Trajectory entropy-based method and system for warning of affected vehicles in a road-bridge collapse event

Abstract

The application relates to the technical field of traffic transportation, and provides a warning method and system for vehicles affected by a road-bridge disaster event based on trajectory entropy. The behavior feature sequence of a vehicle in a time period before and after a target road-bridge disaster event is extracted, and the trajectory entropy value of the vehicle is calculated through a trajectory entropy model, so as to identify the vehicle affected by the disaster event. Based on the vehicle set affected by the disaster event, the warning range of the target road-bridge disaster event is dynamically determined according to real-time traffic flow data. Through the vehicle trajectory big data, the abnormal degree of the vehicle driving behavior is quantified, the vehicle affected by the disaster is accurately, efficiently and dynamically identified, the warning range is scientifically determined, reliable data support is provided for emergency response and traffic relief, and the processing efficiency and safety of the road-bridge disaster event are effectively improved.

Description

Technical Field

[0001] This application relates to the field of transportation technology, and in particular to a method and system for early warning of affected vehicles in road and bridge disasters based on trajectory entropy. Background Technology

[0002] Road and bridge structures are vulnerable to sudden events such as geological disasters, extreme weather, and traffic accidents during operation, leading to serious disasters such as bridge collapses and roadbed damage. These events not only disrupt traffic but also pose a fatal threat to vehicles in motion. Therefore, after a disaster, quickly and accurately identifying affected vehicles and delineating a scientifically defined warning area is crucial for providing decision support to transportation authorities and emergency departments in carrying out targeted rescue, evacuation, and traffic control measures. This is essential for improving road traffic safety and emergency management capabilities. Summary of the Invention

[0003] The purpose of this application is to provide a method and system for early warning of affected vehicles in road and bridge disasters based on trajectory entropy, so as to solve or alleviate the problems existing in the prior art.

[0004] To achieve the above objectives, this application provides the following technical solution: This application provides a method for early warning of affected vehicles in road and bridge disaster events based on trajectory entropy, including: extracting the time period before the occurrence of the target road and bridge disaster event. and the time period after the occurrence The system analyzes the behavioral characteristic sequences of vehicles; based on these sequences, it calculates the trajectory entropy values ​​of vehicles using a constructed trajectory entropy model to identify vehicles affected by the disaster; and it then identifies the set of vehicles affected by the disaster. Based on real-time traffic flow data, the early warning range for target road and bridge disaster events is dynamically defined.

[0005] Preferably, the time period prior to the occurrence of the target road and bridge disaster event is obtained. and the time period after the occurrence Trajectory data of vehicles within the target road and bridge ; trajectory data After data cleaning, the vehicle's trajectory points are matched to the actual road segments where the target road and bridge disaster occurred based on the map matching algorithm, resulting in a structured trajectory sequence of the vehicle. Extract the vehicle's behavioral feature sequence from the vehicle's structured trajectory sequence.

[0006] Preferred, with vehicles At any moment trajectory points Determine the vehicle as the center. At any moment The set of candidate road segments ; Calculation time vehicle In the candidate path set The Middle Candidate road sections Observation probability and calculation from time 1 At the time vehicle From candidate road sections Move to candidate road segment transition probability ; Based on observation probability and transition probability To determine the vehicles at the time of the target road and bridge disaster. The candidate road segment sequence in which it is located; Responding to vehicles when the target road or bridge collapse occurs If the candidate road segment sequence matches the road segment where the target road / bridge disaster occurred, then the vehicle will be... The corresponding trajectory points are vertically projected onto the geometric alignment of the matching road segment to obtain the vehicle. Structured trajectory sequences.

[0007] Preferably, each trajectory point data item in the structured trajectory sequence includes at least: a timestamp, location information, and speed information; the vehicle's behavioral feature sequence is extracted from the vehicle's structured trajectory sequence, including: based on the vehicle's behavior at the time of the target road and bridge disaster event. The projected position and timestamp of each trajectory point are used to calculate the vehicle. The instantaneous speed of the vehicle at the time of the target road and bridge disaster was determined. velocity sequence; and through vehicles Calculate the speed difference and time difference between two consecutive trajectory points to determine the vehicle's position at the time of the target road / bridge disaster event. The acceleration sequence.

[0008] Preferably, each trajectory point data item in the structured trajectory sequence further includes: a road segment identifier (link_id); extracting the vehicle's behavioral feature sequence from the vehicle's structured trajectory sequence also includes: The structured sequence of vehicles is divided into several consecutive road segments based on the changes in the road segment identifier link_id; Calculate the travel time ratio for each road segment, and arrange the travel time ratios of all road segments in chronological order to obtain the travel time ratio sequence for the vehicles.

[0009] Preferably, the structured sequence of the vehicle is divided into several continuous road segments according to the change of the road segment identifier link_id, including: sorting the structured sequence of the vehicle in ascending order of timestamps, and traversing the trajectory points of the structured sequence according to the time sequence; In response to the timestamp-ordered ascending sequence trajectory points If the corresponding road segment identifier (link_id) matches the vehicle's current road segment ID, then the trajectory point will be... Add to the end of the list of trajectory points for the vehicle's current route; among them, It is a positive integer, and ; Response to trajectory points If the corresponding road segment identifier link_id does not match the vehicle's current road segment ID, then the vehicle will be sorted by timestamp in ascending order. trajectory points The timestamp record is the end time of the vehicle's current road segment journey. Therefore, the vehicle's current road segment journey ends, and the time is recorded as a trajectory point. The corresponding road segment identifier link_id creates a new road segment trip.

[0010] The preferred trajectory entropy model is: In the formula, For vehicles The trajectory entropy value, For vehicles Driving status in trajectory data The probability of occurrence This is a set of vehicle driving states.

[0011] Preferably, based on the time period prior to the occurrence of the target road / bridge disaster event. The trajectory entropy values ​​of vehicles within the road are used to construct the baseline range of trajectory entropy for the target road and bridge. ;in, The period before the target road and bridge disaster occurred. The mean trajectory entropy of vehicles inside the vehicle, The period before the target road and bridge disaster occurred. Standard deviation of the trajectory entropy of vehicles inside the vehicle; In response to ,or, Then determine the time period after the target road and bridge disaster event. Vehicles inside For vehicles affected by the disaster; in, For entropy adjustment parameters, For the preset threshold, The time period following the target road and bridge disaster Vehicles inside The trajectory entropy value, The period before the target road and bridge disaster occurred. Inner vehicle The trajectory entropy value.

[0012] Preferably, in the road network topology, the set of vehicles affected by the disaster event is determined by tracing back upstream from the point of failure of the target road / bridge. The location of the furthest vehicle is the initial upstream boundary of the warning range. ; Based on the average deceleration and stagnation of vehicles after the target road and bridge disaster, a traffic wave velocity model is used to predict the congestion propagation speed of the congestion impact upstream after the disaster. ; According to the formula: Determine the upstream boundary of the warning area .

[0013] This application embodiment also provides an early warning system for affected vehicles in road and bridge disaster events based on trajectory entropy. The system dynamically delineates the early warning range of a target road and bridge disaster event using any of the aforementioned prediction methods for affected vehicles in road and bridge disaster events based on trajectory entropy. The system includes: The vehicle behavior feature unit is configured to extract the time period preceding the target road / bridge disaster event. and the time period after the occurrence The behavioral characteristic sequence of vehicles inside; The vehicle impact range unit is configured to calculate the trajectory entropy value of the vehicle based on the vehicle's behavioral characteristic sequence through a constructed trajectory entropy model, in order to identify vehicles affected by the disaster event. The warning scope is divided into units, configured based on the set of vehicles affected by the disaster event. Based on real-time traffic flow data, the early warning range for target road and bridge disaster events is dynamically defined.

[0014] Beneficial effects: The early warning method and system for affected vehicles in road and bridge disaster events based on trajectory provided in this application embodiment extracts the time period before the occurrence of the target road and bridge disaster event. and the time period after the occurrence The system analyzes the behavioral characteristic sequences of vehicles and calculates their trajectory entropy values ​​using a constructed trajectory entropy model to identify vehicles affected by disaster events. It then uses this set of vehicles affected by disaster events as a basis for further analysis. The warning range for target road and bridge disaster events is dynamically determined based on real-time traffic flow data.

[0015] Therefore, by utilizing vehicle trajectory big data, the abnormality of vehicle driving behavior can be quantified to accurately, efficiently, and dynamically identify vehicles affected by disasters, enabling the scientific delineation of the early warning range, providing reliable data support for emergency response and traffic management, and effectively improving the efficiency and safety of handling road and bridge disasters. Attached Figure Description

[0016] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. Wherein: Figure 1 This is a flowchart illustrating an early warning method for affected vehicles in a road and bridge disaster event based on a trajectory, according to some embodiments of this application. Figure 2 A logical framework diagram of an early warning method for affected vehicles in a road and bridge disaster event based on trajectory, provided according to some embodiments of this application; Figure 3 This is a schematic diagram of vehicle driving status distribution during a period prior to a disaster event, according to an embodiment of this application. Figure 4 This is a schematic diagram illustrating the distribution of vehicle driving status over a period of time following a disaster event, according to an embodiment of this application. Figure 5 This is a comparative diagram of the traditional static warning range division and the dynamic warning range division provided in this application according to the embodiments of this application; Figure 6 This is a schematic diagram of the structure of an early warning system for affected vehicles in a road and bridge disaster event based on a trajectory, according to some embodiments of this application. Detailed Implementation

[0017] The present application will now be described in detail with reference to the accompanying drawings and embodiments. Various examples are provided by way of explanation and not by way of limitation. In fact, those skilled in the art will understand that modifications and variations can be made to the present application without departing from the scope or spirit of the present application. For example, a feature shown or described as part of one embodiment may be used in another embodiment to produce yet another embodiment. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0018] Currently, research on determining the warning range for vehicles affected by disasters mainly includes static delineation methods based on fixed geometric distances, macroscopic estimation methods based on traffic flow theory, and simple spatiotemporal filtering methods based on vehicle trajectories. Among them, the static delineation method based on fixed geometric distances delineates a fixed circular or strip-shaped area as the warning range, centered on the disaster point, based on experience or simple rules (such as "X kilometers upstream"). Its significant drawback is that it does not consider the actual road network structure (such as ramps and interchanges), real-time traffic flow status (such as vehicle speed and density), and the driving intentions of the vehicles themselves. This results in the warning range being either too large (including a large number of unaffected vehicles, causing a waste of warning resources and excessive panic) or too small (missing vehicles that are truly in danger, causing serious safety hazards).

[0019] Macroscopic estimation methods based on traffic flow theory utilize macroscopic traffic flow parameters (such as average vehicle speed and occupancy) acquired by fixed sensors such as loop detectors and cameras, and then extrapolate the congestion backtracking distance using traffic wave theory (shock wave model). However, this method relies heavily on the deployment density and integrity of fixed sensors, and its reliability drops significantly in areas with insufficient sensor coverage. Furthermore, macroscopic traffic flow parameters cannot accurately reflect the microscopic behavioral changes of individual vehicles, making it difficult to distinguish between vehicles hindered by disasters and those moving slowly due to ordinary congestion or their own reasons (such as stopping for rest), resulting in limited identification accuracy.

[0020] With the widespread adoption of vehicle-to-everything (V2X) and floating car technologies, obtaining real-time vehicle trajectory data has become possible. Existing simple spatiotemporal filtering methods based on vehicle trajectories identify affected vehicles by determining whether their trajectory points enter a pre-defined fixed geofence or whether they stop moving within a specific time period. This method's logic is overly simplistic and cannot effectively distinguish between abnormal stagnation caused by disasters and normal waiting at traffic lights, congestion queues, or temporary stops, resulting in high false alarm and false negative rates. Furthermore, it lacks effective predictive capabilities for vehicles that have not yet reached the disaster site but are about to enter the danger zone.

[0021] Whether it's a static delineation method based on fixed geometric distances, a macroscopic estimation method based on traffic flow theory, or a simple spatiotemporal filtering method based on vehicle trajectories, all suffer from shortcomings such as inaccurate delineation of warning ranges, low level of intelligence, inability to dynamically adapt, and inability to accurately identify affected vehicles at the individual level. The core problem lies in the lack of intelligent analysis capable of profoundly quantifying the abnormal impact of disaster events on individual vehicle behavior and deeply integrating road network topology and real-time traffic conditions.

[0022] Based on this, this embodiment provides an early warning method for affected vehicles in road and bridge disasters based on trajectory entropy. It makes full use of vehicle trajectory big data, quantifies the degree of abnormality of vehicle driving behavior through models, accurately, dynamically and efficiently identifies vehicles affected by disasters, and scientifically delineates the early warning range, providing a reliable decision-making basis for emergency command and improving rescue efficiency and road traffic safety.

[0023] like Figures 1 to 5 As shown, this early warning method for affected vehicles in road and bridge disaster events based on trajectory entropy includes: Step S101: Extract the time period before the target road and bridge disaster event. and the time period after the occurrence The behavioral characteristic sequence of vehicles inside.

[0024] In this embodiment, the time period prior to the occurrence of the target road and bridge disaster event is obtained. and the time period after the occurrence Within the area where the target road and bridge disaster occurred, all collectable vehicle trajectory data. Each trajectory data point includes at least the vehicle's unique identifier (ID), timestamp, latitude and longitude coordinates, instantaneous speed, and direction of travel.

[0025] The acquired vehicle trajectory data Preprocessing is performed. This includes data cleaning to remove obvious outliers (such as speeds exceeding reasonable limits or coordinate drift points); and using map matching algorithms and high-precision road network maps, matching vehicle trajectory points (GPS, BeiDou, etc.) to the actual road segments where the target road and bridge disaster occurred to obtain a structured trajectory sequence of the vehicle, and extracting the vehicle's behavioral feature sequence from the structured trajectory sequence.

[0026] For trajectory interruptions caused by signal loss, a path inference algorithm is used for reasonable interpolation and completion. The path inference algorithm, given partial trajectory points, reasonably infers the missing portion of the vehicle's trajectory based on road network constraints, vehicle motion models, and motion patterns. Commonly used methods include geometric interpolation (linear interpolation, spline interpolation), road network-based matching inference (shortest path algorithm (Dijkstra, A*)), and statistical and machine learning methods (Hidden Markov Model (HMM), Kalman filtering, Recurrent Neural Network (RNN)). In this embodiment, a path inference algorithm based on high-precision road network topology is used to complete the missing segments in the trajectory data caused by signal loss.

[0027] Specifically, first, the structured trajectory sequence of the vehicle is traversed, and when the time interval between two consecutive trajectory points in the structured trajectory sequence... Exceeding the preset threshold If the time is 30 seconds, the vehicle trajectory is determined to be interrupted, and adjacent trajectory points before and after the trajectory interruption are marked; among them, the last trajectory point before the vehicle trajectory interruption is... The first trajectory point after the interruption is .

[0028] Then, the trajectory points Mapping to the nearest road network node via map matching , track points Mapping to the nearest road network node via map matching The Dijkstra shortest path algorithm is used to calculate the path from each node in the road network topology. To road network nodes Optimal connected path Among them, the connected path Composed of a series of orderly road sections composition.

[0029] Next, the average speed of the vehicle during the trajectory interruption is estimated. Among them, average driving speed For trajectory points Average speed and connected path within a preset time period before the point The smaller of the two speed limits for the corresponding road segment, and expressed at a fixed time resolution. (e.g., 5 seconds), according to the formula: Calculate the vehicle's travel distance And along the connected path The geometric shape of the line generates a series of interpolation points. and for interpolation points timestamp In the interval Linear allocation is performed within the range; For trajectory points The corresponding timestamp, For trajectory points The corresponding timestamp.

[0030] Then, map matching is performed on the generated interpolation point sequence to assign it the correct road segment identifier (link_id) and corresponding road attributes, and it is then inserted into the vehicle trajectory data. From the preprocessed structured trajectory sequence, a complete and uninterrupted structured trajectory sequence of the vehicle is obtained. Then, the vehicle's behavioral feature sequence is extracted from the complete and uninterrupted structured trajectory sequence.

[0031] When extracting vehicle behavior feature sequences from complete, uninterrupted structured trajectory sequences, the process begins by acquiring the vehicle's GPS trajectory sequence. Each trajectory point in the sequence must contain at least latitude, longitude, and a timestamp. Simultaneously, high-precision road network data (high-precision electronic road network map) containing links and nodes is acquired. Each link includes attributes such as road class, number of lanes, speed limit, direction, and geometry. Then, based on the vehicle... At any moment trajectory points Determine the vehicle as the center. At any moment The set of candidate road segments In the vehicle's GPS trajectory sequence, a trajectory point at a certain time is used as the center, and a reasonable search radius (usually 1000 m) is used. (meters), retrieve all road segments falling within this search radius from high-precision road network data, and call these searched road segments candidate road segments for the trajectory point. The set of candidate road segments is the candidate road segment set corresponding to the trajectory point of the vehicle in this event.

[0032] Next, calculate the time. vehicle In the candidate path set The Middle Candidate road sections Observation probability To characterize the vehicle in the first Candidate road sections The observed trajectory points The possibility of [the following is unclear and requires context: "the trajectory points are calculated using a Gaussian distribution model"]. To candidate road sections vertical distance (i.e., trajectory points) To candidate road sections The shorter the distance of the corresponding line segment, the closer the distance, the more likely the vehicle is to be on the first line segment. Candidate road sections The observed trajectory points The greater the probability, the higher the likelihood. Specifically, according to the observation probability model: Calculation time vehicle In the candidate path set The Middle Candidate road sections Observation probability In the formula, This represents a typical error in GPS devices. rice.

[0033] At the same time, calculate from time 10:00 At the time vehicle From candidate road sections Move to candidate road segment transition probability This characterizes the probability of a vehicle moving from a candidate road segment in the previous time step to a candidate road segment in the current time step. Specifically, it is based on the Euclidean distance between two consecutive trajectory points and the distance between two candidate road segments (candidate road segments...). Candidate road sections The shortest path distance along the road network determines the vehicle's transition probability. Specifically, according to the transition probability model: Calculate from time 1 At the time vehicle From candidate road sections Move to candidate road segment transition probability In the formula, For vehicles At any moment trajectory points With time trajectory points The Euclidean distance between them, where the trajectory points With trajectory points These are two consecutive GPS coordinate points; For vehicles In the candidate road segment Move to candidate road segment The shortest path distance along the road network; This is a control parameter used to control the width of the transition probability distribution.

[0034] In this embodiment, the matching accuracy and algorithm robustness are balanced by adjusting parameter β, which can be determined through heuristic settings, data-driven optimization, and other methods. Specifically, in the heuristic settings, the positioning error is based on the positioning system (such as GPS, BeiDou, etc.). Configure settings. The range of values ​​is to For example, regarding positioning errors For consumer-grade GPS devices, the adjustment parameter β can be set to an empirical value between 30 meters and 75 meters.

[0035] In data-driven optimization, the control parameter β is determined based on historical trajectory datasets. Specifically, firstly, historical trajectory datasets of actual driving paths are collected as the training set; then, within a pre-defined parameter space (such as...), the control parameter β is determined. Perform a grid search within a meter; for each candidate The matching algorithm of this embodiment is run on the training set, and the accuracy of the matching results against the real path is calculated. The matching algorithm with the highest accuracy is selected. The value serves as the final control parameter.

[0036] In addition, control parameters can be established. The mapping relationship between the control parameters and the road environment characteristics is used to determine the control parameters. For example, in highway areas with sparse road networks and wide lanes, larger [vehicles / engines] are used. To enhance robustness; in complex urban areas with dense road networks and parallel auxiliary roads or elevated structures, a smaller value is used. Values ​​are used to improve matching accuracy. Environmental features can be obtained from attributes such as road grade, number of lanes, and area type in high-precision road network data.

[0037] Then, based on the observation probability and transition probability To determine the vehicles at the time of the target road and bridge disaster. The candidate road segment sequence in which it is located. Specifically, a dynamic programming algorithm is used to find the globally optimal candidate road segment sequence, according to the model: Determine the vehicles at the time of the target road and bridge disaster. The candidate road segment sequence in which it is located; where, For candidate path set The number of candidate road sections. All are positive integers.

[0038] Finally, the vehicles at the time of the target road and bridge disaster. The candidate road segment sequence is matched with the road segment where the target road / bridge disaster event occurred (road segment ID matching). If the vehicle was at the time of the target road / bridge disaster event... If the candidate road segment sequence matches the road segment where the target road / bridge disaster event occurred, then the vehicle is projected using a projection method. The corresponding trajectory points are vertically projected onto the geometric alignment of the matching road segment to determine the vehicle's precise position on that segment, thus obtaining the vehicle's position. The structured trajectory sequence. Each trajectory point data item in the structured trajectory sequence contains at least: timestamp, link identifier, location information, and speed information.

[0039] Then, the vehicle's behavioral feature sequence is extracted from the vehicle's structured trajectory sequence; wherein, the behavioral feature sequence includes at least: a speed sequence. , acceleration sequence Travel time ratio sequence Specifically, based on the vehicles at the time of the target road / bridge disaster event. The projected position and timestamp of each trajectory point are used to calculate the vehicle. The instantaneous speed of the vehicle at the time of the target road and bridge disaster was determined. velocity sequence ; and by calculating the speed difference and time difference between two consecutive trajectory points of the vehicle, the vehicle at the time of the target road and bridge disaster event is calculated. acceleration sequence .

[0040] The structured trajectory sequence obtained through map matching is based on the vehicle. The projected position and timestamp of each trajectory point are used to calculate the vehicle. The instantaneous velocity is used to obtain a more accurate instantaneous velocity sequence than the original velocity measured by weighing. Among them, according to the vehicle Any two consecutive trajectory points in a structured trajectory sequence and trajectory points Travel distance along the road network and travel time difference Identify the vehicle at the trajectory point instantaneous speed .

[0041] Specifically, depending on the vehicle Any two consecutive trajectory points in a structured trajectory sequence and trajectory points timestamp Calculate vehicles At the trajectory point and trajectory points Travel time difference when traveling along the road network ( Meanwhile, in determining the vehicle At the trajectory point and trajectory points Travel distance when traveling along the road network At that time, if the vehicle trajectory points and trajectory points Matched to the same road segment (i.e., trajectory points) and trajectory points If the corresponding road segment identifier (link_id) is the same, then the travel distance is... For trajectory points and trajectory points The distance of the curve along the shape line of the set on this driving segment; if the vehicle trajectory points and trajectory points Matched to different road segments (i.e., trajectory points) and trajectory points If the corresponding road segment identifiers (link_id) are different, then the travel distance is... For trajectory points To the trajectory point The total length of the shortest feasible path along the road network topology, which is determined in a high-precision road network using a shortest path algorithm (such as Dijkstra's algorithm).

[0042] Then, according to the formula: vehicle at the trajectory point instantaneous speed For vehicles The structured trajectory sequence is traversed to calculate the vehicle... The instantaneous velocity across all trajectory segments yields the vehicle's velocity at the time of the target road / bridge collapse event. velocity sequence .

[0043] In this embodiment, the travel time ratio sequence of the vehicle is used. Quantifying the degree of decline in road traffic efficiency. Specifically, the structured sequence of vehicles is divided into several consecutive road segment journeys according to the changes in the road segment identifier link_id. For each road segment journey, the ratio of the actual time taken by the vehicle to the baseline time taken in free-flow conditions is calculated, yielding the vehicle's journey time ratio for that road segment. Then, the journey time ratios of vehicles across all road segments are arranged in chronological order to obtain a sequence of vehicle journey time ratios. Among these, the vehicle's journey time ratio in the first... Trip to each section Free-flow reference time According to the formula: In the formula, For the first Trip to each section Length, For the first Trip to each section The free-flow velocity is obtained from the road network attribute library or historical data.

[0044] In a specific example, when dividing a vehicle's structured sequence into several consecutive road segments based on changes in the link_id, the vehicle's structured sequence is first sorted in ascending order by timestamp, and then the trajectory points of the structured sequence are traversed in chronological order. Specifically, for the first trajectory point arranged in chronological order, the link_id is empty, indicating the start of the vehicle's journey. A new journey object is created using the link_id corresponding to the first trajectory point as the segment identifier for the journey. And use the timestamp of the first trajectory point as the trip object. The trip start time is used to add the first trajectory point to the trip object. The list of trajectory points.

[0045] When sorted in ascending order by timestamp ( It is a positive integer, and ) trajectory points If the corresponding road segment identifier (link_id) matches the vehicle's current road segment ID, the vehicle will travel on the current road segment, and the trajectory points will be... Add to the end of the list of trajectory points for the vehicle's current route; when trajectory points If the corresponding road segment identifier `link_id` does not match the vehicle's current road segment ID, then the vehicle has left the original road segment and entered a new road segment. The new segment will be sorted by timestamp in ascending order. trajectory points The timestamp record is the end time of the vehicle's current road segment journey. Therefore, the vehicle's current road segment journey ends, and the time is recorded as a trajectory point. The corresponding road segment identifier, link_id, is used to create a new road segment trip. and the trajectory points The timestamp records the route journey. The start time of the journey, the trajectory points Add to newly created route trip middle.

[0046] After traversing all trajectory points, the current journey may still be under construction (i.e., the vehicle is traveling on the last road segment). In this case, the timestamp of the last trajectory point is recorded as the end time of the current journey, and the journey is added to the road segment journey set. Finally, the road segment journey set is output. Each element in the road segment journey set represents the complete travel record of the vehicle on a specific, continuous road segment. Each road segment journey contains the unique identifier of the road segment traveled by the vehicle (link_id), the timestamps of the vehicle's entry time (Start_time) and exit time (end_time), the actual travel time of the vehicle on the road segment journey (duration = end_time - Start_time), all matching trajectory points of the vehicle on the road segment journey, and the road segment attributes (length, free-flow speed, etc.).

[0047] Step S102: Based on the vehicle's behavioral characteristic sequence, calculate the vehicle's trajectory entropy value using the constructed trajectory entropy model to identify vehicles affected by the disaster event.

[0048] Disasters disrupt normal vehicle driving patterns. Before a disaster, vehicle driving states are relatively stable; after a disaster, affected vehicles may exhibit abnormal states such as rapid acceleration or prolonged stagnation, leading to drastic changes in the vehicle driving state distribution. In this embodiment, a trajectory entropy model is constructed to quantify the degree of abnormality in the driving state of each vehicle before and after a disaster. Specifically, based on the extracted vehicle behavioral characteristics, the trajectory entropy model is used as follows: Calculate vehicles trajectory entropy In the formula, For vehicles Driving status A collection of vehicles Driving status This includes, but is not limited to, "high-speed driving," "slow-speed driving," "rapid acceleration," and "stationary," among which, the vehicle Driving status By vehicle velocity sequence , acceleration sequence Travel time ratio sequence definition; For vehicles Driving status in the driving trajectory The probability of occurrence.

[0049] In a specific application scenario, define a set of formal states. When the vehicle speed And vehicle acceleration At that time, it is determined that the vehicle is in motion. (Free-flow: high-speed travel); when the vehicle speed And vehicle acceleration At that time, it is determined that the vehicle is in motion. (Following the flow: moving slowly at low speed); when the vehicle speed At that time, it is determined that the vehicle is in motion. (Congestion and slow traffic); when vehicle speed And duration At that time, it is determined that the vehicle is in motion. (Stagnation).

[0050] Meanwhile, statistics were compiled on the time period before the target road and bridge disaster occurred. Within the time frame, the frequency of each driving state is counted, and its probability distribution is calculated to obtain the probability of each driving state occurring in the vehicle's trajectory before the disaster event. Then, the trajectory entropy value of the vehicle before the target road / bridge disaster event is calculated. Similarly, the time frame of the vehicle after the target road / bridge disaster event is statistically analyzed. Within the vehicle, the frequency of each driving state is counted, and its probability distribution is calculated to obtain the probability of each driving state occurring in the vehicle's trajectory after the disaster event. Then, the trajectory entropy value of the vehicle after the disaster event on the target road and bridge is calculated.

[0051] By calculating trajectory entropy values, the distribution changes of vehicle driving states before and after a disaster event are depicted. By comparing the obtained vehicle trajectory entropy values ​​with preset thresholds, the groups of vehicles affected by the disaster event can be identified. Specifically, firstly, a baseline entropy is established: based on the time period before the target road / bridge disaster event. The trajectory entropy values ​​of vehicles within the road are used to construct the baseline range of trajectory entropy for the target road and bridge. In other words, the statistics cover the period prior to the target road / bridge disaster. The distribution of trajectory entropy values ​​of all or sampled vehicles within the area is used to establish a baseline range for the trajectory entropy of the target road and bridge. Among these, The period before the target road and bridge disaster occurred. The mean trajectory entropy of vehicles inside the vehicle, The period before the target road and bridge disaster occurred. Standard deviation of the trajectory entropy of vehicles inside the vehicle.

[0052] Then, abnormal vehicle identification is performed based on the established baseline entropy and the vehicle's trajectory entropy value. Specifically, when... ,or, Then determine the time period after the target road and bridge disaster event. Vehicles inside These are vehicles affected by the disaster. Among them, This is the entropy adjustment parameter, which can take the value of 2 or 3. For the preset threshold ( 0.5), The time period following the target road and bridge disaster Vehicles inside The trajectory entropy value, The period before the target road and bridge disaster occurred. Inner vehicle The trajectory entropy value.

[0053] like Figure 3 As shown in the diagram, the probability distribution of vehicle driving states reveals a relatively dispersed and balanced distribution. Most vehicle events occur in the efficient "free flow" and "car-following flow" states, while "congested slow-moving" and "stagnant" states account for a relatively low proportion. The relatively flat probability distribution of vehicle driving states and the high calculated trajectory entropy value indicate rich and uncertain vehicle operating states, reflecting a healthy traffic flow. Figure 4 As shown, the probability distribution of vehicle driving states reveals that the distribution becomes highly concentrated and uneven, with the probability of the "stagnant" state soaring and becoming the dominant state, while the "free-flowing" state is almost nonexistent. Due to the highly concentrated probability distribution of the driving states (focused on the "stagnant" state), the uncertainty is greatly reduced, and the calculated trajectory entropy value drops sharply, intuitively quantifying the enormous impact and interference of the disaster event on the vehicle driving state.

[0054] By capturing and quantifying the probability distribution changes and trajectory entropy changes of vehicle driving status, it is possible to accurately determine whether a vehicle has been affected by a disaster event. This goes beyond simple position or speed judgment and deeply mines the "abnormal behavior patterns" hidden in the trajectory data, thereby achieving higher-precision identification.

[0055] Step S103: Based on the set of vehicles affected by the disaster event Based on real-time traffic flow data, the early warning range for target road and bridge disaster events is dynamically defined.

[0056] All satisfied or Vehicles are marked as affected vehicles, and a set of vehicles affected by the disaster event is constructed. In conjunction with the road network topology and real-time traffic flow data, the early warning range after a target road or bridge disaster event is dynamically defined.

[0057] Traditional static early warning methods define a fixed circular or fan-shaped area (e.g., 5 kilometers upstream) centered on the disaster site (the location of the disaster event). This circular area fails to accurately reflect the actual road network, covering a large number of unrelated fields, buildings, or vehicles on parallel roads (false alarms). It cannot reflect the true spread of the disaster's impact; if the congestion echo has already propagated to 8 kilometers, a large number of affected vehicles will be missed within a 5-kilometer radius (underreporting). In contrast, this application's dynamic early warning system based on trajectory entropy can accurately identify all "affected vehicles" and determine the location of the farthest affected vehicle (the farthest affected vehicle). Combining a traffic wave propagation model, it calculates the theoretically farthest distance the congestion should propagate based on the time since the disaster event and the estimated wave velocity. (Theoretical propagation boundary) Compare the location of the farthest vehicle with the theoretical propagation boundary, and take its upstream as the upstream boundary of the warning range.

[0058] First, in the road network topology, trace back upstream from the point of failure of the target road / bridge disaster event to determine the set of vehicles affected by the disaster event. The location of the furthest vehicle is the initial upstream boundary of the warning range. Specifically, In the formula, For vehicles along the road network to the disaster site The actual distance.

[0059] Then, based on the average deceleration and stagnation of vehicles after the target road and bridge disaster, the congestion propagation speed of the congestion impact on the upstream side is predicted using a traffic wave velocity model. This allows for the determination of the final upstream boundary of the warning area. .

[0060] In a specific example, the traffic wave velocity model adopts a shock wave model, and the wave velocity is estimated based on the average deceleration characteristics of vehicles and road characteristics after a disaster event. Specifically, firstly, the free-flow velocity of vehicles is determined. and average deceleration This can be achieved by querying the design value of the free-flow velocity of the road segment upstream of the disaster event from a high-precision road network attribute database; or by querying the time period before the disaster event. The free-flow velocity of vehicles is determined by statistically analyzing 85% of the vehicle's trajectory. Here, based on the identified set of affected vehicles... The post-disaster trajectory was analyzed, and the average deceleration of each vehicle from normal speed to rest was calculated. The average or median of all vehicles was then taken as the average deceleration of the vehicles. ( (Negative values). Simultaneously, the average reaction time of vehicle drivers was determined. (Empirical values ​​for traffic engineering, usually taken as...) (seconds) and average safe parking distance (usually taken as...) rice).

[0061] Then, a simplified shockwave model based on vehicle following behavior is adopted: Calculate the speed of congestion propagation of vehicles In the formula, the negative sign indicates that the wave propagates in the opposite direction (i.e., upstream) of the vehicle's travel.

[0062] After obtaining the speed of vehicle congestion propagation Then, calculate the time period following the occurrence of the target road and bridge disaster event. The theoretical maximum distance that congestion shock waves can propagate upstream. ( This allows for adjustments to the upstream boundary of the warning area, determining the final upstream boundary of the warning area. In this embodiment, the early warning boundary model is used: Determine the upstream boundary of the warning area This ensures that early warnings are effectively issued to vehicles about to enter the disaster-affected area. Simultaneously, vehicles merging on all roads within the boundary and at key nodes (such as interchanges and entrance ramps) also face risk. Therefore, by extending the early warning range to all vehicles potentially heading towards the disaster site at key road network nodes such as interchanges and ramps, based on vehicle travel direction, the warning system extends the warning range to all vehicles possibly heading towards the disaster site. By identifying the associated road sections, it effectively avoids missing early warnings and forms an irregular, precise polygonal early warning area extending along the road network.

[0063] Final output warning range A polygonal geographic region, including the location of the disaster event and its upstream boundary. This includes the area enclosed by the road boundaries. Furthermore, it connects to traffic guidance screens, navigation apps, and other terminals to issue warning information to vehicles within the warning area. For example, it pushes red congestion warnings and voice prompts such as "Bridge ahead has collapsed, please detour immediately" to all vehicle users within the warning area.

[0064] In this way, warning areas are dynamically delineated based on the actual location of affected vehicles and road structure. The warning range changes dynamically as affected vehicles move and traffic waves spread, effectively guiding the deployment of rescue forces, fixed-point traffic control, and the accurate release of warning information by navigation software, greatly improving emergency response efficiency.

[0065] In the early warning system provided in this embodiment, the data layer, acting as a processing pipeline from raw data to usable information, plays a crucial role in map matching and structuring, forming the foundation for subsequent analysis. The analysis layer implements the entire process from feature extraction to state pattern distribution. The decision layer makes decisions throughout the entire process, from trajectory entropy calculation and abnormal vehicle identification to affected vehicle marking, clarifying the judgment logic for identifying affected vehicles and calculating trajectory entropy values ​​based on the probability distribution of vehicle driving states, ensuring the accuracy and robustness of affected vehicle identification. In the early warning generation layer, continuous "areas" (early warning ranges) are generated from discrete "points" (affected vehicles), and the scientific validity and practicality of early warning generation are effectively ensured through traffic wave models and road network topology adjustments. The output layer displays the entire process of early warning information release, push notifications, and emergency response. Thus, a progressive system is achieved from the occurrence of a disaster event to the issuance of an early warning, demonstrating the advanced nature of data-driven decision-making in intelligent traffic emergency management.

[0066] Compared to traditional fixed-range early warning methods, the early warning method in this embodiment uses an information-theoretic indicator—trajectory entropy—to deeply quantify the impact of disasters at the micro-level of individual behavior. This effectively distinguishes between disaster-induced stagnation and normal parking, greatly improving the accuracy of vehicle identification and the precision of the early warning range. During the early warning process, the warning range is dynamically adjusted based on the real-time identified locations of affected vehicles and traffic flow conditions, rather than remaining fixed. This fully and accurately reflects the actual spread of the disaster's impact, avoiding resource waste caused by an excessively large range and preventing omissions due to an excessively small range, thus achieving precise and efficient emergency command. The entire early warning process does not rely on fixed hardware sensors but utilizes widely available floating car data, resulting in broad coverage and strong resistance to single-point failures. Furthermore, the early warning process deeply integrates big data mining, machine learning (state classification), and traffic engineering theory (traffic waves), achieving automated and intelligent analysis from data to decision-making, saving valuable time for emergency response.

[0067] like Figure 6 As shown, this embodiment also provides an early warning system for affected vehicles in road and bridge disaster events based on trajectory entropy. The system dynamically delineates the early warning range of the target road and bridge disaster event using the prediction method for affected vehicles in road and bridge disaster events based on trajectory entropy from any of the above embodiments. The system includes: Vehicle behavior feature unit 601 is configured to extract the time period prior to the occurrence of the target road and bridge disaster event. and the time period after the occurrence The behavioral characteristic sequence of vehicles inside; The vehicle impact range unit 602 is configured to calculate the trajectory entropy value of the vehicle based on the vehicle's behavioral characteristic sequence and through a constructed trajectory entropy model, so as to identify the vehicles affected by the disaster event. The early warning range division unit 603 is configured as a set of vehicles affected by the disaster event. Based on real-time traffic flow data, the early warning range for target road and bridge disaster events is dynamically defined.

[0068] The early warning system for affected vehicles in road and bridge disasters based on trajectory entropy provided in this embodiment can implement the steps and processes of the early warning method for affected vehicles in road and bridge disasters based on trajectory entropy in any of the above embodiments, and achieve the same technical effect, which will not be described in detail here.

[0069] In the description of this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0070] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for early warning of affected vehicles in road and bridge disaster events based on trajectory entropy, characterized in that, include: Extract the time period before the target road and bridge disaster event. and the time period after the occurrence The behavioral characteristic sequence of vehicles inside; Based on the vehicle's behavioral characteristic sequence, the trajectory entropy value of the vehicle is calculated through the constructed trajectory entropy model in order to identify vehicles affected by disaster events; Based on the collection of vehicles affected by the disaster. Based on real-time traffic flow data, the early warning range for target road and bridge disaster events is dynamically defined.

2. The method according to claim 1, characterized in that, The time period prior to the target road and bridge disaster event was obtained. and the time period after the occurrence Trajectory data of vehicles within the target road and bridge ; trajectory data After data cleaning, the vehicle's trajectory points are matched to the actual road segments where the target road and bridge disaster occurred based on the map matching algorithm, resulting in a structured trajectory sequence of the vehicle. Extract the vehicle's behavioral feature sequence from the vehicle's structured trajectory sequence.

3. The method according to claim 2, characterized in that, With vehicles At any moment trajectory points Determine the vehicle as the center. At any moment The set of candidate road segments ; Calculation time vehicle In the candidate path set The Middle Candidate road sections Observation probability and calculation from time 1 At the time vehicle From candidate road sections Move to candidate road segment transition probability ; Based on observation probability and transition probability To determine the vehicles at the time of the target road and bridge disaster. The candidate road segment sequence in which it is located; Responding to vehicles when the target road or bridge collapse occurs If the candidate road segment sequence matches the road segment where the target road / bridge disaster occurred, then the vehicle will be... The corresponding trajectory points are vertically projected onto the geometric alignment of the matching road segment to obtain the vehicle. Structured trajectory sequences.

4. The method according to claim 2, characterized in that, Each trajectory point data item in the structured trajectory sequence must contain at least: timestamp, location information, and speed information; the vehicle's behavioral feature sequence is extracted from the vehicle's structured trajectory sequence, including: Based on the vehicles at the time of the target road and bridge disaster event The projected position and timestamp of each trajectory point are used to calculate the vehicle. The instantaneous speed of the vehicle at the time of the target road and bridge disaster was determined. velocity sequence; and through vehicles Calculate the speed difference and time difference between two consecutive trajectory points to determine the vehicle's position at the time of the target road / bridge disaster event. The acceleration sequence.

5. The method according to claim 2, characterized in that, Each trajectory point data item in the structured trajectory sequence also includes: a road segment identifier (link_id); Extracting vehicle behavior feature sequences from structured trajectory sequences also includes: The structured sequence of vehicles is divided into several consecutive road segments based on the changes in the road segment identifier link_id; Calculate the travel time ratio for each road segment, and arrange the travel time ratios of all road segments in chronological order to obtain the travel time ratio sequence for the vehicles.

6. The method according to claim 5, characterized in that, The structured sequence of vehicles is divided into several consecutive road segments based on the changes in the road segment identifier link_id, including: The structured sequence of vehicles is sorted in ascending order by timestamp, and the trajectory points of the structured sequence are traversed according to the time sequence. In response to the timestamp-ordered ascending sequence trajectory points If the corresponding road segment identifier (link_id) matches the vehicle's current road segment ID, then the trajectory point will be... Add to the end of the list of trajectory points for the vehicle's current route; among them, It is a positive integer, and ; Response to trajectory points If the corresponding road segment identifier link_id does not match the vehicle's current road segment ID, then the vehicle will be sorted by timestamp in ascending order. trajectory points The timestamp record is the end time of the vehicle's current road segment journey. Therefore, the vehicle's current road segment journey ends, and the time is recorded as a trajectory point. The corresponding road segment identifier link_id creates a new road segment trip.

7. The method according to claim 2, characterized in that, The trajectory entropy model is: In the formula, For vehicles The trajectory entropy value, For vehicles Driving status in trajectory data The probability of occurrence This is a set of vehicle driving states.

8. The method according to claim 1, characterized in that, Based on the time period before the target road and bridge disaster occurred The trajectory entropy values ​​of vehicles within the road are used to construct the baseline range of trajectory entropy for the target road and bridge. ;in, The period before the target road and bridge disaster occurred. The mean trajectory entropy of vehicles inside the vehicle, The period before the target road and bridge disaster occurred. Standard deviation of the trajectory entropy of vehicles inside the vehicle; In response to ,or, Then determine the time period after the target road and bridge disaster event. Vehicles inside For vehicles affected by the disaster; in, For entropy adjustment parameters, For the preset threshold, The time period following the target road and bridge disaster Vehicles inside The trajectory entropy value, The period before the target road and bridge disaster occurred. Inner vehicle The trajectory entropy value.

9. The method according to claim 1, characterized in that, In the road network topology, by tracing back upstream from the point of failure of the target road / bridge disaster event, the set of vehicles affected by the disaster event can be determined. The location of the furthest vehicle is the initial upstream boundary of the warning range. ; Based on the average deceleration and stagnation of vehicles after the target road and bridge disaster, a traffic wave velocity model is used to predict the congestion propagation speed of the congestion impact upstream after the disaster. ; According to the formula: Determine the upstream boundary of the warning area .

10. A warning system for affected vehicles in road and bridge disaster events based on trajectory entropy, characterized in that, The system dynamically delineates the early warning range of a target road and bridge disaster event using the trajectory entropy-based prediction method for affected vehicles in road and bridge disaster events as described in any one of claims 1-9. The system comprises: The vehicle behavior feature unit is configured to extract the time period preceding the target road / bridge disaster event. and the time period after the occurrence The behavioral characteristic sequence of vehicles inside; The vehicle impact range unit is configured to calculate the trajectory entropy value of the vehicle based on the vehicle's behavioral characteristic sequence through a constructed trajectory entropy model, in order to identify vehicles affected by the disaster event. The warning scope is divided into units, configured based on the set of vehicles affected by the disaster event. Based on real-time traffic flow data, the early warning range for target road and bridge disaster events is dynamically defined.

Images