of accidents, and infer the points where anomalies occurred. S3. Based on the accident risk level, use edge nodes to issue lightweight tasks, and verify the speed, trajectory deviation and driving behavior of abnormal vehicles according to emergency response standards, and formulate a variety of emergency strategies based on the verification results. S4. Use a digital twin model to simulate the traffic conditions in the target area, evaluate the effectiveness of various emergency strategies, and then select the optimal emergency strategy to issue to the abnormal vehicle and other vehicles around it.
2. The method for detecting and responding to traffic anomalies on highways according to claim 1, characterized in that, The method involves collecting real-time vehicle data using pre-configured on-board terminals and roadside equipment, and then combining this data with the driver's historical behavior to predict the driving trajectory. Identifying abnormal vehicles whose deviations exceed a set deviation threshold by comparing the actual path with the predicted trajectory includes the following steps: S11. Real-time vehicle driving data, including speed, position, acceleration, and steering angle, of each vehicle in the target area is obtained through pre-configured vehicle terminals and roadside equipment. S12. Retrieve historical behavioral data from the driver, including common driving habits, historical trajectories, and records of abnormal behaviors; S13. Based on the vehicle driving data and the historical behavior data, generate the driving trajectory of each vehicle within a preset time period using a trajectory prediction model. S14. Compare the actual driving path of the vehicle with the driving trajectory within a preset time period in the future, and calculate the deviation between the two. S15. Compare the deviation with a preset deviation threshold, identify vehicles whose deviation exceeds the preset deviation threshold, and mark them as abnormal vehicles.
3. The method for detecting and responding to traffic anomalies on highways according to claim 1, characterized in that, The analysis of abnormal vehicle behavior patterns, combined with spatiotemporal clustering algorithms to analyze the spatial distribution and temporal characteristics of accident types, to infer the accident type, assess the accident risk level, and predict the anomaly occurrence point includes the following steps: S21. The predicted driving trajectory of the abnormal vehicle is spatiotemporally segmented, and the trajectory is divided into multiple trajectory segments based on the time window. The spatial features, temporal features and kinematic features of the predicted trajectory segments are extracted. S22. Use a spatiotemporal clustering algorithm to cluster the segmented trajectory segments, perform similarity analysis based on the spatial, temporal and kinematic features of the predicted trajectory segments, and mark abnormal trajectory segments during the clustering process. S23. Based on the abnormal trajectory segments, infer the type of accident that occurred and conduct a risk level assessment of the abnormal vehicles; S24. Based on the risk level assessment results, mark the location of the abnormal vehicle and infer the point of occurrence of the anomaly.
4. The method for detecting and responding to traffic anomalies on highways according to claim 3, characterized in that, The process of clustering the segmented trajectory segments using a spatiotemporal clustering algorithm, performing similarity analysis based on the spatial, temporal, and kinematic features of the predicted trajectory segments, and marking abnormal trajectory segments during the clustering process includes the following steps: S221. Standardize the spatial, temporal, and kinematic features of the acquired predicted trajectory segment; S222. Set the weights of the standardized spatial features, temporal features, and kinematic features, and define the spatiotemporal distance metric between predicted trajectory segments; S223. Set the spatial neighborhood threshold based on typical vehicle speed and road speed limit, set the spatial neighborhood threshold based on specific time windows, and set the minimum number of points by observing the expected number of trajectory segments in the same lane. S224. Based on the spatiotemporal distance metric, spatial neighborhood threshold, temporal neighborhood threshold, and minimum number of points, perform spatiotemporal clustering on all trajectory segments to obtain the cluster label and noise label for each trajectory segment. S225. Identify the trajectory segment marked as noise as an abnormal trajectory segment.
5. The method for detecting and responding to traffic anomalies on highways according to claim 4, characterized in that, The expression for the spatiotemporal distance metric between predicted trajectory segments is defined as follows: ； In the formula, Represents trajectory points With trajectory points The spatiotemporal distance between them; Weights representing spatial features; Weights representing time features; Weights representing kinematic characteristics; Represents trajectory points With trajectory points Spatial characteristics and The Euclidean distance between them; Represents trajectory points With trajectory points In terms of time characteristics and The time difference between them; Represents trajectory points With trajectory points Kinematic characteristics between and The kinematic differences between them.
6. The method for detecting and responding to traffic anomalies on highways according to claim 4, characterized in that, The process of performing spatiotemporal clustering on all trajectory segments based on spatiotemporal distance metrics, spatial neighborhood thresholds, temporal neighborhood thresholds, and minimum number of points to obtain cluster labels and noise labels for each trajectory segment includes the following steps: S2231. Use spatiotemporal distance metrics to obtain all neighboring trajectory segments in the spatial and temporal neighborhoods of each trajectory segment; S2232. Check whether the neighborhood of the current trajectory segment contains at least a minimum number of trajectory segments. If it contains at least a minimum number of trajectory segments, determine the current trajectory segment as the core point and create a new cluster based on the core point. S2233. Starting from the core point, add all trajectory segments within the neighborhood of the core point. If the trajectory segments within the neighborhood meet the condition of minimum number of points, continue to expand the current cluster. S2234. If the current trajectory segment cannot be classified into any cluster or is not the core point of any cluster and does not have enough neighborhood points, then the current trajectory segment is marked as noise. S2235. Assign each trajectory segment to a cluster, or label it as noise, and generate a cluster label and a noise label for each trajectory segment.
7. The method for detecting and responding to traffic anomalies on highways according to claim 6, characterized in that, The process of inferring the type of accident based on abnormal trajectory segments and assessing the risk level of abnormal vehicles includes the following steps: S231. Based on abnormal trajectory segments, analyze the kinematic characteristics of sudden speed changes, rapid acceleration or deceleration, and sharp turns to make a preliminary inference about the current accident type. S232. Based on the spatial characteristics of abnormal trajectory segments, determine whether they occur in specific road sections and whether there is persistent lane departure, and refine the prediction of accident types. S233. Based on the accident type, calculate the trajectory deviation of the abnormal trajectory segment; evaluate the speed change characteristics of the abnormal trajectory segment, and count the number of vehicles involved in the abnormal trajectory segment. S234. Weighted fusion of trajectory deviation, speed change, and number of vehicles to generate a comprehensive risk score; S235. Compare the comprehensive risk score with the preset risk level classification standard, and classify the abnormal vehicles into low-risk, medium-risk and high-risk categories based on the comparison results.
8. The method for detecting and responding to traffic anomalies on highways according to claim 1, characterized in that, Based on the accident risk level, lightweight tasks are issued using edge nodes, and the speed, trajectory deviation, and driving behavior of abnormal vehicles are verified according to emergency response standards. Based on the verification results, various emergency strategies are formulated, including the following steps: S31. Based on the accident risk level, distribute the lightweight task to the edge node; S32. At edge nodes, verify the speed, trajectory deviation, and driving behavior of abnormal vehicles based on emergency response standards; S33. Based on the verification results, assess the dangers of abnormal vehicle speed, trajectory deviation, and driving behavior, and formulate corresponding emergency strategies based on the assessment results.
9. The method for detecting and responding to traffic anomalies on highways according to claim 1, characterized in that, The process of using a digital twin model to simulate traffic conditions in a target area, evaluating the effectiveness of various emergency strategies, and then selecting the optimal emergency strategy to issue to the abnormal vehicle and other vehicles in its vicinity includes the following steps: S41. Obtain static environmental information and dynamic vehicle information of the target area; S42. Construct a static twin model of the scene based on the static environment information, and then construct a dynamic twin model of the vehicle based on the dynamic vehicle information; S43. Construct a digital twin model of the target area based on the static twin model of the scene and the dynamic twin model of the vehicle, and simulate the traffic conditions of the target area; S44. Input multiple emergency strategies into the digital twin model of the target area and perform simulation calculations to predict the impact and evolution of emergency strategies on traffic flow. S45. Evaluate the simulation results of multiple emergency strategies, select the optimal emergency strategy based on the evaluation results, and distribute the optimal emergency strategy to the abnormal vehicle and other relevant vehicles around it.