A real-time anomaly detection method and system based on vehicle-road cloud cooperation

Abstract

The application discloses a real-time anomaly detection method and system based on vehicle-road cloud cooperation. By acquiring vehicle trajectory, vehicle running state, road topology and road running state time series data, determining the road section matching result and extracting the vehicle abnormal event, the candidate abnormal section is formed according to the vehicle passing order; the vehicle abnormal event is spatio-temporally matched with the road running state, the common deviation interval of multiple vehicles is extracted and the road section abnormal event is formed; the single vehicle abnormal deviation is decomposed by the common deviation interval, the road attribution deviation interval and the vehicle residual abnormality are obtained, and the real-time anomaly detection result is generated by continuously updating and re-matching, so that the multi-vehicle common abnormality is attributed according to the common action of the road, the vehicle residual abnormality of the single vehicle exceeding the common deviation range is reserved, and the vehicle-road abnormality relationship is updated with the continuous change of the road abnormality and the vehicle abnormality.

Description

Technical Field

[0001] This application relates to the field of vehicle-road-cloud collaborative data processing technology, and in particular to a real-time anomaly detection method and system based on vehicle-road-cloud collaboration. Background Technology

[0002] With the continuous deployment of intelligent connected vehicles, roadside sensing devices, and cloud-based traffic management platforms, information such as vehicle operating status, vehicle trajectory, road topology, and road operating conditions can be continuously acquired during the same road travel process. When a vehicle decelerates, brakes, lateral deviates, vibrates, or changes its trajectory, the anomaly may originate from the vehicle's own operating state or be caused by road factors such as construction, road flooding, obstacles, accidents blocking lanes, traffic congestion, or lane changes. To improve the reference value of anomaly detection results for road operation management and vehicle status judgment, it is necessary to further distinguish the relationship between vehicle anomalies and road anomalies and continuously track the changes of anomalies in time and road space.

[0003] Current methods for detecting and processing vehicle-road anomalies typically involve setting anomaly thresholds or normal baselines for both vehicle and road operating states. Individual vehicle anomalies are identified based on vehicle-side data such as speed, acceleration, braking status, and trajectory changes, while road-side data such as average speed, traffic flow, and road occupancy are used to identify road anomalies. In scenarios involving multiple vehicles, the number of anomalous vehicles, the frequency of anomalies, and changes in vehicle status are typically statistically analyzed within fixed time windows or fixed road areas. This data is then combined with road status information to determine if a common anomaly exists in a specific road area. For identified vehicle and road anomalies, the corresponding anomaly records are updated according to the current detection cycle or adjacent time windows.

[0004] The above-mentioned processing methods still have insufficient adaptability in scenarios where multiple vehicles travel sequentially along the same road. Different vehicles pass the same abnormal location at different times, and similar vehicle anomalies caused by the same road factor are often scattered across different times and locations. Furthermore, the amplitude of the anomalies affected by the same road factor is not entirely consistent across vehicles, and some vehicles may even have their own anomalies superimposed. When multiple vehicles experience similar anomalies sequentially on the same road segment, it is difficult to distinguish between common road anomalies and individual vehicle anomalies by combining vehicle travel processes and road operating conditions. It is easy to categorize anomalies of a vehicle that exceed the common range of change as road anomalies. As road and vehicle anomalies continue to develop, the original vehicle-road correspondence will change with road location and occurrence time, making it difficult to update the vehicle-road anomaly relationship in a timely manner as the anomaly objects continuously change. Summary of the Invention

[0005] In view of the above-mentioned actual situation, this application proposes a real-time anomaly detection method and system based on vehicle-road-cloud collaboration to solve the problems in existing vehicle-road anomaly detection, such as the difficulty in distinguishing common road anomalies from individual vehicle anomalies when multiple vehicles have similar anomalies in the same road segment, and the difficulty in updating the vehicle-road anomaly relationship in real time when the anomaly objects are constantly changing.

[0006] A real-time anomaly detection method based on vehicle-road-cloud collaboration includes:

[0007] S1. Obtain vehicle trajectory time-series data, vehicle operation status time-series data, road topology data, and road operation status time-series data; determine and update road segment matching results based on vehicle trajectory time-series data and road topology data.

[0008] S2. Based on the road segment matching results, extract vehicle abnormal events and corresponding abnormal deviation intervals from the vehicle operation status time series data, determine the vehicle passage order of each vehicle abnormal event, and merge vehicle abnormal events according to the vehicle passage order to form candidate abnormal segments.

[0009] S3. Perform spatiotemporal matching of vehicle anomaly events in the candidate anomaly segment and road operation status time series data, extract the common deviation interval of multiple vehicles from the matched vehicle anomaly events, and confirm the anomaly of the matched road operation status accordingly to form a road segment anomaly event.

[0010] S4. Decompose the abnormal deviation intervals of the corresponding vehicle abnormal events into the common deviation intervals of the multiple vehicles to obtain the road attribution deviation intervals and vehicle residual anomalies, and combine them with the road segment abnormal events to form a vehicle-road anomaly determination result.

[0011] S5. According to the time sequence of road segment abnormal events and vehicle residual abnormalities in the vehicle-road abnormality determination result, continuously update the road segment abnormal events and vehicle residual abnormalities respectively, rematch the updated road segment abnormal events and vehicle residual abnormalities, and generate real-time abnormality detection results.

[0012] The real-time anomaly detection method and system based on vehicle-road-cloud collaboration proposed in this application realizes hierarchical identification of common road anomalies and individual vehicle anomalies, enabling anomalies occurring together by multiple vehicles to be attributed to the joint effects of the road, while retaining residual anomalies of individual vehicles that exceed the common deviation range, and can update the vehicle-road anomaly relationship in real time as road anomalies and vehicle anomalies continue to change. Attached Figure Description

[0013] Figure 1 This is a flowchart of a real-time anomaly detection method based on vehicle-road-cloud collaboration.

[0014] Figure 2This is a schematic diagram illustrating the formation of candidate abnormal segments based on the merging of vehicle abnormal event sequences.

[0015] Figure 3 A schematic diagram illustrating the selection of abnormal vehicle events and alignment of abnormal deviation intervals in a matching event sequence;

[0016] Figure 4 A schematic diagram illustrating the road attribution bias interval and the formation of vehicle residual anomalies;

[0017] Figure 5 A diagram illustrating the rematching of updated road segment anomalies and vehicle residual anomalies;

[0018] Figure 6 This is a schematic diagram of a real-time anomaly detection system based on vehicle-road-cloud collaboration, provided as an embodiment of this application. Detailed Implementation

[0019] The following describes a specific implementation of a real-time anomaly detection method based on vehicle-road-cloud collaboration, with reference to the accompanying drawings. This embodiment addresses a scenario where vehicle state anomalies and road operation anomalies are intertwined during continuous vehicle travel along a road. It processes vehicle trajectory time-series data, vehicle operation state time-series data, road topology data, and road operation state time-series data under a unified time reference and a unified road spatial reference, sequentially generating road segment matching results, candidate anomaly segments, road segment anomaly events, vehicle-road anomaly determination results, and real-time anomaly detection results.

[0020] The data used in this embodiment is accessed through the vehicle-side data interface, the roadside perception interface, and the road infrastructure data interface. Vehicle identity, road markings, sampling time, and road location serve as the basic identifiers used for subsequent data attribution and object connection, maintaining consistent data meaning throughout each processing step.

[0021] The vehicle-road-cloud collaboration referred to in this application means that vehicle-side data on vehicle trajectory and vehicle operation status are generated, road-side data on road operation status are generated, and the cloud or edge computing side organizes vehicle-side data, road-side data and road topology data according to a unified time reference and a unified road spatial reference, and performs cross-vehicle anomaly association, road anomaly confirmation, vehicle-road anomaly attribution and real-time anomaly update around the same road traffic process.

[0022] The vehicle side generates vehicle trajectory time-series data and vehicle operating status time-series data. The vehicle trajectory time-series data is used to describe the continuous change of the vehicle's position over time, and the vehicle operating status time-series data is used to describe the continuous change of the vehicle's motion state or control state over time.

[0023] Roadside data generates time-series data on road operation status, describing the changes in road operation status over time within a specific road space. Road topology data describes road segments, road nodes, road traffic directions, segment lengths, and directed connections between different road segments within the road network.

[0024] Cloud or edge computing nodes undertake the corresponding computational tasks in road segment matching, vehicle anomaly event extraction, road operation anomaly handling, multi-vehicle common deviation handling, road attribution handling, and continuous updating of anomaly objects. In a deployment system, S1 to S5 are executed sequentially by the same computing node; in a distributed deployment mode, the processing result of the previous process, along with the corresponding object identifier, is passed to the next process, and the data formation order of S1 to S5 remains consistent.

[0025] In one implementation, the vehicle trajectory timing data is data output by a GNSS positioning device, the vehicle operation status timing data is data output by a vehicle CAN bus, an inertial measurement device, or a vehicle control system, the road operation status timing data is data output by a roadside radar, video traffic detection equipment, or traffic flow detection facility, and the road topology data is electronic road network data.

[0026] Each data source carries its corresponding sampling time or state time and is unified to the same time reference. Vehicle trajectory location, road state unit, and subsequent vehicle anomaly event road location and road anomaly range are unified to the same road spatial reference. Basic time synchronization, road coordinate transformation, data caching, and data transmission are performed according to the data interface specifications of the deployed system.

[0027] Vehicle trajectory time-series data records changes in vehicle spatial location in chronological order of sampling time, and includes at least the trajectory location, the corresponding sampling time, and the vehicle's identity. Multiple consecutive trajectory locations are arranged according to their sampling time to constitute the trajectory time series of the same vehicle.

[0028] Vehicle operating status time-series data refers to the motion or operating status data of the same vehicle formed at consecutive sampling times, including one or more of the following: vehicle speed, longitudinal acceleration, lateral acceleration, braking status, and steering changes. Different state indicators in the vehicle operating status correspond to corresponding anomaly types, and subsequent vehicle abnormal events are generated according to the corresponding state indicators and their abnormal deviation directions.

[0029] Road topology data describes the structure of the road network, including at least road segment identifiers, road segment lengths, road traffic directions, and directed connections between different road segments. After vehicle trajectories are mapped to the road network, the vehicle's spatial location on the road is described by matching road segments and their corresponding locations within those segments.

[0030] Road operation status time-series data records changes in road operation status according to continuous time intervals, including one or more of the following: average speed of road segment, traffic flow, road occupancy, queuing status, and lane traffic status. Each road operation status record corresponds to a specific road spatial range and time interval, and different road operation status indicators correspond to specific road operation benchmarks and anomaly detection conditions.

[0031] Vehicle identity is used to maintain object consistency for the same vehicle across vehicle trajectory time-series data, vehicle operating status time-series data, vehicle anomaly events, and vehicle residual anomalies. Road identifiers and road locations are used to maintain spatial consistency among vehicle trajectories, road topology, vehicle anomaly events, and road operating status.

[0032] The sampling time in vehicle trajectory time-series data, the temporal position in vehicle operation status time-series data, and the status time in road operation status time-series data are organized according to a unified time benchmark. The vehicle trajectory position, the road position corresponding to the vehicle abnormal event, and the road range corresponding to the road operation status are organized according to a unified road spatial benchmark. Based on this, vehicle abnormal objects and road abnormal objects have a matching basis in both road spatial and temporal dimensions.

[0033] The road segment matching result is the processing result formed after mapping vehicle trajectory time-series data to the road network, including the currently matched road segment, the location within the current road segment, and the road segment travel stage. The currently matched road segment is used to determine the road topology segment to which the vehicle currently belongs, the location within the current road segment is used to determine the vehicle's position along the corresponding road travel direction, and the road segment travel stage is used to describe the continuous travel process formed by the vehicle within the currently matched road segment.

[0034] Candidate anomaly segments are continuous road ranges formed by organizing vehicle anomaly events according to their traffic order and merging them based on the continuity of the road positions corresponding to adjacent vehicle anomaly events along the traffic direction. Candidate anomaly segments are associated with the vehicle anomaly events and their corresponding traffic orders, and are used to limit the scope of road analysis for subsequent road operation status.

[0035] Road segment anomalies are identified by spatiotemporally matching these abnormal road segments with vehicle anomalies in candidate anomaly segments, and then confirming the anomaly based on the shared deviation intervals of the matched vehicle anomalies. Each road segment anomaly retains its corresponding road range and duration, and is associated with the shared deviation intervals of the vehicles used for this anomaly confirmation.

[0036] The vehicle-road anomaly determination result is the processing result formed after decomposing the abnormal deviations of vehicle anomaly events corresponding to road segment anomaly events. For the corresponding vehicle anomaly event, its abnormal deviation interval is divided into a road attribution deviation interval and a vehicle residual anomaly according to the common deviation interval of multiple vehicles. The vehicle-road anomaly determination result saves the road segment anomaly event, the road attribution deviation interval, the vehicle residual anomaly, and the corresponding source.

[0037] Real-time anomaly detection results are generated by continuously updating road segment anomaly events and vehicle residual anomalies, and then re-matching the updated road segment anomaly events and vehicle residual anomalies according to road spatial and temporal conditions. These real-time anomaly detection results are used to store the re-matching results and the corresponding road anomaly and vehicle anomaly objects.

[0038] The specific formation processes of candidate road segments, candidate road segment sequences, road segment matching sequences, vehicle abnormal events, vehicle abnormal event sequences, road operation abnormal segments, matching event sequences, multi-vehicle common deviation intervals, road attribution deviation intervals, remaining deviation intervals, and vehicle residual anomalies will be elaborated in the corresponding sections later.

[0039] In this embodiment, the abnormal deviation interval, the multi-vehicle common deviation interval, and the road attribution deviation interval all use deviation amplitude as the description dimension; the occurrence period and the duration period use time as the description dimension; and the road location and the road range use road space as the description dimension.

[0040] Zero deviation is the deviation reference position corresponding to the vehicle's operating state and the vehicle's operating reference. Abnormal deviation direction is the direction of amplitude extension when the vehicle's operating state deviates abnormally from the vehicle's operating reference.

[0041] The abnormal deviation range is the range of deviation amplitudes formed by extending from the zero deviation starting point along the current abnormal deviation direction to the maximum abnormal deviation amplitude of the vehicle abnormal event. It is used to describe the amplitude boundary of the vehicle abnormal event in the corresponding abnormal direction.

[0042] The common deviation range of multiple vehicles is a continuous deviation amplitude range that is jointly covered by different vehicles that meet the preset number of vehicles after multiple abnormal events of the same type and in the same abnormal deviation direction have been processed by a unified deviation benchmark.

[0043] The road attribution deviation range is the deviation amplitude portion of the abnormal deviation range corresponding to the vehicle abnormal event that falls within the current road joint effect explanation range. Its cutoff position is constrained by both the multi-vehicle joint deviation range and the current vehicle's own abnormal deviation range.

[0044] The occurrence period describes the time range corresponding to abnormal vehicle events or residual abnormalities. The duration period describes the time range corresponding to abnormal segments of road operation or abnormal events on road sections.

[0045] Time overlap refers to the occurrence period and the corresponding duration period sharing a common time range. Road location falling within the road range means that the road location corresponding to the vehicle abnormality event or vehicle residual abnormality is within the corresponding road range, and is judged according to a unified road spatial benchmark combined with the corresponding road boundary tolerance.

[0046] The processing parameters corresponding to the confluence of time boundaries and the boundaries of road ranges will be explained further later.

[0047] In this embodiment, as Figure 1 As shown, a real-time anomaly detection method based on vehicle-road-cloud collaboration includes S1 to S5. S1: Acquire vehicle trajectory time-series data, vehicle operating status time-series data, road topology data, and road operating status time-series data; determine and update road segment matching results based on vehicle trajectory time-series data and road topology data.

[0048] S2. Based on the road segment matching results, extract abnormal vehicle events and corresponding abnormal deviation intervals from the vehicle operation status time series data, determine the vehicle passage order of each abnormal vehicle event, and merge the abnormal vehicle events according to the vehicle passage order to form candidate abnormal segments.

[0049] S3. Perform spatiotemporal matching on the time series data of vehicle anomaly events and road operation status in the candidate anomaly segment, extract the common deviation interval of multiple vehicles from the matched vehicle anomaly events, and confirm the anomaly of the matched road operation status accordingly to form road segment anomaly events.

[0050] S4. Decompose the abnormal deviation intervals of corresponding vehicle abnormal events by using the common deviation interval of multiple vehicles to obtain the road attribution deviation interval and vehicle residual abnormality, and combine them with road segment abnormal events to form the vehicle-road abnormality judgment result.

[0051] S5. According to the time sequence of road segment abnormal events and vehicle residual abnormalities in the vehicle-road abnormality judgment results, continuously update the road segment abnormal events and vehicle residual abnormalities respectively, rematch the updated road segment abnormal events and vehicle residual abnormalities, and generate real-time abnormality detection results.

[0052] S1 is used to establish the spatial correspondence between vehicle trajectories and road networks. After the vehicle trajectory time series data and road topology data are matched, the vehicle's operating status has a corresponding matching road segment and its position within the road segment, and subsequent abnormal vehicle events are used to obtain the road location.

[0053] S2 is used to organize deviations in the operating status of a single vehicle into vehicle anomaly events with a road traffic sequence. After a vehicle anomaly event is formed, the vehicle traffic sequence is determined based on the event's temporal location and road location, and then candidate anomaly segments are formed based on the continuity of the vehicle traffic sequence and road location.

[0054] S3 is used to establish the correspondence between abnormal road operation status and common abnormal responses of multiple vehicles. Candidate abnormal segments limit the analysis scope of road operation status. After the road operation status forms abnormal road operation segments, it is matched with abnormal vehicle events according to road location and time range. Matching abnormal vehicle events forms a common deviation interval for multiple vehicles, which is used to confirm the corresponding abnormal road operation segments and form road segment abnormal events.

[0055] S4 is used to decompose the deviations of the vehicle anomaly events involved in the current road anomaly confirmation. The common deviation interval of multiple vehicles defines the range of deviation amplitudes corresponding to the combined effect of the road. The road common explanation portion of the corresponding vehicle anomaly deviation forms the road attribution deviation interval, and the remaining deviations form the vehicle residual anomaly.

[0056] S5 is used for continuous processing of road segment anomalies and vehicle residual anomalies. The two types of anomalies are updated according to their respective continuity conditions. The updated road segment anomalies and vehicle residual anomalies are then re-matched according to road location, road range, time of occurrence, and duration to form real-time anomaly detection results.

[0057] In S1, vehicle trajectory time-series data and road topology data form road segment matching results. Subsequent vehicle anomaly handling uses the current matched road segment, the location within the current road segment, and the road segment passage stage from the road segment matching results.

[0058] In S2, vehicle operation status time-series data is combined with road segment matching results to form vehicle abnormal events and abnormal deviation intervals. Vehicle abnormal events further carry vehicle identity, time-series location, road location, abnormality type, abnormal deviation direction, and abnormal deviation interval, and based on this, the vehicle passage order is determined and candidate abnormal segments are formed.

[0059] In S3, road operation status time-series data forms road operation anomaly segments within the road analysis range corresponding to candidate anomaly sections. These road operation anomaly segments participate in the spatiotemporal matching of vehicle anomaly events based on road range and duration. Matched vehicle anomaly events participate in multi-vehicle common deviation processing based on anomaly type, anomaly deviation direction, and anomaly deviation interval. The resulting multi-vehicle common deviation interval is then associated with the corresponding set of vehicle anomaly events.

[0060] In S4, the road attribution bias interval and vehicle residual anomalies are included in the vehicle-road anomaly determination results. Vehicle residual anomalies retain the corresponding vehicle identity, road location, and time of occurrence for subsequent continuous processing of vehicle residual anomalies.

[0061] In S5, road segment anomalies and vehicle residual anomalies are continuously updated. Updated road segment anomalies retain the road range and duration, while updated vehicle residual anomalies retain the road location and occurrence time. The corresponding spatial and temporal attributes are then re-matched.

[0062] Road topology data maintains a consistent road structure meaning in road segment matching, road location sorting, and road spatial continuity judgment; vehicle identity maintains a consistent object identity in vehicle operating status, vehicle abnormal events, and vehicle residual anomaly handling; abnormal deviation intervals and multi-vehicle common deviation intervals maintain a consistent deviation amplitude meaning in common deviation handling and road attribution handling.

[0063] When vehicle trajectories are mapped to the road network, road intersections, adjacent parallel roads, and changes in the position of consecutive road segments can result in multiple candidate road segments corresponding to the same trajectory location. For continuous vehicle trajectories, this embodiment forms a candidate road segment sequence according to the trajectory time sequence. It determines the road segment matching sequence by combining the changes in candidate road segments corresponding to adjacent trajectory locations, changes in positions within road segments, and directed connectivity relationships in the road topology. Based on the road segment matching sequence, it continuously updates the vehicle's currently matched road segment, its position within the current road segment, and the road segment travel stage.

[0064] Specifically, the formation of road segment matching results includes the following processing:

[0065] S11. Map the continuous trajectory positions in the vehicle trajectory time series data to the road network described by the road topology data, determine the candidate road segments and the corresponding positions within each trajectory position, and form a candidate road segment sequence consisting of the candidate road segments and the corresponding positions within each road segment according to the trajectory time series.

[0066] S12. Based on the changes in candidate road segments and the changes in positions within road segments of adjacent trajectory positions in the candidate road segment sequence, and combined with the directed connectivity in the road topology data, determine the matching road segments and the corresponding positions within each candidate road segment to form a road segment matching sequence.

[0067] S13. According to the temporal changes of the matched road segment and the corresponding position within the road segment in the road segment matching sequence, update the vehicle's current matched road segment and the current position within the current road segment, and determine the road segment passage stage by the continuous changes in the position within the road segment, and form the road segment matching result with the current matched road segment, the current position within the current road segment and the road segment passage stage.

[0068] Candidate road segments are road topology segments selected from the road network described by road topology data based on the current trajectory position in the vehicle trajectory time series data. One trajectory position corresponds to one or more candidate road segments, and each candidate road segment is associated with the position of that trajectory position within the corresponding road segment.

[0069] The location within a road segment is represented by the directed road coordinates of the candidate road segment itself. The starting point of the candidate road segment along the road's travel direction is used as the local location starting point. The cumulative road length along the road centerline is calculated, and the cumulative length corresponding to the projected position after mapping the trajectory position to the road centerline is taken as the location within the road segment. When the road centerline has a curved geometry, the location within the road segment is obtained according to the arc length of the road centerline.

[0070] For the current trajectory position, search the road topology segments in the adjacent area of ​​the corresponding road space to obtain the candidate road set corresponding to the current trajectory position; for each road topology segment in the candidate road set, determine the projection position and projection distance from the current trajectory position to the center line of the corresponding road, and convert the projection position into the position within the corresponding road segment.

[0071] The spatial proximity range of a road is set by combining vehicle positioning accuracy, road geometric width, and road network density. When multiple road topology segments that meet the spatial mapping conditions exist in a road intersection area, adjacent parallel road area, or road segment switching area, the corresponding road topology segments are retained together in the candidate road segment set at the current trajectory position.

[0072] The vehicle's direction of motion is obtained based on the spatial changes between the positions of continuous valid trajectories. For each candidate road segment, the difference between the vehicle's direction of motion and the road's direction of travel in that candidate road segment is recorded, so that the candidate road segment simultaneously has three parameters related to the selection of subsequent road segments: projected distance, position within the road segment, and direction difference.

[0073] The candidate road segment set and the corresponding position within each road segment are arranged according to the sampling time in the vehicle trajectory time series data to form a candidate road segment sequence. Each time series position in the candidate road segment sequence corresponds to a trajectory position, and one or more candidate road segments corresponding to that trajectory position and the corresponding position within each road segment are saved.

[0074] For adjacent trajectory positions in the candidate road segment sequence, the candidate road segment set, intra-segment position and sampling time corresponding to the previous trajectory position and the next trajectory position are obtained respectively, and the road traffic continuity relationship between candidate road segments is processed.

[0075] When the preceding and following candidate road segments belong to the same road topology, the continuous traffic relationship of vehicles in the current road segment is determined based on the changes in position along the road travel direction within the two road segments. When the preceding and following candidate road segments belong to different road topologies, the directed connection chain from the preceding candidate road segment to the following candidate road segment is obtained based on the directed connection relationship in the road topology data, and the road path distance formed along the directed connection chain between the two trajectory positions is calculated.

[0076] When the vehicle trajectory sampling interval covers multiple road topology segments, the directed connection chain includes the intermediate road segments located between consecutive candidate road segments. The sampling time interval corresponding to adjacent trajectory positions, trajectory spatial displacement, intra-segment position changes, and the road path distance corresponding to the directed connection chain jointly participate in continuous road matching.

[0077] For candidate road segment combinations corresponding to adjacent trajectory positions, those combinations that satisfy the conditions of directional road connection, intra-segment position change, and road path distance are retained. When a trajectory position corresponds to multiple candidate road segments that satisfy the above conditions, candidate disambiguation is performed in a predetermined order.

[0078] In one implementation, the candidate road segments are sorted according to the projected distance from the current trajectory position to the center line of each candidate road, and the candidate road segments with smaller projected distances are given priority. When the projected distances of multiple candidate road segments fall within the same preset distance tolerance range, the directional difference between the vehicle's direction of movement and the road traffic direction of each candidate road segment is compared, and the candidate road segments with smaller directional differences are given priority. When both the projected distance and the directional difference fall within the same corresponding judgment range, a unique matching road segment is obtained according to the fixed sorting of the road topology road segment identifiers.

[0079] The matching road segments and corresponding locations within each trajectory location are arranged according to the trajectory time sequence to form a road segment matching sequence.

[0080] The currently matched road segment represents the road topology segment where the vehicle is located at the current processing time, and the position within the current road segment represents the vehicle's position along the current matching road segment's travel direction. For each new valid trajectory position obtained, the currently matched road segment and the position within the current road segment are updated according to the matching result of that trajectory position in the road segment matching sequence.

[0081] A road segment travel phase refers to the continuous travel time process formed by a vehicle within the same currently matched road segment. When a vehicle is continuously in the same matched road segment and its position within the corresponding road segment changes continuously along the road travel direction, the corresponding trajectory time period is included in the same road segment travel phase. When a vehicle enters the next matched road segment through a directed connection, the road segment travel phase corresponding to the previous road segment ends at the road segment switching position, and a new road segment travel phase is established for the next matched road segment.

[0082] When the trajectory sampling time interval reaches the preset stage separation time, or when the corresponding road path change between consecutive trajectory positions reaches the preset stage separation condition, the current valid trajectory position is used as the stage boundary, and subsequent valid trajectory positions enter a new road segment passage stage.

[0083] Each road segment's passage phase records at least the corresponding matched road segment, the phase start time, the phase end time, and the position within consecutive road segments within the phase. The current matched road segment, the position within the current segment, and the current passage phase together form the road segment matching result.

[0084] Vehicle operation status time-series data and vehicle trajectory time-series data adopt a unified vehicle identity and a unified time reference. For a state sampling moment in vehicle operation status, the corresponding matching road segment is determined according to the road segment traffic phase to which that state sampling moment belongs.

[0085] When the state sampling time is located between two adjacent trajectory sampling times and the two trajectory samples correspond to the same matching road segment, the position within the corresponding road segment at the state sampling time is determined based on the changes in the positions of the two trajectory sampling times and the corresponding road segments. When the two trajectory samples cross one or more road topology segments via a directed connection chain, the road positions corresponding to the two trajectory samples are converted to the same directed path cumulative position, and the corresponding road position is determined according to the time position of the state sampling time between the two trajectory sampling times.

[0086] Through the above-mentioned time and road space alignment, the state sampling in the vehicle operation status time sequence is associated with the corresponding road segment traffic stage and road location.

[0087] After obtaining the road segment matching results, vehicle anomalies are assigned corresponding road locations and temporal locations. The time of anomaly occurrence describes the event's position within the vehicle's operational sequence, while the road location describes the event's position within the road network. The vehicle passage order organizes different vehicles' corresponding anomalies according to the same directed road passage process. Based on this processing relationship, vehicle anomalies are first generated during the road segment passage phase. Then, the vehicle passage order is determined based on the event's road location and temporal location, and consecutive vehicle anomalies at road locations are merged along the vehicle passage order.

[0088] Specifically, the formation of candidate anomalous segments includes the following processing:

[0089] S21. Determine the road segment passage stage corresponding to the vehicle based on the road segment matching results, determine the vehicle operation benchmark from the vehicle operation status time series data, use the vehicle operation benchmark as a reference to perform deviation detection on the vehicle operation status within the road segment passage stage, obtain the abnormal deviation interval, and generate the vehicle abnormal event corresponding to the abnormal deviation interval.

[0090] S22. Determine the corresponding road location based on the temporal position of each vehicle abnormal event in the vehicle operation status time series data and the road segment matching result. Determine the positional order of each road location along the road traffic direction based on the road topology data. Combine the temporal position of each vehicle abnormal event to determine the vehicle traffic order. Form a vehicle abnormal event sequence according to the vehicle traffic order.

[0091] S23. Merge adjacent vehicle abnormal events along the road location in the direction of traffic along the sequence of vehicle abnormal events, determine the continuous road range covered by the merged vehicle abnormal events, form candidate abnormal sections with the continuous road range, and retain the vehicle traffic order of the vehicle abnormal events that constitute the candidate abnormal sections.

[0092] The abnormal deviation range is the range of deviation amplitude formed by the vehicle's operating state relative to the vehicle's operating reference along the abnormal deviation direction.

[0093] The vehicle operating baseline is determined based on the vehicle's identity, vehicle operating status indicators, and the current road traffic flow, serving as a normal reference for detecting deviations in vehicle operating status. Corresponding vehicle operating baselines and anomaly detection conditions are established for vehicle operating status indicators such as vehicle speed, longitudinal acceleration, and lateral acceleration.

[0094] In one implementation, a sample of the vehicle's most recent period of normal operation is stored for the current vehicle, and a normal vehicle operation window is formed according to vehicle operation status indicators. The median value of the corresponding status indicator in the normal vehicle operation window is used to form the vehicle operation benchmark, and anomaly trigger thresholds and anomaly recovery thresholds are set according to the fluctuation range of the corresponding status indicator in the normal vehicle operation window.

[0095] When a vehicle enters a new road segment, once the number of normal vehicle status samples in the current segment reaches the preset baseline sample number, the vehicle operation baseline is updated based on the normal samples in the current segment. When the current segment is in the process of accumulating normal samples, the vehicle operation baseline formed by the most recent normal vehicle operation window before the vehicle entered the current road segment is adopted.

[0096] When a vehicle enters an abnormal operating state, the vehicle operating baseline corresponding to the current abnormal type is maintained, ensuring that each state sample during the abnormal process deviates from the normal reference calculation formed before the abnormality occurred. Once the vehicle state returns to the normal range and continuously reaches the preset number of recovery samples or the recovery duration, subsequent normal state samples are used to update the vehicle operating baseline.

[0097] For vehicle operation status sampling during the road segment's traffic phase, the vehicle operation status values ​​are compared with the corresponding vehicle operation benchmarks to obtain the deviation amplitude and abnormal deviation direction. When the deviation amplitude reaches the abnormal trigger condition corresponding to the current abnormal type, and continuously reaches the preset abnormal holding sampling number or abnormal holding duration, the starting sampling time that reaches the corresponding condition is determined as the starting point of the current vehicle abnormal event.

[0098] When vehicle operating states of the same anomaly type and in the same direction of deviation continuously meet the anomaly maintenance conditions, the related state samples are grouped into the same vehicle anomaly event. When the vehicle operating states continuously meet the anomaly recovery conditions to reach the preset recovery sampling quantity or recovery duration, the end time of the current vehicle anomaly event is determined. When the direction of deviation changes and the changed deviation direction meets the corresponding anomaly triggering condition, the two vehicle anomaly events are divided based on the position of the direction change.

[0099] Vehicle anomaly events should at least save the vehicle identity, anomaly type, direction of anomaly deviation, time period of occurrence, representative time sequence location, representative road location, and anomaly deviation range.

[0100] The vehicle operating status sampling time with the largest abnormal deviation amplitude within the period of the vehicle abnormal event is taken as the representative time sequence position. When multiple sampling times correspond to the same maximum abnormal deviation amplitude, the representative time sequence positions are obtained in the order of the sampling times from front to back.

[0101] The representative road location is obtained from the road segment matching results based on the representative temporal location. When the representative temporal location is consistent with the trajectory sampling time, the matching road segment and the location within the road segment corresponding to the trajectory sampling are used; when the representative temporal location is between adjacent trajectory sampling times, time alignment is performed based on the matching records of adjacent valid road segments.

[0102] When adjacent road segment matching records correspond to the same matching road segment, the position within the corresponding road segment is determined based on the time position of the representative temporal position between the sampling times of adjacent trajectories. When adjacent road segment matching records pass through a directed connection chain, the road positions of the two matching records are converted to the same directed path cumulative position, and then the corresponding directed path cumulative position is obtained based on the time position of the representative temporal position between the sampling times of adjacent trajectories, and the representative road position is obtained accordingly.

[0103] The abnormal deviation interval extends from the zero deviation starting point along the current abnormal deviation direction to the maximum abnormal deviation amplitude during the current vehicle abnormal event period. The actual vehicle state sampling during the vehicle abnormal event continues to maintain the original time sequence record, and the abnormal deviation interval serves as the amplitude boundary used for subsequent multi-vehicle common deviation processing.

[0104] When the same vehicle generates multiple time-separated abnormal events during continuous road traffic, each abnormal event saves its representative time sequence location, representative road location, occurrence time period, and abnormal deviation range, and is then processed in the vehicle traffic sequence.

[0105] When different vehicles' abnormal events are located at different road positions during the same directional road passage process, the road positions of each abnormal event are assigned to the corresponding directional passage path, and the vehicle passage order is determined by combining the road position sequence and the event time sequence.

[0106] Furthermore, the formation of a vehicle anomaly event sequence includes the following processing:

[0107] S221. Based on the temporal position of each vehicle abnormal event in the vehicle operation status time series data, determine the corresponding road position from the road segment matching results, classify the road position into the corresponding directed travel path according to the directed connection relationship in the road topology data, and determine the position order of each road position along the directed travel path.

[0108] S222. Determine the vehicle passage order corresponding to each abnormal vehicle event based on the positional order and the temporal position of each abnormal vehicle event.

[0109] S223. Arrange the abnormal vehicle events according to the order of vehicle passage to obtain the sequence of abnormal vehicle events.

[0110] A directed traffic path consists of one or more road topology segments that are continuously connected in the direction of traffic from the road topology data. It is used to place abnormal vehicle events that participate in the same vehicle traffic order calculation under a unified road direction and a unified road distance coordinate system.

[0111] The corresponding representative road location is obtained based on the representative temporal location of the vehicle anomaly event. This, combined with the road segment matching sequence before and after the representative temporal location, determines the actual directed road segment sequence traversed by the vehicle. Vehicle anomalies belonging to the same directed connection chain and with the same road travel direction are grouped into the same directed travel path processing group.

[0112] When a road intersection corresponds to multiple subsequent road segments, the subsequent road segment that the vehicle actually enters is determined based on the road segment matching sequence already formed by each vehicle. Abnormal events of vehicles traveling along the same subsequent directed connection chain are grouped into the same directed travel path processing group, while other subsequent directed connection chains form their respective processing groups.

[0113] Establish a unified cumulative position for a directed travel path. Select the fixed path starting point in the directed travel path as the starting point of the cumulative position, and accumulate the length of each road topology segment along the road travel direction according to the road topology connection order.

[0114] For a road location located on a segment of a directed travel path, its cumulative position on the directed path is obtained by adding the sum of the lengths of the complete road segments traversed by the directed travel path before that segment to the position within the corresponding segment of the current road location.

[0115] After accumulating the aforementioned positions, vehicle anomalies located on different road topology segments are incorporated into a unified road distance coordinate system. Road positions with smaller cumulative positions along the travel direction correspond to the beginning of the path, while those with larger cumulative positions correspond to the end of the path. The representative road positions for each vehicle anomaly are ordered according to the cumulative positions along the directed path. Events participating in the same vehicle passage order calculation share the same starting point of the cumulative position along the directed path, the same road travel direction, and the same road length reference. Subsequent calculations of equivalent passage times will be based on this shared information. , and All positions are represented by the cumulative position of the directed path.

[0116] Vehicle passage order represents the chronological order in which various vehicle anomalies occur, based on the sequence of the corresponding vehicles traveling along the same directed passage path. Since the representative road locations of each vehicle anomaly are distributed across different points along the directed passage path, the representative temporal locations of each event are converted to a unified reference road location, and the vehicle passage order is determined according to the conversion results.

[0117] For the set of abnormal vehicle events participating in the sorting of the current directed traffic path, a unified reference road position is determined on the directed traffic path, and its cumulative position on the directed path is denoted as... The reference road location is located within the coverage area of ​​the road location of the abnormal event of the currently participating vehicle or its corresponding adjacent road location, and remains consistent in the current vehicle passage order calculation.

[0118] For vehicle abnormal event i, its representative time sequence position is denoted as: The cumulative position of the directed path corresponding to the road location is denoted as Obtain a valid trajectory sampling point from the time series of vehicles adjacent to the vehicle event, moving along the same road direction. The corresponding sampling time is recorded as... The cumulative position of the corresponding directed path is denoted as .

[0119] The cumulative displacement of the directed path between the selected nearest valid trajectory sampling point and the road location represented by the vehicle abnormal event is greater than the preset minimum valid road displacement threshold, and the two correspond to the same road travel direction.

[0120] Based on the cumulative displacement and time changes of roads during the movement of vehicles in the vicinity of the abnormal event, the temporal location of the abnormal vehicle event is converted to a unified reference road location, and the corresponding equivalent passage time is: ,in This indicates that the abnormal vehicle event i has been converted to the reference road location. The equivalent passage time afterwards; This indicates the temporal change between the location of the vehicle anomaly event and the sampling time of the nearest valid trajectory; It represents the cumulative road displacement formed along the same directional travel path within the corresponding time range.

[0121] In one implementation, neighboring trajectory sampling points are obtained from the trajectory sampling records prior to the time sequence location represented by the vehicle anomaly event, arranged chronologically from nearest to farthest, until the cumulative displacement of the directed path between the sampling point and the representative road location reaches a preset minimum effective road displacement threshold. This trajectory sampling point is then used as the reference point. and The corresponding valid sampling points.

[0122] Reference road location A fixed cumulative position is used within the road location coverage area of ​​the currently participating vehicle anomaly events. In one implementation, the midpoint between the minimum and maximum cumulative positions corresponding to the participating vehicle anomaly events is used as the reference road position. When the distribution range of the currently participating events in the road space exceeds a preset reference conversion range, corresponding event processing groups are divided according to the adjacent road positions on the directed path, and a reference road position is determined for each processing group.

[0123] Equivalent travel time As a sorting time quantity formed after different vehicle events have been converted using a unified road reference location, the original representative temporal position of abnormal vehicle events Continue to save in the event log.

[0124] For the set of vehicle abnormal events For any abnormal vehicle event h, its corresponding equivalent passage time is determined using the same conversion method described above, and denoted as h. Let the set of abnormal vehicle events participating in the sorting under the same directed travel path be . The initial sorting position of vehicle abnormal event i, obtained based on the equivalent passage time, is: ,in This represents the initial sorting position of vehicle abnormal event i based on the equivalent passage time. This indicates the number of elements in the set, where each element corresponds to an equivalent passage time earlier than [previous time]. Abnormal vehicle events.

[0125] Vehicle anomalies occurring at equivalent passage times in different system time-resolved locations are categorized according to... The events are arranged chronologically. When the equivalent passage times corresponding to multiple vehicle anomalies are at the same system time resolution position, they are arranged according to their representative time sequence position; when the representative time sequence positions are at the same time resolution position, they are arranged according to the fixed order of the cumulative positions of the corresponding directed paths; when the aforementioned sorting parameters are the same, the arrangement position is determined according to the fixed order of the vehicle anomaly event identifiers.

[0126] The abnormal vehicle events are arranged according to the established vehicle passage order to obtain a vehicle abnormal event sequence. The vehicle abnormal event sequence saves the arrangement position of the vehicle events and continues to associate each abnormal vehicle event with the vehicle identity, representative time sequence position, representative road position, occurrence time, abnormality type, abnormality deviation direction, and abnormality deviation range.

[0127] After the vehicle anomaly event sequence is formed, adjacent vehicle anomaly events are obtained sequentially according to their event order. The vehicle anomaly event sequence is used to determine the currently participating adjacent events for comparison, and the continuity of road positions along the road travel direction is used to determine the road merging relationship between adjacent events.

[0128] For a pair of adjacent vehicle anomaly events, obtain the representative road locations and corresponding directed path cumulative locations for both events. When the two road locations are on the same directed travel path, determine the road distance between them along the road travel direction according to the directed path cumulative location; when the two road locations are on road segments that are continuously connected according to the road topology, calculate the corresponding road distance according to the connected directed road path.

[0129] When the road locations corresponding to two adjacent vehicle anomaly events are continuously connected along the road travel direction, and the road distance between them falls within the preset continuous road interval range, the two adjacent vehicle anomaly events are grouped into the same merged event group.

[0130] After a merged event group is formed, the next vehicle anomaly event in the sequence of vehicle anomalies is obtained. A new pair of adjacent events is formed by the vehicle anomaly event at the end of the current merged event group along the road travel direction and the next vehicle anomaly event, and the judgment continues according to the same road continuity condition. Vehicle anomalies that continuously meet the merging condition are added to the current merged event group in sequence.

[0131] Road fork locations are merged according to the aforementioned directed traffic path processing groups. For vehicle anomalies along the same subsequent directed traffic path, road continuity judgment continues, and different subsequent directed traffic paths form corresponding merged event groups.

[0132] In the merged event group, the road position represented by the vehicle anomaly event at the foremost position along the road travel direction is taken as the starting position of the continuous road range, and the road position represented by the vehicle anomaly event at the last position along the road travel direction is taken as the ending position of the continuous road range. The road space covered by the corresponding directional travel path between the two positions forms a continuous road range.

[0133] Candidate anomaly segments are formed by continuous road areas, and the vehicle anomaly events constituting the candidate anomaly segments and their vehicle passage order are associated with and saved to the candidate anomaly segments. The original vehicle identity, representative road location, occurrence time, anomaly type, anomaly deviation direction, and anomaly deviation range of each vehicle anomaly event are retained for subsequent road operation anomaly handling and vehicle event matching.

[0134] Figure 2In the sequence of vehicle anomalies, adjacent vehicle anomalies correspond to road locations within a directed traffic path. The road locations corresponding to adjacent vehicle anomalies are then assessed for continuity along the traffic direction. If this condition is met, the adjacent vehicle anomalies proceed to the "merging adjacent vehicle anomalies" process. The corresponding road locations are then defined as a continuous road range along the directed traffic path, forming a candidate anomaly segment. Another assessment branch establishes segment boundaries at the adjacent event locations for the current merged event group, and subsequent vehicle anomalies proceed to the next merging process according to the vehicle anomaly event sequence.

[0135] Figure 2 The sequence of abnormal vehicle events in the figure adopts the event sorting results formed above, and the directed traffic paths adopt the road spatial benchmark formed above. The continuous road range in the figure corresponds to the road range covered by the currently merged abnormal vehicle events, and the candidate abnormal segment is associated with the abnormal vehicle events and vehicle traffic order that constitute the segment.

[0136] Once candidate anomaly segments are formed, a set of abnormal vehicle events organized according to vehicle traffic order has accumulated within the corresponding continuous road area. Road operation status time-series data records the changes in the road's spatial operation status over time. During processing, abnormal road operation segments are first formed from the road operation status. Then, abnormal vehicle events are matched with specific abnormal road operation segments based on road location and time. Subsequently, common deviation intervals between multiple vehicles are extracted from the matched vehicle events, and these common deviation intervals are used to confirm the anomalies in the corresponding abnormal road operation segments.

[0137] Specifically, the formation of abnormal events on road sections includes the following handling:

[0138] S31. Define the road analysis range of the road operation status time series data based on the candidate abnormal sections, determine the road operation benchmark from the road operation status time series data, use the road operation benchmark as a reference to perform deviation detection on the road operation status within the road analysis range, obtain the abnormal road operation segments, and determine the road range and duration of the abnormal road operation segments.

[0139] S32. Obtain the corresponding road location and occurrence time from the vehicle abnormal events in the candidate abnormal section, match the abnormal road operation segments with the vehicle abnormal events according to the road location and occurrence time, filter the vehicle abnormal events whose road location falls within the road range and whose occurrence time coincides with the duration, and form a matching event sequence according to the vehicle passage order.

[0140] S33. Obtain the anomaly type and anomaly deviation direction from the vehicle anomaly events in the matching event sequence. Select vehicle anomaly events with the same anomaly type and the same anomaly deviation direction. Perform common deviation processing on the selected vehicle anomaly events to determine the common deviation interval of multiple vehicles. Use the common deviation interval of multiple vehicles to confirm the abnormal road operation segment and form the road segment anomaly event. The road segment anomaly event retains the road range and duration of the abnormal road operation segment.

[0141] The scope of road analysis is determined based on the continuous road range of the candidate anomaly segment. The candidate anomaly segment is associated with at least the continuous road range and the vehicle anomaly events constituting the segment. Road operation status records covering the corresponding road space are obtained from the road operation status time series data based on the continuous road range.

[0142] When storing road operating status according to road state units, road state units corresponding to the spatial occurrence of candidate anomaly segments within a continuous road range are included in the current road analysis scope based on road topology identifiers and road locations. Road state units are arranged along the road's travel direction, each with a corresponding road start and end position, and their corresponding road operating status is saved according to the current state time.

[0143] When the continuous road range of a candidate anomaly segment spans multiple road state units, the road state units are obtained in a continuous arrangement according to the directed road connection order. When the boundary of a candidate anomaly segment falls within a road state unit, the road state unit containing that boundary location is included in the road analysis range, and the corresponding spatial boundary processing is performed according to a unified road spatial benchmark.

[0144] Road operation benchmarks are established for road state units and road operation state indicators respectively. For average speed of road segments, traffic flow, road occupancy, queuing status or lane traffic status, corresponding normal operation references and abnormal judgment conditions are stored respectively.

[0145] In one implementation, a normal time window is formed from the normal road operation state samples before the current road state unit anomaly occurs. The median value of the corresponding road operation state index in the normal time window is used as the road operation benchmark, and anomaly triggering conditions and anomaly recovery conditions are set according to the fluctuation range of the corresponding state index within the window.

[0146] When the current road state unit enters a road anomaly process, the road operation benchmark established before the anomaly began is maintained, and the corresponding road state deviation during the anomaly process is continuously calculated based on this normal reference. After the road operation state returns to the normal range and continuously reaches the preset number of recovery samples or the recovery duration, the road operation benchmark continues to be updated based on subsequent normal road state samples.

[0147] Deviation detection is performed on each road state unit within the road analysis scope according to road operation status indicators. When the current road operation status reaches the abnormal trigger condition relative to the corresponding road operation benchmark and continues to reach the preset abnormal holding sampling number or abnormal holding duration, a corresponding road state abnormality record is generated. The road state abnormality record is associated with at least the road state unit, the road operation status indicator type, the abnormality start time, and the abnormality end time.

[0148] For the same road operation status index, road status anomaly records that are spatially adjacent and whose abnormal times are consecutive are spatiotemporally linked. When adjacent road status units along the road direction form consecutive abnormal states within overlapping or consecutive abnormal time ranges, the corresponding abnormal records are included in the same road operation anomaly segment; when the same road status unit continuously meets the abnormal conditions in a continuous state cycle, the corresponding abnormal time is continuously merged into the current road operation anomaly segment.

[0149] The abnormal road operation segment consists of the aforementioned spatiotemporally continuous records of abnormal road conditions. The road extent is determined by the continuous road space covered by the road state unit constituting the current abnormal road operation segment along the road travel direction; the duration is determined by the time range continuously covered by the corresponding abnormal road condition records; and the road operation status index type follows the road operation status index constituting the current abnormal road operation segment.

[0150] Within the same road analysis scope, road operation anomaly segments are generated for different road operation status indicators. Road operation anomaly segments generated by average road speed, road operation anomaly segments generated by road occupancy, and road operation anomaly segments generated by other indicators are each stored with their respective indicator types and corresponding numerical scales.

[0151] Once an abnormal road operation segment is formed, the source road condition anomaly record or corresponding source identifier is saved, and the road range and duration are used for subsequent vehicle event spatiotemporal matching. Road operation status index values ​​continue to be saved according to the original index dimensions, and subsequent common deviations on the vehicle side adopt the unified scale of the vehicle abnormal deviation itself.

[0152] For the current abnormal road segment, a set of vehicle anomaly events is obtained from the candidate anomaly segments that form the current road analysis scope. The vehicle anomaly events continue to use the previously determined representative road location, time period, and vehicle passage order.

[0153] The corresponding road location here uses the representative road location of the vehicle anomaly event. The representative road location of the vehicle anomaly event is compared with the road range of the current abnormal road segment, and the time period of the vehicle anomaly event is compared with the duration of the current abnormal road segment.

[0154] A vehicle anomaly event is identified as a matching vehicle anomaly event for the current road operation anomaly segment when the road location falls within the road range of the current road operation anomaly segment, and when the time of the vehicle anomaly event and the duration of the current road operation anomaly segment have a common time range.

[0155] The common time range of the occurrence period and the duration period is determined according to a unified time benchmark. The spatial attribution between the road location and the road range is determined according to the aforementioned road location and the road range of the current abnormal road operation segment, and the boundary tolerance adopts the unified road spatial tolerance.

[0156] After filtering a road operation anomaly segment and obtaining multiple vehicle anomaly events, these events are then arranged according to the previously established vehicle passage order. The filtering process only changes the set of events entering the current road operation anomaly segment, while maintaining the relative order of the original vehicle passage, thereby forming a matching event sequence corresponding to the current road operation anomaly segment.

[0157] For example, in the candidate abnormal segment, vehicle abnormal events A, B, C, and D are stored in order of vehicle passage. When the current abnormal road segment only meets the road location and time conditions with A, C, and D, the matching event sequence is arranged according to the existing vehicle passage order of A, C, and D.

[0158] When the same vehicle anomaly event simultaneously meets the road range and duration conditions of multiple road operation anomaly segments within the same processing cycle, a deterministic attribution is performed on the candidate road operation anomaly segments.

[0159] In one implementation, the common duration of the vehicle anomaly event and the duration of each road operation anomaly segment are first compared, and the road operation anomaly segment with the larger common duration is identified as the priority assignment object; when the common duration is the same, the spatial assignment depth of the road location represented by the vehicle anomaly event in each road range is compared, that is, the road distance from the road location along the directed road direction to the nearest boundary of the corresponding road range, and the road operation anomaly segment with the larger spatial assignment depth is given priority; when the corresponding conditions are still at the same judgment position, the assignment result is determined according to the fixed order of the road operation anomaly segment identifiers.

[0160] After attribution processing, each vehicle anomaly event has a definite source road operation anomaly segment in the current road attribution processing chain. The matching event sequence saves the corresponding road operation anomaly segment identifier, as well as the vehicle identity, vehicle passage order, anomaly type, anomaly deviation direction, and anomaly deviation range of each vehicle anomaly event within it.

[0161] When S31 forms a road operation anomaly segment, the source of the corresponding road operation status has been retained. Therefore, in S32, the road location and time correspondence between the vehicle anomaly event and the road operation anomaly segment are established, and the spatiotemporal correspondence between the vehicle anomaly event and the corresponding road operation status is established.

[0162] After the matching event sequence is formed, the anomaly type and anomaly deviation direction are obtained from each vehicle anomaly event.

[0163] The anomaly type corresponds to the vehicle operating status index from which the vehicle anomaly event originates, and the anomaly deviation direction corresponds to the direction in which the vehicle operating status deviates from the vehicle operating baseline. Vehicle anomalies in the matched event sequence are grouped according to their anomaly type and anomaly deviation direction. Vehicle anomalies with the same anomaly type and the same anomaly deviation direction are placed into the same common deviation processing group.

[0164] The common deviation processing group must at least retain the identifier of the abnormal road operation segment, the abnormality type, the abnormal deviation direction, and the set of vehicle abnormal event identifiers constituting the processing group. After a vehicle abnormal event enters the common deviation processing group, it continues to retain the vehicle identity, vehicle passage order, time of occurrence, representative road location, and abnormal deviation interval.

[0165] When multiple anomaly types or multiple abnormal deviation directions exist within the same road operation anomaly segment, corresponding common deviation processing groups are formed. Each common deviation processing group independently forms a corresponding multi-vehicle common deviation interval, and the formation result is bound to the set of vehicle anomaly events in the current group.

[0166] Abnormal deviations participating in the same common deviation processing group adopt the same physical dimension or are processed by a unified deviation scale. The abnormal deviation range after unified deviation scale processing continues to maintain the abnormal deviation direction of each vehicle's abnormal event, and participates in subsequent common deviation processing with the maximum abnormal deviation amplitude under the unified deviation scale.

[0167] Figure 3 In the process, the matched event sequence enters the "Same anomaly type and consistent anomaly deviation direction?" judgment node. Vehicle anomaly events that meet the corresponding conditions enter the "Alignment along the same deviation direction with zero deviation as a common starting point" processing, and the corresponding vehicle anomaly events constitute the current common deviation processing group. In the same common deviation processing group, each vehicle anomaly event already has a unified anomaly type and anomaly deviation direction, and the subsequent common deviation range is determined according to the actual coverage of different vehicles in the unified deviation amplitude space.

[0168] Furthermore, the formation of the common deviation range for multiple vehicles involves the following processing:

[0169] S331. Align the abnormal deviation intervals of the selected vehicle abnormal events with zero deviation as the common starting point along the same deviation direction to obtain the aligned abnormal deviation intervals.

[0170] S332. Using the endpoints of each aligned abnormal deviation interval, divide the continuous deviation range along the deviation direction into multiple deviation sub-intervals, and count the number of vehicles covered by each deviation sub-interval.

[0171] S333: Merge adjacent deviation sub-intervals where the number of covered vehicles reaches a preset vehicle number threshold along the deviation direction to form a common deviation interval for multiple vehicles.

[0172] For the current common deviation processing group, obtain the abnormal deviation intervals corresponding to the abnormal events of each vehicle within the group. According to the aforementioned definition of abnormal deviation intervals, each abnormal deviation interval extends from the zero deviation starting point along the current abnormal deviation direction to the maximum abnormal deviation amplitude of the abnormal event of that vehicle.

[0173] During the alignment process, the zero-deviation positions of all vehicle anomalies are mapped to the same common starting point, and the anomaly deviation directions corresponding to the current common deviation processing group are unified to the same amplitude extension direction. After alignment, each vehicle's anomaly deviation interval has a common zero-deviation starting point and a common extension direction, and the cutoff position of each interval is determined by the maximum anomaly deviation amplitude of each vehicle's anomaly event.

[0174] Figure 3 The diagram shows a common starting point of zero deviation, the same deviation direction, and multiple aligned abnormal deviation intervals. Each abnormal deviation interval extends from the same common starting point of zero deviation along the same deviation direction and has its own ending point. This graphical relationship corresponds to the interval alignment result formed by S331.

[0175] Let the maximum abnormal deviation amplitude of vehicle abnormal event i in the current common deviation handling group be . Obtain the non-zero cutoff endpoints of each abnormal deviation interval within the current group along the common deviation direction. Merge the values ​​of the same endpoints and then arrange them in order of distance from the common zero deviation starting point from the nearest to the furthest point: ,in Indicates the common starting point of zero deviation. This represents the endpoint of the positive deviation after deduplication.

[0176] From adjacent endpoints to The range of continuous deviation amplitudes between them constitutes a deviation sub-interval. Since all abnormal deviation intervals within the current group extend continuously in the same direction from a common zero deviation starting point, when the maximum abnormal deviation amplitude of vehicle abnormal event i is... Reaching or exceeding the endpoint At that time, the abnormal deviation range corresponding to the abnormal event of the vehicle was completely covered by... This represents the current deviation sub-interval at the far boundary.

[0177] Therefore, the number of vehicles covered in each deviation sub-interval is counted based on the deviation depth corresponding to its farthest endpoint. The further the endpoint is from the common starting point of zero deviation, the greater the common deviation depth required for the corresponding vehicle's abnormal deviation. The number of covered vehicles is counted according to their different vehicle identities.

[0178] Let the current common deviation processing group be... The vehicle identity of vehicle abnormal event i is The corresponding maximum abnormal deviation amplitude is Then the deviation endpoint The corresponding number of covered vehicles is: ,in This indicates that the deviation endpoint has been reached. The number of vehicles corresponding to different deviation depths. This indicates the number of elements in the set, and the elements in the set are identified by vehicle identity. When the same vehicle corresponds to multiple vehicle anomaly events in the current common deviation processing group, the same vehicle identity participates in the coverage count as a set element at the same deviation depth. In other words, one vehicle identity corresponds to one vehicle coverage source at one deviation depth.

[0179] The preset vehicle quantity threshold is denoted as This is used to define the minimum number of different vehicles required to form a common deviation range among multiple vehicles. The value of is at least 2.

[0180] Since the abnormal deviation intervals for each vehicle extend continuously from the common starting point of zero deviation, as the deviation endpoints gradually move away from the common starting point of zero deviation along the abnormal deviation direction, the number of vehicle identities reaching the corresponding deviation depth remains the same or gradually decreases. The deviation sub-intervals covering the number of vehicles reaching the preset vehicle number threshold extend continuously outward from the zero deviation side.

[0181] The farthest effective endpoint from the positive deviation endpoints is obtained when the number of covered vehicles reaches a preset vehicle number threshold. The maximum common deviation of multiple vehicles is: ,in This indicates the maximum common deviation value among multiple vehicles in the current common deviation treatment group.

[0182] Extending from the common starting point of zero deviation along the direction of the current abnormal deviation to The continuous deviation amplitude range forms a common deviation interval for multiple vehicles. When processed according to deviation sub-intervals, the number of vehicles covered reaches... Adjacent deviation sub-intervals are continuously merged from the side of the common zero-deviation starting point, with the farthest valid endpoint corresponding to... .

[0183] The formation of an effective multi-vehicle common deviation interval is conditional upon at least one positive deviation endpoint satisfying a vehicle number threshold. Therefore, the zero-deviation common starting point only serves as the alignment reference for each vehicle's abnormal deviation interval, and the actual common abnormal amplitude boundary is determined by the positive deviation endpoint that meets the multi-vehicle coverage requirement.

[0184] Once the common deviation interval for multiple vehicles is formed, the current abnormal road segment identifier, common deviation processing group identifier, abnormal type, abnormal deviation direction, maximum common deviation value of multiple vehicles, corresponding vehicle abnormal event identifier set, and corresponding vehicle identity set are written into the common deviation processing record.

[0185] When the same abnormal segment of road operation corresponds to multiple common deviation processing groups, each common deviation processing record saves its own multi-vehicle common deviation interval and vehicle event source. Subsequent road attribution obtains the corresponding vehicle abnormal event according to the corresponding common deviation processing record.

[0186] Road operation anomaly segments are formed by the road operation state relative to the road operation baseline, while multi-vehicle common deviation intervals are formed by vehicle anomaly events that have completed spatiotemporal matching with the road operation anomaly segment. The two types of processing results establish object correspondence through road operation anomaly segment identification and common deviation processing records.

[0187] The road operation anomaly segment saves the road range, duration, and road operation status index type, used to record the road space, time range, and source of the road status index where the current road anomaly occurred. The common deviation processing record saves the anomaly type, anomaly deviation direction, multi-vehicle common deviation interval, and corresponding vehicle anomaly event set, used to record the multi-vehicle common response corresponding to the current road operation anomaly segment.

[0188] Road operation status indicators are stored according to their own physical dimensions, while vehicle abnormal deviations are stored according to the unified deviation scale corresponding to the common deviation processing group. The two types of values ​​are linked through the event objects of the current abnormal road operation segment.

[0189] When the current road operation anomaly segment has completed the spatiotemporal matching of vehicle anomaly events, and at least one corresponding common deviation processing group forms an effective multi-vehicle common deviation interval, the current road operation anomaly segment is confirmed with the corresponding multi-vehicle common deviation interval, thus forming a road segment anomaly event.

[0190] The abnormal road segment event retains the road range, duration, and road operation status index type of the abnormal road segment from which the abnormal event originates, and saves the identifier of the abnormal road segment from which the abnormal event originates, as well as one or more common deviation processing records involved in the current confirmation.

[0191] Each common deviation processing record maintains the binding relationship between the anomaly type, the direction of the abnormal deviation, the common deviation interval for multiple vehicles, and the corresponding set of abnormal events for the vehicle. When multiple valid common deviation processing groups correspond to the same road operation anomaly segment, the corresponding multiple common deviation processing records are jointly associated with the current road segment's anomaly event.

[0192] When performing road attribution for vehicle anomaly events, the corresponding vehicle anomaly event is obtained from the set of vehicle anomaly event identifiers in the common deviation processing record, and the multi-vehicle common deviation interval bound to this set is called for deviation decomposition. In this way, the road range and duration of the abnormal segment along the road are entered into the road segment anomaly event, and the multi-vehicle common deviation interval is entered into the subsequent vehicle anomaly attribution process along the common deviation processing record.

[0193] After an abnormal event occurs on a road segment, the corresponding common deviation processing record saves the common deviation interval of multiple vehicles, the abnormality type, the abnormal deviation direction, and the set of abnormal events of the vehicles involved in forming the common deviation interval. The common deviation interval of multiple vehicles in the same common deviation processing record is used to represent the common deviation depth of multiple vehicles in the current abnormality type and the current abnormal deviation direction. The abnormal deviation interval of each vehicle abnormal event represents the abnormal deviation depth reached by the corresponding vehicle itself.

[0194] Road attribution processing uses common deviation processing records as processing units, aligning and decomposing the common deviation intervals of multiple vehicles with the abnormal deviation intervals of each vehicle's abnormal events in the records. The common coverage area formed by the two intervals from the common zero deviation starting point along the same abnormal deviation direction is taken as the explanation range of the current vehicle's road interaction. The portion of the vehicle's abnormal deviation along this direction that exceeds the corresponding explanation boundary continues to form the vehicle's residual anomaly.

[0195] Specifically, the formation of the vehicle-road anomaly determination result includes the following processing:

[0196] S41. Based on the common deviation interval of multiple vehicles and the abnormal deviation interval of the corresponding vehicle abnormal event, determine the road attribution deviation interval for each corresponding vehicle abnormal event.

[0197] S42. Divide the abnormal deviation interval of the corresponding vehicle abnormal event according to the road attribution deviation interval, and form the remaining deviation interval by extending the part of the road attribution deviation boundary along the abnormal deviation direction.

[0198] S43. Identify the abnormal deviations within the remaining deviation interval as the vehicle residual anomalies of the corresponding vehicle abnormal events, and form the vehicle-road anomaly judgment result based on the road attribution deviation interval and road segment abnormal events, combined with the vehicle residual anomalies; retain the vehicle identity, road location and occurrence time of the corresponding vehicle abnormal event for the vehicle residual anomaly.

[0199] Each road attribution process corresponds to a common bias processing record formed above.

[0200] The common deviation processing record is associated with at least the current road segment abnormal event identifier, the source road operation abnormal segment identifier, the multi-vehicle common deviation interval, the abnormality type, the abnormal deviation direction, and the vehicle abnormal event identifier set. When processing the current record, the corresponding vehicle abnormal event is obtained one by one according to the vehicle abnormal event identifier set, and the abnormal deviation interval, maximum abnormal deviation amplitude, vehicle identity, representative road location, occurrence time, and abnormal state sampling of the event's internal association are read.

[0201] The shared deviation intervals and vehicle anomaly events in the current records originate from the same shared deviation processing group mentioned above. Therefore, they use the same anomaly type, the same anomaly deviation direction, and the same deviation scale. When an anomaly event on a road segment is associated with multiple shared deviation processing records, each record performs road attribution according to its own shared deviation intervals and vehicle anomaly event set, and each attribution result continues to retain the corresponding shared deviation processing record identifier.

[0202] The vehicle anomaly events have already been attributed according to the road operation anomaly segments. Therefore, the current vehicle anomaly events will continue to use the already identified road operation anomaly segments and road segment anomaly event sources. Within the current processing cycle, a vehicle anomaly event will enter the corresponding road attribution processing according to the established source chain.

[0203] Road attribution here refers to the explanation and attribution during the anomaly detection process. The road segment anomaly event has recorded the road range and duration of the road condition anomaly, and the multi-vehicle common deviation interval further provides the deviation amplitude boundary of the vehicle-side common response. This deviation amplitude boundary is used to determine the portion of the corresponding single-vehicle anomaly that falls within the scope of the road's collective effect explanation. The specific physical causes in the real-world environment are further determined based on road field perception, equipment diagnostic data, and operational records.

[0204] The maximum abnormal deviation amplitude corresponding to different vehicles has its own value. When the same multi-vehicle common deviation interval acts on abnormal events of different vehicles in the group, the road attribution deviation amplitude is also constrained by the upper bound of the abnormal deviation of the corresponding vehicle itself.

[0205] Furthermore, the formation of the road attribution bias interval includes the following processing:

[0206] S411. For each vehicle abnormal event corresponding to the common deviation interval of multiple vehicles, align the common deviation interval of multiple vehicles with the abnormal deviation interval of the vehicle abnormal event along the same deviation direction with zero deviation as the common starting point to obtain the interval alignment result.

[0207] S412. Determine the maximum deviation amplitude along the deviation direction for the two deviation intervals in the interval alignment result, and determine the smaller of the two values ​​as the road attribution deviation amplitude.

[0208] S413. Starting from zero deviation, extend along the deviation direction to the road attribution deviation amplitude to form the road attribution deviation range for the abnormal vehicle event.

[0209] For vehicle abnormal event i in the current common deviation processing record, obtain the multi-vehicle common deviation interval and the abnormal deviation interval of vehicle abnormal event i corresponding to the record.

[0210] Both intervals are described in amplitude form. The original direction of the abnormal deviation is the actual deviation direction of the vehicle's operating state relative to the vehicle's operating reference. Within the same common deviation processing record, this direction is uniformly used as the amplitude extension direction, and each deviation amplitude is expressed according to the distance from the zero deviation starting point along the current abnormal deviation direction.

[0211] In the aforementioned implementation method that uses the original deviation scale for common deviation processing, vehicle anomaly events continue to be attributed to roads using the same original deviation scale. In the aforementioned implementation method that uses a unified deviation scale, the common deviation processing record saves the corresponding scale identifier and scale conversion parameters. The current vehicle anomaly event obtains the corresponding maximum anomaly deviation amplitude according to the same scale conversion rule, so that the common deviation interval of multiple vehicles and the anomaly deviation interval of a single vehicle are at the same amplitude coordinate.

[0212] After scaling is achieved, the zero-deviation positions of the two intervals are mapped to the same common starting point, and the current abnormal deviation direction is set to the same amplitude extension direction. The two intervals thus form an interval alignment result with the same starting point and direction. Let the maximum deviation amplitude of the current multi-vehicle common deviation interval be... The maximum abnormal deviation amplitude of vehicle abnormal event i is .in Read the current common deviation processing record directly. Read the abnormal deviation range corresponding to the current vehicle abnormal event.

[0213] In the same common deviation processing record This record maintains consistency with the vehicle anomaly events it is linked to, and each vehicle has its own unique record. .

[0214] Since the two deviation intervals extend in the same direction from the same zero deviation starting point, the far boundary of their common coverage area is determined by the smaller of the two maximum deviation amplitudes. The road attribution deviation amplitude for vehicle anomaly event i is: ,in This represents the road attribution bias magnitude corresponding to vehicle abnormal event i.

[0215] when At that time, the vehicle's own abnormal deviation range reaches the amplitude endpoint first, and the road attribution deviation amplitude is taken as... .

[0216] when When the common deviation interval of multiple vehicles and the abnormal deviation interval of vehicles have the same amplitude endpoint, the road attribution deviation amplitude is taken as the common endpoint.

[0217] when When multiple vehicles share a common deviation range, the amplitude endpoint is reached first, and the road attribution deviation amplitude is taken as... .

[0218] Using the common zero-deviation starting point as the starting point of the road attribution deviation interval, it extends along the current abnormal deviation direction to... At the corresponding location, a road attribution deviation interval is formed for vehicle anomaly event i. The road attribution deviation interval stores the source vehicle anomaly event identifier, common deviation processing record identifier, anomaly type, anomaly deviation direction, and current deviation scale.

[0219] In the implementation using a unified deviation scale, the scale conversion parameters are saved based on the current common deviation processing records. The corresponding boundary is mapped to the state deviation coordinates of the source vehicle abnormal event, forming the road attribution boundary used for subsequent state sampling and segmentation within the vehicle event.

[0220] Figure 4 In this context, both the common deviation interval for multiple vehicles and the abnormal deviation interval for corresponding vehicle abnormal events extend from the common starting point of zero deviation along the same deviation direction. "Determining the smaller of the two maximum deviation amplitudes as the road attribution deviation amplitude" corresponds to the process of determining the far boundary where the two unidirectional intervals jointly cover. "Extending along the deviation direction to form the road attribution deviation interval" corresponds to extending from the common starting point of zero deviation to... The process of interval formation.

[0221] After the road attribution bias interval is formed, the road attribution bias amplitude is used. The corresponding position serves as the deviation decomposition boundary for vehicle abnormal event i.

[0222] The original abnormal deviation range of vehicle abnormal events has been extended from the common starting point of zero deviation to the maximum abnormal deviation amplitude. From the common starting point of zero deviation to The amplitude range at the corresponding location is the road attribution bias interval, extending from outside the decomposition boundary along the current abnormal bias direction to... The amplitude range of the corresponding endpoint is the remaining deviation interval.

[0223] The residual anomaly amplitude of vehicle event i is: ,in This indicates the maximum residual deviation depth of vehicle abnormal event i exceeding the current road attribution boundary.

[0224] when At that time, a residual deviation interval with actual amplitude length is formed from the road attribution boundary to the endpoint of the vehicle's maximum abnormal deviation, and the vehicle residual anomaly is established accordingly.

[0225] when When the current abnormal vehicle event corresponds to the road attribution deviation range, it extends to the endpoint of the vehicle's maximum abnormal deviation. The vehicle event is then recorded as the complete road attribution state under the current common deviation processing record.

[0226] Figure 4 The phrase "dividing the abnormal deviation intervals of corresponding vehicle abnormal events according to the road attribution deviation interval" corresponds to the above amplitude division. The road attribution deviation interval and the remaining deviation interval are connected along the same abnormal deviation direction. The remaining deviation interval in the figure continues to enter the "determining vehicle residual abnormality" process. Figure 4 use The diagram illustrates the relationship, therefore the figure shows the remaining deviation interval with actual length, the actual interval length according to... and The corresponding value is determined. This describes the maximum residual deviation depth of the current vehicle anomaly event. To determine the actual time position corresponding to this residual deviation during the occurrence of the vehicle anomaly event, the samples of the already associated anomaly states within the source vehicle anomaly event are further divided into samples on a sample-by-sample basis.

[0227] For each abnormal state sampled during the period in which the abnormal event occurred, the abnormal deviation amplitude of the sample relative to the vehicle's operating baseline is obtained and mapped to the current abnormal deviation direction according to the deviation scale used in the current common deviation processing record. When the sampling deviation amplitude is within the road attribution deviation interval, the sample is included in the road common interpretation sample; when the sampling deviation amplitude crosses the road attribution boundary, the portion of the amplitude exceeding the boundary is included in the vehicle residual abnormal sample.

[0228] Sampling with deviation amplitudes exactly at the road attribution boundary is included in the road attribution boundary. Vehicle residual anomaly sampling occurring continuously over time forms an actual occurrence sub-period. When vehicle residual anomaly sampling occurs across multiple time segments within the source vehicle anomaly event, multiple actual occurrence sub-periods are formed.

[0229] Each actual occurrence sub-time period is further associated with the road location of the corresponding abnormal state sampling. The temporal and road-space correspondence between vehicle operation state sampling and road segment matching results has been established above. Therefore, based on the sampling time of each state sampling within the actual occurrence sub-time period, the corresponding road location is obtained from the corresponding road segment matching record to form the road location record corresponding to that actual occurrence sub-time period.

[0230] For cases where a vehicle moves continuously within the actual sub-period, the road location record saves multiple consecutive road locations according to the sampling time sequence; for cases where subsequent processing requires the use of a single event road location, the vehicle residual anomaly continues to retain the representative road location of the source vehicle anomaly event, while also saving the road location sequence corresponding to the actual sub-period.

[0231] In this way, the remaining deviation interval is responsible for recording the amplitude boundary, the vehicle residual anomaly sampling is responsible for recording the actual state sampling of the residual deviation, the actual occurrence sub-time period is responsible for recording the time and location, and the road location recording is responsible for recording the corresponding road spatial location.

[0232] against Vehicle anomaly events are represented by residual deviation ranges, vehicle residual anomaly amplitudes, and corresponding actual residual samples to form vehicle residual anomaly objects.

[0233] One vehicle residual anomaly object corresponds to one source vehicle anomaly event and one common deviation processing record. The vehicle residual anomaly must at least store the source vehicle anomaly event identifier, common deviation processing record identifier, vehicle identity, representative road location, time of occurrence, anomaly type, anomaly deviation direction, residual deviation range, and vehicle residual anomaly amplitude.

[0234] Vehicle identity is inherited from the source vehicle's abnormal events, and is used for the continuous processing of residual abnormalities of the same vehicle organization.

[0235] The road location is inherited along the source vehicle anomaly event, used to store the main road spatial identity of the anomaly event. The road location record corresponding to the actual occurrence sub-time period is also written to the vehicle residual anomaly, used for subsequent road spatial matching according to specific time ranges.

[0236] The occurrence time period is saved along the source vehicle anomaly event, which assumes the overall temporal identity of the vehicle anomaly event; the actual occurrence sub-time period saves the time range in which the vehicle residual deviation actually crosses the road attribution boundary. When a vehicle residual anomaly has multiple actual occurrence sub-time periods, each sub-time period saves the corresponding start and end times and road position records.

[0237] The anomaly type and anomaly deviation direction are inherited along the source vehicle anomaly event. The remaining deviation range and vehicle residual anomaly amplitude adopt the deviation scale used in the current common deviation processing record; in the implementation using a unified deviation scale, the source deviation boundary corresponding to the scale conversion parameter is also saved.

[0238] Vehicle residual anomalies generated by the same vehicle in different vehicle anomaly events retain their respective source event identifiers. Further analysis is then conducted based on vehicle identity and temporal continuity to determine the continuity relationship between these vehicle residual anomalies.

[0239] When multiple common deviation processing records exist in the same road segment abnormal event, the residual abnormality of each vehicle continues to retain the identifier of the common deviation processing record to maintain a fixed source correspondence between the vehicle residual abnormality and the multi-vehicle common deviation interval used in this attribution.

[0240] After the common deviation processing records associated with the current road segment abnormal events are completed and the vehicle road attribution is completed, the road segment abnormal events, road attribution records, and vehicle residual abnormalities are organized into a vehicle-road abnormality judgment result.

[0241] For a common deviation processing record, the vehicle-road anomaly determination result saves the current road segment anomaly event identifier, common deviation processing record identifier, multi-vehicle common deviation interval, and the corresponding vehicle anomaly event set.

[0242] For each corresponding vehicle anomaly event, a road attribution record is generated. The road attribution record includes at least the vehicle anomaly event identifier, vehicle identity, anomaly type, anomaly deviation direction, maximum anomaly deviation amplitude of the vehicle, road attribution deviation amplitude, and road attribution deviation range, and is associated with the current common deviation processing record.

[0243] When the residual anomaly amplitude of a vehicle is greater than zero, the road attribution record continues to associate the corresponding residual anomaly object with the vehicle; when the residual anomaly amplitude of a vehicle is equal to zero, the road attribution record saves the complete road attribution state of the current vehicle event.

[0244] The object relationships in the vehicle-road anomaly determination results are maintained in the following order of origin: road segment anomaly events are associated with common deviation processing records, common deviation processing records are associated with multi-vehicle common deviation intervals and corresponding vehicle anomaly event sets, each vehicle anomaly event forms its own road attribution deviation interval, and the corresponding remaining deviations further form vehicle residual anomalies.

[0245] Subsequently, road segment anomaly events and residual anomalies of active vehicles are obtained from the vehicle-road anomaly determination results. Road segment anomaly events continue to be processed continuously using road range and duration; vehicle residual anomalies continue to be processed continuously using vehicle identity, occurrence time, actual occurrence sub-time period, and corresponding road location.

[0246] During the continuous updating of vehicle residual anomalies, the actual occurrence sub-time period and road location of the source vehicle anomaly event are maintained according to the original time-road spatial binding. When multiple vehicle residual anomalies continue to form, the actual occurrence sub-time periods of each source are retained as constituent units. The time tolerance used for continuous judgment between adjacent anomalies plays a role in connecting the anomaly life cycle, and the actual occurrence sub-time periods continue to serve as the data source for subsequent vehicle-road spatiotemporal rematching.

[0247] After road attribution processing is completed, road segment anomalies and vehicle residual anomalies enter a continuous processing phase. Road segment anomalies are described by road extent and duration, while vehicle residual anomalies are described by vehicle identity, occurrence time, and actual occurrence sub-time period. Both types of anomalies are recorded in chronological order and activity logs based on their respective continuity conditions. The updated road segment anomalies and vehicle residual anomalies are then re-matched according to road space and actual occurrence time.

[0248] Specifically, the generation of real-time anomaly detection results includes the following processing:

[0249] S51. According to the time sequence of road segment abnormal events and vehicle residual abnormalities in the vehicle-road abnormality determination results, perform continuous matching of adjacent road segment abnormal events that are continuous and connected in the road travel direction, and perform continuous matching of vehicle residual abnormalities that belong to the same vehicle and occur in connected time periods, to obtain the continuous matching results of road segment abnormal events and vehicle residual abnormalities.

[0250] S52. Based on the continuous matching results of road segment abnormal events and vehicle residual abnormalities, the road segment abnormal events and vehicle residual abnormalities that form continuous matching are continued and updated respectively, and the road segment abnormal events and vehicle residual abnormalities that do not form continuous matching are terminated and updated, so as to obtain the updated road segment abnormal events and vehicle residual abnormalities.

[0251] S53. Rematch the updated vehicle residual anomalies with the updated road segment anomalies according to the road location and occurrence time. Form a rematch result for vehicle residual anomalies and road segment anomalies whose road location falls within the corresponding road range and whose occurrence time coincides with the corresponding duration time. Retain the road segment anomalies and vehicle residual anomalies that have not formed a rematch. Generate real-time anomaly detection results based on the rematch results and the retained anomalies.

[0252] Obtain road segment anomaly events and vehicle residual anomalies from the vehicle-road anomaly determination results.

[0253] For road segment anomaly events, based on the location of the road segment within the road topology, the road segment is assigned to the corresponding directed traffic path or a road sequence with continuous directed connections. Road segment anomaly events within the same directed traffic path are arranged according to the start time of the duration; if the start times are the same, they are arranged according to the end time of the duration; when two time parameters are at the same system time resolution position, they are arranged according to the fixed order of the road segment anomaly event identifiers, forming a time sequence of road segment anomaly events under the corresponding road path.

[0254] When a road fork corresponds to multiple subsequent directed paths, each subsequent path stores its own time series of abnormal events. When the source road spans the fork location, the corresponding continuous path is determined based on the connection direction of its actual road range in the road topology.

[0255] For vehicle residual anomalies, they are first grouped according to vehicle identity, and then arranged according to the start time of the occurrence period. When the start time of the occurrence period is the same, they are arranged according to the end time of the occurrence period; when the corresponding time parameters are still at the same system time resolution position, they are arranged according to the fixed order of vehicle residual anomaly identifiers, forming a vehicle residual anomaly time sequence corresponding to the same vehicle.

[0256] Adjacent road segment anomalies in the time series of anomalies on the same road segment are included in road continuous matching; residual anomalies of adjacent vehicles in the time series corresponding to the same vehicle are included in vehicle continuous matching.

[0257] A continuous anomaly process is stored in real-time using an activity anomaly log. The road segment anomaly activity log must at least store the current directed road path, the source road segment anomaly events participating in the continuation, the current continuous processing boundary, and the duration of the last actual anomaly. The vehicle residual anomaly activity log must at least store the vehicle identity, the source vehicle residual anomaly participating in the continuation, the current time boundary, and the last actual occurrence time period.

[0258] Continuous processing proceeds from the activity anomaly record onwards. Newly arriving anomaly events are sequentially entered into the corresponding activity records for continuous evaluation according to a stable chronological order. For the current road segment's anomaly activity record and the next road segment's anomaly event in the chronological order, the actual road extent and actual duration of the last source road segment anomaly event in the activity record are obtained, as well as the road extent and duration of the next road segment anomaly event.

[0259] Along the direction of road traffic, the spatial interval of the road is determined by the end of the previous actual road area and the beginning of the next actual road area. When two road areas overlap or their directed road boundaries meet, Set to 0; when there are continuous directed road intervals between two road ranges. Take the road topology distance calculated along the corresponding directed travel path.

[0260] It has the dimension of road length, and its calculation is always carried out along the directional road spatial reference established above.

[0261] The time interval between the end time of the actual duration of an abnormal event on a previous source road segment and the start time of the duration of the next event was determined. When two actual durations share a common time or their time boundaries are adjacent, Set to 0; if a time interval exists, take the actual time length between the boundaries of the two time periods.

[0262] Let the threshold for continuous intervals in road space be . The threshold for the continuous interval of road anomalies is: The two are respectively with and They have the same dimensions. The continuity of the abnormal events in the current two road segments is: ,in It is a dimensionless quantity.

[0263] when When there is a directed connection between two road ranges along the current road travel direction, a continuous matching result of road segment abnormal events is formed.

[0264] The above maximum value processing subjectes both the road spatial interval and the time interval to their own continuous threshold constraints. In other words, the current continuous matching simultaneously satisfies both the road spatial continuity condition and the time continuity condition.

[0265] After a continuous match is formed, the abnormal event from the next source road segment is added to the abnormal activity record of the current road segment, and the end time of its duration and the actual road range boundary along the road travel direction are updated to the current boundary used for the next continuous judgment.

[0266] When multiple candidate road segment anomalies occur within the same processing cycle and are adjacent in time, they are processed sequentially according to the aforementioned determined time order. After each new source road segment anomaly is added, the updated current boundary is used to continue processing the next candidate event, ensuring that the formation order of the continuous chain is deterministic.

[0267] When the source road segment anomaly events have different road operation status index types, each source index type and the corresponding common deviation processing record are saved separately along with the source event. Continuous matching forms the continuity relationship of the road anomaly life cycle.

[0268] Vehicle residual anomaly continuous processing uses vehicle identity as the aggregation key.

[0269] For the current vehicle residual anomaly activity record corresponding to the same vehicle and the next vehicle residual anomaly in the time sequence, obtain the end time of the occurrence period of the last source vehicle residual anomaly in the activity record, and the start time of the occurrence period of the next vehicle residual anomaly.

[0270] When two events occur at the same time or have adjacent boundaries, the time interval is... Set to 0; when there is a time interval between the two occurrences, Take the corresponding time length.

[0271] Let the allowable continuous time interval for vehicle residual anomalies be . The continuity corresponding to the two vehicle residual anomalies is: ,in and It has the dimension of time. It is a dimensionless quantity.

[0272] When two vehicle residual anomalies have the same vehicle identity and At that time, a continuous matching result of vehicle residual anomalies is formed.

[0273] The continuous relationship of vehicle residual anomalies is established around the time process of the same vehicle. The residual anomalies of the source vehicles participating in continuous matching continue to save their own anomaly type, anomaly deviation direction, residual deviation interval, source vehicle anomaly event identifier, and common deviation processing record identifier.

[0274] When multiple vehicle residual anomalies have different anomaly types or different anomaly deviation directions, the updated vehicle residual anomaly activity records maintain a continuous lifecycle based on the vehicle identity, and save each anomaly type and anomaly deviation direction according to the source records.

[0275] The aforementioned actual occurrence sub-time periods and road location records are entered into the current activity record along with the source vehicle residual anomaly. When a source vehicle residual anomaly has multiple actual occurrence sub-time periods, each sub-time period is still saved as an independent time-road spatial record.

[0276] After a continuous match is formed, the residual anomaly of the next vehicle is added to the current vehicle activity record, and the end time of its occurrence period is updated to the current time boundary used for the next continuous judgment of vehicle residual anomalies.

[0277] Once a road segment anomaly event is matched continuously, the source road segment anomaly events that participated in the current continuous process will be organized together into an updated road segment anomaly event.

[0278] The updated road segment anomaly event record saves the active road path, activity start time, current activity end time, and all source road segment anomaly event identifiers. Simultaneously, it saves the road extent and actual duration of each source road segment anomaly event as a paired road anomaly support record.

[0279] Each road anomaly support record corresponds to an actual road area that has been confirmed as an anomaly and the actual duration of that area, and is further associated with the source road operation status indicator type and common deviation processing record.

[0280] There is a gap between abnormal events from adjacent source road sections. When there is a road spatial interval within the range, the road interval is registered as a road continuous tolerance interval; there exists a road spatial interval within the range. When the time interval falls within the specified range, this time interval is registered as a continuous tolerance interval. The continuous tolerance interval is used to maintain the updated status of road segment anomaly events, and the road anomaly support record continues to preserve the road range and duration for which actual road anomaly evidence was obtained.

[0281] Therefore, the updated road segment anomaly events have two types of temporal and spatial information: the activity lifecycle uses the overall activity boundary, and the spatiotemporal matching of road anomalies uses the source road anomaly support record.

[0282] After vehicle residual anomalies form a continuous match, the source vehicle residual anomalies that participate in the current continuous process are collectively organized into an updated vehicle residual anomaly.

[0283] The updated vehicle residual anomaly saves the vehicle identity, activity start time, current activity end time, and all source vehicle residual anomaly identifiers. Each source vehicle residual anomaly is paired with its corresponding road location in the aforementioned actual occurrence sub-time period to form a vehicle residual anomaly support record.

[0284] A vehicle residual anomaly support record shall include at least the actual occurrence sub-time period, the road location or road location sequence corresponding to the sub-time period, the anomaly type, the direction of the anomaly deviation, the remaining deviation interval, and the source vehicle residual anomaly identifier.

[0285] There is a relationship between residual anomalies from adjacent vehicle sources. When the time interval falls within the specified range, this time interval is registered as the continuous tolerance interval for vehicle anomalies. The activity time boundary is updated as new continuous source anomalies emerge, and the vehicle residual anomaly support record is still saved according to the actual occurrence sub-time period formed by sampling the true anomaly state.

[0286] In S52, the termination continuation for abnormal objects that have not formed a continuous match is executed after the corresponding activity abnormal record reaches the continuous waiting deadline. Before the continuous waiting deadline, the activity abnormal record remains in a waiting state and continues to receive subsequent similar abnormal objects for continuous matching. During real-time data processing, the current activity record enters a waiting state after the last actual source abnormality ends. The road segment abnormal activity record uses the end time of the last actual duration period as the time waiting starting point and the last actual road range as the benchmark for subsequent road spatial continuity judgment; the vehicle residual abnormal activity record uses the end time of the last actual occurrence period as the vehicle continuous waiting starting point.

[0287] When a new abnormal object arrives within the waiting period, a continuous judgment is performed according to the corresponding continuity and object identity conditions. When the current candidate abnormal object forms a continuous match, the activity record continues to be updated; when the current candidate object corresponds to another road path, another vehicle identity, or the continuity is outside the corresponding continuous range, the corresponding candidate object enters its own activity processing, while the original activity record continues to be maintained until its continuous waiting deadline.

[0288] The abnormal activity record for the road segment begins from the end time of the last actual duration period. Then it enters a closed state; the vehicle residual abnormal activity record is recorded from the end time of the last actual occurrence period. It then enters a closed state. The activity record after closure retains existing source anomaly support records and continuous tolerance interval records.

[0289] After the updated road segment anomalies and vehicle residual anomalies have completed their respective continuation processing, the vehicle-road association is re-established according to the road space and the actual anomaly time.

[0290] The updated road segment anomaly events are obtained from the internal road anomaly support records, which provide the corresponding road range and actual duration. The updated vehicle residual anomalies are obtained from the internal vehicle residual anomaly support records, which provide the road location and actual occurrence sub-time period.

[0291] In the rematching process, candidate vehicle-road object pairs are first formed based on the road range of each road anomaly support record in the updated road segment anomaly event and the road location of the vehicle residual anomaly after the update. Then, spatiotemporal verification is performed on the actual anomaly support records within the candidate object pairs.

[0292] For a vehicle residual anomaly support record, the road location within it is compared with the road range of a road anomaly support record. When saving the road location sequence in the vehicle residual anomaly support record, the road location corresponding to the actual occurrence time is obtained one by one for road spatial determination.

[0293] When a vehicle residual anomaly location falls within the road range corresponding to the current road anomaly support record, the actual occurrence sub-time period corresponding to that road location is compared with the actual duration period corresponding to the current road range. If both have a common time, the current vehicle residual anomaly support record and the road anomaly support record form an actual spatiotemporal matching unit.

[0294] When a vehicle residual anomaly support record corresponds to multiple road locations, the road locations and times continue to be used according to the aforementioned sampling time correspondence. In this way, when a vehicle moves along the road within the same actual occurrence sub-time period, the time and road location maintain a paired relationship.

[0295] When an updated road segment anomaly event has multiple road anomaly support records, each one is compared with the vehicle residual anomaly support record in the above spatiotemporal verification; when an updated vehicle residual anomaly has multiple support records, each support record participates in the verification separately.

[0296] As long as at least one actual spatiotemporal matching unit is formed, that is, a rematching result is formed between the current updated road segment anomaly event and the updated vehicle residual anomaly. When a vehicle residual anomaly forms actual spatiotemporal matches with multiple updated road segment anomaly events simultaneously, corresponding rematching records are formed respectively, and each record saves its own road segment anomaly event identifier and actual matching unit.

[0297] The rematch record should at least save the road segment abnormal event identifier, vehicle residual abnormal identifier, vehicle identity, matching road range, matching road location, actual occurrence sub-time period, corresponding actual duration period, and the common time range of the two.

[0298] Figure 5 In the updated road segment anomaly events, the corresponding road range and duration are defined, while the updated vehicle residual anomalies correspond to the road location and occurrence time. Figure 5 The “spatial matching (road spatial comparison)” in the figure corresponds to the comparison between road location and road range, and the “temporal matching (time interval comparison)” corresponds to the comparison between occurrence time and duration time. The “overlapping time period” in the figure represents the time portion that is actually covered by both types of time ranges.

[0299] Both road spatial comparison and temporal comparison are used to determine "Does the road location fall within the corresponding road range and the time period coincide with the corresponding duration?". When a corresponding road spatial and temporal coincidence relationship is formed, the results are re-matched; when corresponding vehicle-road objects form their own independent storage relationships, road segment abnormal events and vehicle residual abnormalities are respectively stored in their respective retention records.

[0300] Figure 5The road range, duration, road location, and occurrence time in the data correspond to the attributes directly read during rematching. The road anomaly support records and vehicle residual anomaly support records within the updated object serve as the actual source of the above attributes, ensuring that the continuous tolerance interval remains in the activity lifecycle record, and the actual anomaly support records are entered into vehicle-road spatiotemporal matching.

[0301] The rematching process uses the previously established road attribution results. The multi-vehicle common deviation interval, road attribution deviation interval, and vehicle residual anomaly amplitude all retain the original attribution records. The rematching results add the current road-time correspondence formed after the two types of anomalies continue to evolve.

[0302] Within the current processing cycle, updated road segment anomalies and vehicle residual anomalies that have been rematched will be written into the rematching results. The remaining updated road segment anomalies and vehicle residual anomalies will have their respective object states saved. The real-time anomaly detection results will be formed by combining the rematching results and the separately saved anomalies.

[0303] Real-time anomaly detection results include at least road segment anomaly event records, vehicle residual anomaly records, and rematch records.

[0304] The road segment anomaly event log stores the current activity status, activity road path, activity time boundary, road anomaly supporting records, and source road segment anomaly events. The vehicle residual anomaly log stores the vehicle identity, activity status, activity time boundary, vehicle residual anomaly supporting records, and source vehicle residual anomalies. The rematching log stores the aforementioned actual spatiotemporal matching units and corresponding object identifiers.

[0305] When an active object enters a waiting state, it continues to be stored in the activity set corresponding to the real-time anomaly detection result; after a new continuous match is formed, the corresponding activity record is updated; after the continuous waiting deadline is reached, it transitions to a closed record.

[0306] When real-time data enters the system, abnormal records of the current active road segment are located based on road path identifiers, and residual abnormal records of the current active vehicle are located based on vehicle identity. New road segment abnormal events and vehicle residual abnormalities are continuously matched with the corresponding activity records. After continuous updates are completed, vehicle-road rematching is performed on active objects that have changed within the current processing cycle.

[0307] Historical anomaly records that have already been saved are retained as object source chains, while continuous processing in the current cycle revolves around newly added anomaly objects and active anomaly records. As a result, road segment anomaly events continue to evolve along the road space and time, while vehicle residual anomalies continue to evolve along the vehicle identity and time. After each new actual anomaly supporting record is formed, the corresponding vehicle-road spatiotemporal associations of both types of objects are updated.

[0308] Based on the aforementioned S1 to S5 and corresponding processing procedures, vehicle trajectory time-series data, vehicle operation status time-series data, road topology data, and road operation status time-series data are processed according to a unified time reference and a unified road spatial reference. Parameter records are established for vehicle operation reference, road operation reference, deviation scale, vehicle quantity threshold, and continuous processing parameters, and these records are kept consistent with the corresponding abnormal objects through parameter versions.

[0309] Vehicle trajectory time-series data, vehicle operation status time-series data, and road operation status time-series data all retain the source sampling time. The computing nodes separately record the data reception time for data transmission and real-time processing scheduling. Event formation, time sorting, determination of occurrence time periods, and determination of duration periods are all based on the unified system time corresponding to the source sampling time.

[0310] When different data sources use different sampling periods, their original sampling times are retained. When the corresponding road position needs to be obtained based on the vehicle's operating status, the aforementioned road position alignment is performed based on the position of the vehicle's state sampling time between adjacent trajectory sampling times. When matching the road operating status with vehicle abnormal events in time, the unified state time, occurrence time, and duration time are directly used for processing.

[0311] When there are discrepancies in the timing of network transmissions, the computing nodes write the data to the corresponding time position according to the source sampling time. Data arriving within the preset transmission waiting range is assigned to the corresponding processing window according to the source sampling time, and the data in the current time window is rearranged according to the source time. After the processing deadline is reached, the corresponding window enters the subsequent object formation process.

[0312] When records with the same sampling identifier are generated due to communication retransmission from the same data source, the records are merged according to the sampling identifier, so that a sample from one source corresponds to a valid data position in the corresponding time series data.

[0313] The road spatial reference is based on the road segment identifiers, directed connectivity, road centerlines, and road segment lengths in the road topology data. Before entering road processing, vehicle positioning coordinates are converted to the coordinate system used in the road topology data, and road operation status records are associated with the corresponding road status units and road location ranges.

[0314] After the vehicle trajectory location is mapped to the road, it is described using matching road segments and locations within road segments. When calculating road distances across multiple road segments, the road coordinates are unified according to the previously established directed path cumulative location, so that the trajectory location, the road location of vehicle abnormal events, the road state unit, and the road range of road segment abnormal events have a common road spatial reference.

[0315] When a vehicle trajectory has a short sampling interval, time alignment is performed based on the positions of adjacent valid trajectories, sampling times, and the continuity of road topology. When the trajectory position deviates from the preceding and following road paths to the point of meeting the positioning screening criteria, the corresponding trajectory sampling is marked as low-confidence sampling, and the current road traffic process is maintained by the preceding and following valid trajectories.

[0316] When smoothing is applied to vehicle and road operating status, the smoothing window length is set according to the corresponding data sampling period and anomaly retention time, ensuring that the processed data retains the direction of anomaly deviations, the sequence of anomaly peaks, and the timing of anomalies. Preprocessing records save the source sampling identifier, the post-processing sampling identifier, and the preprocessing parameter version.

[0317] Vehicle operating baselines are maintained separately according to vehicle identity and vehicle operating status indicators. Road operating baselines are maintained separately according to road status units and road operating status indicators.

[0318] In one implementation, the most recent normal operating window is saved for the vehicle's operating status, and the median value of the corresponding status index within the window is used as the vehicle's operating benchmark. The length of the normal operating window is set according to the vehicle operating status sampling period, the normal rate of change of vehicle status, and the abnormal retention time.

[0319] When a vehicle enters a new road segment, once the number of normal state samples in the current stage reaches the preset baseline number, the vehicle's operating baseline is updated using the normal samples from the current stage. If the current stage is in the normal sample accumulation phase, the vehicle's operating baseline formed during its most recent normal operating window continues to be used.

[0320] Once a vehicle enters an abnormal operation phase, the vehicle operation baseline used at the start of the abnormality is written into the current vehicle abnormality record, and this baseline version is maintained throughout the duration of the current abnormality. Once the vehicle status returns to the normal range and the required number of recoverable samples or the recovery duration are reached consecutively, subsequent normal samples enter a new normal operation window, forming the subsequent vehicle operation baseline version.

[0321] Road operation benchmarks are managed in the same manner. For road state units and road operation state indicators, state samples are obtained from the normal time window prior to the occurrence of a road anomaly, and the window statistics form the road operation benchmark. After a road state enters an active anomaly process, the current road operation anomaly segment continues to use the road operation benchmark version corresponding to the start of the anomaly. Once the road state recovers to the normal range and meets the recovery conditions, subsequent normal samples form a new road operation benchmark version.

[0322] Road conditions exhibiting clear temporal patterns are used to establish a normal sample set based on road condition units, date attributes, and time segments. For example, normal sample records are established for weekday morning rush hour, weekday off-peak hour, and holidays, respectively, and real-time road conditions are used to obtain corresponding road operation benchmarks based on the current date attribute and time segment.

[0323] Anomaly detection parameters include at least the anomaly trigger threshold, the number of samples to maintain an anomaly or the duration of an anomaly maintenance, the anomaly recovery threshold, and the number of samples to restore or the duration of restoration. The anomaly trigger threshold is determined based on the fluctuation range of normal samples relative to the normal baseline, and the anomaly recovery threshold corresponds to the normal state regression range.

[0324] Each baseline record stores the baseline identifier, status indicator type, sample time range, statistical method, and parameter version. When a vehicle anomaly event or road operation anomaly segment is formed, the corresponding baseline identifier and anomaly judgment parameter version are written. An existing anomaly object uses the parameter version corresponding to its initial creation throughout its lifecycle; subsequently created anomaly objects use the updated parameter version.

[0325] Vehicle anomaly types are established based on vehicle operating status indicators. Speed ​​status, longitudinal acceleration status, lateral acceleration status, braking status, and steering status each form a corresponding anomaly type. Vehicle anomaly events in a common deviation handling group adopt the same anomaly type and the same anomaly deviation direction.

[0326] When the state deviations of each vehicle are already within the same physical unit and similar operating scale under the same anomaly type, the corresponding state deviation amplitude is directly used to process the anomaly deviation range.

[0327] When different vehicles exhibit significantly different normal operating scales under the same anomaly type, the deviations in the state of the source vehicles are converted to a unified relative deviation scale. Taking a speed reduction anomaly as an example, a relative decrease value is formed based on the vehicle's speed operating baseline and the current speed reduction magnitude, ensuring that the speed anomalies corresponding to different vehicles are brought into the same proportional scale.

[0328] When using a unified deviation scale, each vehicle anomaly event saves the source vehicle's operating baseline, the source anomaly deviation amplitude, the unified deviation amplitude, the scale type, and the scale conversion parameters. The common deviation processing record saves the scale type and scale version used by the current processing group.

[0329] The aforementioned method generates the maximum common deviation value for multiple vehicles within a unified deviation scale. The aforementioned method obtains the maximum abnormal deviation amplitude of vehicle abnormal events within the same scale. And form the road attribution bias amplitude. .

[0330] When further determining vehicle residual anomalies based on internal state sampling of vehicle anomaly events, the road attribution bias boundary is mapped to the corresponding source state bias coordinates based on the scale conversion parameters saved for that vehicle anomaly event. Each state sample is compared with the source vehicle state quantity to obtain the actual residual sample, the actual occurrence sub-time period, and the road location record.

[0331] Road operation status indicators continue to retain their own physical dimensions. Average road speed, road occupancy, traffic flow, etc., each generate their own road anomaly records. Road operation anomaly segments and vehicle common deviations are correlated through road location, time range, and anomaly event identity, with vehicle-side deviation scales and road-side state scales stored separately.

[0332] Reference road location used for equivalent travel time calculation The order is determined before the current vehicle incident handling group enters the passage order calculation and remains consistent during the sorting process of that handling group.

[0333] In one implementation, the directed path cumulative position corresponding to each vehicle abnormal event in the current processing group is obtained, and the middle road position between the minimum cumulative position and the maximum cumulative position is used as the reference road position.

[0334] In a fixed road monitoring scenario, a fixed reference section is set according to the current directed road path, and the cumulative position of the directed path at the reference section is used as the reference point. The same reference location is used uniformly by the same vehicle incident handling group.

[0335] When the road location span of a vehicle anomaly event within the current processing group reaches the preset reference conversion distance, the event is divided into several consecutive processing groups according to the adjacent road locations along the directed path, and each processing group establishes its own reference road location. Events within the group use the same... Complete the calculation of equivalent passage time.

[0336] The road location corresponding to vehicle abnormal event i Sampling locations of neighboring trajectories Once the cumulative displacement of the directed paths between them reaches the minimum effective road displacement threshold, the corresponding trajectory sampling enters the equivalent passage time formula.

[0337] The minimum effective road displacement threshold is set based on the vehicle positioning accuracy, road position quantization accuracy, and vehicle trajectory sampling interval. Its value is greater than the typical displacement scale corresponding to the current road position measurement error, so that the local road displacement and time change have a stable conversion basis.

[0338] When the road displacement between the most recent trajectory sample and the representative road location is below the effective displacement threshold, the system continues to acquire earlier neighboring trajectory samples along the vehicle trajectory time sequence until a trajectory sample pair that meets the effective displacement condition is formed.

[0339] The preset vehicle number threshold in common deviation processing is denoted as Its value corresponds to at least two different vehicles. When using a fixed vehicle number threshold in a fixed traffic scenario, [the threshold will be...]. Save along with the road scene configuration.

[0340] When using dynamic vehicle quantity configuration, the current vehicle quantity threshold is formed based on the number of valid different vehicle identities within the current common deviation processing group and the preset ratio, and remains consistent during the formation of the current common deviation processing record.

[0341] The number of valid distinct vehicle identities in the current common deviation processing group has reached Then, the calculation proceeds to the shared deviation interval of the multiple vehicles. The number of valid vehicle identities is within... In the following cases, the current common deviation processing record remains in the candidate state, and the current set of valid vehicle identities will continue to be updated after subsequent matching vehicle abnormal events.

[0342] Common deviation processing record saves the current The system includes vehicle identity sets and threshold configuration versions, ensuring a unified processing standard for the aforementioned vehicle quantity statistics and subsequent road attribution.

[0343] The road continuity range parameters used in the formation of candidate anomaly sections are determined based on road topology granularity, vehicle positioning accuracy, and road traffic scenario.

[0344] The aforementioned road space continuity interval threshold Road anomaly time continuous interval threshold Vehicle residual anomaly continuous time interval threshold All configuration values ​​are greater than zero.

[0345] Configured based on road topology segment length, road state unit length, and road location accuracy. When the road state spatial unit is long, it is set according to the allowable road interval between adjacent road anomaly records. When the road topology is finely divided, it should be configured according to the corresponding road length scale.

[0346] The configuration is based on the road operation status update cycle and the duration of short-term abnormal road fluctuations. When the road operation status is updated at fixed intervals, the time corresponding to one or more road status update cycles is used as the basis for parameter configuration.

[0347] Based on the vehicle operation status sampling period and the duration scale of abnormal short-term fluctuations in the vehicle, the parameter configuration is based on the time length corresponding to several vehicle status sampling periods.

[0348] In implementations with historical operational data, statistical distributions are obtained from the time intervals between adjacent road status records during normal operation, short-term intermittent intervals within vehicle anomaly segments, and spatial segmentation intervals of road status, and then formed based on a preset high quantile range. , and The initial configuration values ​​are then used to create parameter versions based on road scenarios and operational strategies.

[0349] Once an abnormal event is established for an active road segment, its continuous processing follows the method used at the time of establishment. and Parameter version. Once an active vehicle residual anomaly is established, its continuous processing continues to use the version corresponding to the one established at the time of establishment. Parameter version. Newly created active objects are processed according to the currently effective parameter version.

[0350] When the continuous road range of a candidate anomaly section is smaller than the smallest spatial unit of road operation status data, obtain the smallest set of road status units that completely cover the continuous road range, and write the corresponding unit identifier, road location range, and actual coverage range of the candidate anomaly section into the road analysis record.

[0351] When a candidate abnormal segment spans multiple road state units, a continuous set of road state units is obtained along the road traffic direction, and each state unit participates in the formation of abnormal road operation segments according to its own road range.

[0352] When road operation status data is in a transmission waiting state during the current processing cycle, candidate abnormal sections are registered as road status waiting objects, and vehicle abnormal events and vehicle passage order are continued to be saved. After the corresponding road operation status enters within the preset data waiting range, the road operation status processing of the candidate abnormal section is restored according to the source status time.

[0353] When the representative road location of a vehicle anomaly event is in an undetermined state, the vehicle identity, occurrence time, anomaly type, anomaly deviation direction, and anomaly deviation range are saved, and the event is registered in the road location waiting set. After the road segment matching results complete the corresponding road location, the vehicle anomaly event enters the vehicle traffic order calculation and candidate anomaly segment merging.

[0354] When the same vehicle anomaly event meets the road location and time conditions with multiple road operation anomaly segments, the current attribution source is determined according to the aforementioned time overlap, spatial attribution depth, and fixed object order, and the corresponding attribution result is written into the source association record between the vehicle anomaly event and the road operation anomaly segment.

[0355] When the time difference between the occurrence time boundary and the duration time boundary falls within the time boundary tolerance, it is treated as a boundary contiguous relationship. When the road distance between the vehicle's road position and the road boundary falls within the spatial boundary tolerance, it is treated as a boundary inclusion relationship.

[0356] The time boundary tolerance is configured based on the system time synchronization accuracy, vehicle state sampling period, and road state update period. The spatial boundary tolerance is configured based on the vehicle positioning accuracy, road geometric accuracy, and road state unit boundary accuracy.

[0357] The aforementioned continuous tolerance intervals for road space, road time, and vehicles are stored as lifecycle attributes of activity anomalies. Road anomaly support records still retain the actual road extent and duration, while vehicle residual anomaly support records still retain the actual occurrence sub-time period and corresponding road location. Vehicle-road re-matching retrieves the spatiotemporal attributes from the actual anomaly support records.

[0358] Each object being processed is identified by a unique object identifier, and cross-stage data correspondence is established through the source object identifier.

[0359] The vehicle abnormal event record includes at least the vehicle abnormal event identifier, vehicle identity, representative time sequence location, representative road location, time period of occurrence, abnormal type, abnormal deviation direction, maximum abnormal deviation amplitude, abnormal deviation range, vehicle operating reference identifier, abnormal judgment parameter version, deviation scale type, and the set of state sampling identifiers that constitute the current vehicle abnormal event.

[0360] When using a unified deviation scale, vehicle anomaly event records further preserve the source deviation amplitude, unified deviation amplitude, and scale conversion parameters. Road attribution processing re-acquires each actual sample within the event based on the state sampling identifier set to form the actual occurrence sub-time period of vehicle residual anomalies.

[0361] The record of abnormal road operation segments includes at least the abnormal road operation segment identifier, the source candidate abnormal section identifier, the type of road operation status index, the road range, the duration, the source road status unit set, the source road status sampling identifier set, the road operation benchmark identifier, and the version of the abnormality judgment parameter.

[0362] The road segment abnormal event record should include at least the road segment abnormal event identifier, the source road operation abnormal segment identifier, the road range, the duration, the road operation status index type, and the common deviation processing record identifier set.

[0363] The common deviation processing record should include at least the common deviation processing record identifier, the identifier of the source road operation anomaly segment, the anomaly type, the direction of the anomaly deviation, the deviation scale type, the scale conversion information, and the vehicle quantity threshold. Valid vehicle identity set, common deviation interval for multiple vehicles, and maximum common deviation value for multiple vehicles. And the corresponding set of vehicle abnormal event identifiers.

[0364] The common deviation intervals of multiple vehicles in the common deviation processing record are bound to the set of vehicle abnormal event identifiers to form a fixed source. The road attribution processing obtains the corresponding vehicle abnormal events one by one according to the record.

[0365] Road attribution records should include at least the common deviation handling record identifier, the source vehicle abnormal event identifier, and the maximum abnormal deviation amplitude of the vehicle. Road attribution bias amplitude Information on road attribution bias intervals, bias scale types, and source scale mapping.

[0366] Vehicle residual anomaly records should include at least the vehicle residual anomaly identifier, the source vehicle anomaly event identifier, the common deviation handling record identifier, the vehicle identity, the location of the representative road, the time of occurrence, the anomaly type, the direction of the anomaly deviation, the remaining deviation range, and the vehicle residual anomaly amplitude. And a set of vehicle residual anomaly support records.

[0367] Each vehicle residual anomaly support record must at least store the actual occurrence sub-time period, the corresponding state sampling identifier set, and the corresponding road location record. When a vehicle moves continuously within an actual occurrence sub-time period, the road location record is configured according to the sampling time and road location correspondence.

[0368] The updated road segment anomaly event log includes at least the active road segment anomaly event identifier, directed road path identifier, activity start time, activity end time, activity status, source road segment anomaly event identifier set, road anomaly support record set, and continuous tolerance interval record.

[0369] Each road anomaly support record stores the actual road extent, corresponding actual duration, source road segment anomaly event identifier, road operation status index type, and corresponding common deviation handling record identifier. Continuous tolerance interval records separately store the spatial tolerance range and temporal tolerance range.

[0370] The updated vehicle residual anomaly record includes at least the active vehicle residual anomaly identifier, vehicle identity, activity start time, activity end time, activity status, source vehicle residual anomaly identifier set, vehicle residual anomaly supporting record set, and vehicle continuous tolerance interval record.

[0371] The actual occurrence time period and corresponding road location are kept in the vehicle residual anomaly support record, and the continuous tolerance time is registered in the activity lifecycle field. The two types of time data respectively serve as the actual source of the anomaly and the connection of the activity lifecycle.

[0372] The rematch record includes at least the rematch record identifier, the active road segment abnormal event identifier, the active vehicle residual abnormality identifier, the vehicle identity, the road abnormality support record identifier, the vehicle residual abnormality support record identifier, the matched road range, the matched road location, the actual duration, the actual occurrence sub-time period, and the common time range.

[0373] Each active object is set to active, waiting, and closed states. The active state corresponds to the current acquisition of a new continuous source of exceptions, the waiting state corresponds to the time process during which the object remains in a continuous waiting range after the last source of exceptions has ended, and the closed state corresponds to the completion of the current continuous lifecycle.

[0374] The source of objects is stored along the following data chain: vehicle trajectory forms road segment matching results, vehicle operating status forms vehicle abnormal events, vehicle abnormal events form candidate abnormal segments, road operating status forms road operating abnormal segments, road operating abnormal segments are associated with common deviation processing records, common deviation processing records are associated with multi-vehicle common deviation intervals and vehicle abnormal event sets, vehicle abnormal events further form road attribution records and vehicle residual anomalies, road segment abnormal events and vehicle residual anomalies enter their respective activity records, and when they actually overlap in time and space, they form rematching records.

[0375] Real-time processing employs an execution method that combines fixed-period updates with exception object triggering.

[0376] Vehicle trajectory time-series data is continuously fed into road segment matching processing. New trajectory locations are updated to show the current matched road segment, the vehicle's position within the current road segment, and the road segment's travel phase. Once a vehicle's operating status enters the corresponding road segment's travel phase, a vehicle anomaly event is generated based on the current vehicle operating baseline and anomaly judgment parameters.

[0377] Road operation status is continuously updated in road analysis according to the road status update cycle. When a candidate abnormal section has been formed, the corresponding road status unit is located according to the road range corresponding to the candidate abnormal section, and the addition of a new road status triggers the update of the road operation abnormal segment.

[0378] After a vehicle anomaly event is generated or updated, the current vehicle anomaly event handling group is located based on the directed traffic path identifier to complete the vehicle traffic order update and candidate anomaly segment processing. After a road operation anomaly segment is generated, the vehicle anomaly event is obtained based on the corresponding candidate anomaly segment, and spatiotemporal matching and common deviation processing are performed.

[0379] Anomalies in active road segments are indexed based on directed traffic path identifiers, road extent, and activity time range. When a new anomaly occurs, the active road segment anomalies with road spatial proximity are first retrieved through the road path index, and then further indexed based on duration and... , The parameters are used for continuous matching.

[0380] Active vehicle residual anomalies are indexed by vehicle identity. When a new vehicle residual anomaly arrives, the corresponding active vehicle residual anomaly record is directly located based on the vehicle identity, and then further categorized by the time period and... The parameters are used for continuous matching.

[0381] Vehicle-road rematching employs road spatial indexing and time range indexing. Based on the road location in the vehicle residual anomaly support record, active road segment anomaly events covering that road location are obtained. Then, objects that currently overlap in time and space are filtered based on the actual occurrence sub-time period and the actual duration in the road anomaly support record.

[0382] The current processing cycle focuses on newly added objects, active objects with updated attributes, and active objects still in a waiting state. When the state of an active object changes, the corresponding index record is updated. Closed objects are written to the history object collection, retaining the source event, parameter version, road attribution records, and re-matching records.

[0383] The road segment abnormal activity log stores the deadline for continuous waiting due to road abnormalities, while the vehicle residual abnormal activity log stores the deadline for continuous waiting due to vehicle abnormalities. When a new abnormal object enters before its corresponding deadline, the activity log is updated according to the continuous matching condition. Upon reaching the corresponding deadline, the activity state transitions to a closed state.

[0384] In one deployment method, the vehicle side collects GNSS trajectory data as well as speed and longitudinal acceleration data from the vehicle's CAN bus. The roadside uses roadside radar to generate average speed and road occupancy data for road segments. Road topology data is provided by the electronic road network, and cloud computing nodes complete S1 to S5 processing.

[0385] Another deployment method involves edge computing nodes matching vehicle trajectory segments, generating vehicle anomaly events, and creating abnormal road operation segments. These nodes then send the corresponding object records and their source identifiers to the cloud. The cloud then uses the vehicle anomaly events, directed road locations, and abnormal road operation segments to generate vehicle traffic order, common deviation handling records, road attribution records, and real-time anomaly detection results.

[0386] GNSS data interface, vehicle CAN data interface, roadside radar traffic parameter interface, video traffic parameter interface, electronic road network spatial index, road centerline projection, data cache, structured message transmission and database storage are accessed according to the corresponding device and computing platform interface specifications, and each object record is continuously written to the corresponding processing chain according to the source sampling identifier, object identifier and parameter version.

[0387] Based on the above-described implementation of the real-time anomaly detection method for vehicle-road-cloud collaboration, this application also discloses a real-time anomaly detection system based on vehicle-road-cloud collaboration. The system is implemented by a computer device executing corresponding programs, and each unit in the system is a program module, a processing module, or a combination of a program module and a processing device that performs corresponding data processing functions. For example... Figure 6 As shown, the system includes the following units:

[0388] The road segment matching unit 110 is used to acquire vehicle trajectory time-series data, vehicle operation status time-series data, road topology data and road operation status time-series data, and to determine and update the road segment matching results based on the vehicle trajectory time-series data and road topology data.

[0389] The candidate abnormal segment forming unit 120 is used to extract vehicle abnormal events and corresponding abnormal deviation intervals from the vehicle operation status time series data according to the road segment matching results, determine the vehicle passage order of each vehicle abnormal event, and merge the vehicle abnormal events according to the vehicle passage order to form candidate abnormal segments.

[0390] The road segment anomaly confirmation unit 130 is used to perform spatiotemporal matching of vehicle anomaly events in the candidate anomaly segment and the road operation status time series data, extract the common deviation interval of multiple vehicles from the matched vehicle anomaly events, and confirm the anomaly of the matched road operation status accordingly to form a road segment anomaly event.

[0391] The vehicle-road anomaly determination unit 140 is used to decompose the abnormal deviation interval of the corresponding vehicle anomaly event by the common deviation interval of the multiple vehicles, to obtain the road attribution deviation interval and vehicle residual anomaly, and to form a vehicle-road anomaly determination result by combining the road segment anomaly event.

[0392] The real-time anomaly detection unit 150 is used to continuously update the road segment anomaly events and vehicle residual anomalies according to the time sequence of the road segment anomaly events and vehicle residual anomalies in the vehicle-road anomaly determination result, and re-match the updated road segment anomaly events and vehicle residual anomalies to generate a real-time anomaly detection result.

[0393] The above description is merely a preferred embodiment of this application. It should be understood that this application is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technical or knowledge. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of this application should fall within the protection scope of this application.

Claims

1. A real-time anomaly detection method based on vehicle-road-cloud collaboration, characterized in that, include: S1. Obtain vehicle trajectory time-series data, vehicle operation status time-series data, road topology data, and road operation status time-series data; determine and update road segment matching results based on vehicle trajectory time-series data and road topology data. S2. Based on the road segment matching results, extract vehicle abnormal events and corresponding abnormal deviation intervals from the vehicle operation status time series data, determine the vehicle passage order of each vehicle abnormal event, and merge vehicle abnormal events according to the vehicle passage order to form candidate abnormal segments. S3. Perform spatiotemporal matching of vehicle anomaly events in the candidate anomaly segment and road operation status time series data, extract the common deviation interval of multiple vehicles from the matched vehicle anomaly events, and confirm the anomaly of the matched road operation status accordingly to form a road segment anomaly event. S4. Decompose the abnormal deviation intervals of the corresponding vehicle abnormal events into the common deviation intervals of the multiple vehicles to obtain the road attribution deviation intervals and vehicle residual anomalies, and combine them with the road segment abnormal events to form a vehicle-road anomaly determination result. S5. According to the time sequence of road segment abnormal events and vehicle residual abnormalities in the vehicle-road abnormality determination result, continuously update the road segment abnormal events and vehicle residual abnormalities respectively, rematch the updated road segment abnormal events and vehicle residual abnormalities, and generate real-time abnormality detection results.

2. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 1, characterized in that, The formation of the road segment matching result includes: S11. Map the continuous trajectory positions in the vehicle trajectory time series data to the road network described by the road topology data, determine the candidate road segments and corresponding positions within each trajectory position, and form a candidate road segment sequence consisting of the candidate road segments and corresponding positions within each road segment according to the trajectory time series. S12. Based on the changes in candidate road segments and the changes in positions within road segments of adjacent trajectory positions in the candidate road segment sequence, and combined with the directed connection relationships in the road topology data, determine the matching road segments and the corresponding positions within each candidate road segment to form a road segment matching sequence. S13. According to the temporal changes of the matched road segment and the corresponding position within the road segment in the road segment matching sequence, update the current matched road segment and the current position within the current road segment, and determine the road segment passage stage by the continuous changes in the position within the road segment, and form a road segment matching result with the current matched road segment, the current position within the current road segment and the road segment passage stage.

3. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 1, characterized in that, The formation of the candidate abnormal segment includes: S21. Determine the road segment passage stage corresponding to the vehicle based on the road segment matching result, determine the vehicle operation benchmark from the vehicle operation status time series data, perform deviation detection on the vehicle operation status within the road segment passage stage with the vehicle operation benchmark as a reference, obtain the abnormal deviation interval, and generate the vehicle abnormal event corresponding to the abnormal deviation interval. S22. Determine the corresponding road location based on the temporal position of each vehicle abnormal event in the vehicle operation status time-series data and the road segment matching result; determine the positional order of each road location along the road traffic direction based on the road topology data; and determine the vehicle traffic order based on the temporal position of each vehicle abnormal event, forming a vehicle abnormal event sequence according to the vehicle traffic order. S23. Merge adjacent vehicle abnormal events along the road location in the direction of travel along the vehicle abnormal event sequence, determine the continuous road range covered by the merged vehicle abnormal events, form candidate abnormal sections with the continuous road range, and retain the vehicle passage order of the vehicle abnormal events constituting the candidate abnormal sections. The abnormal deviation range is the range of deviation amplitude formed by the vehicle's operating state relative to the vehicle's operating reference along the abnormal deviation direction.

4. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 1, characterized in that, The formation of abnormal events on the aforementioned road section includes: S31. Define the road analysis range of the road operation status time series data according to the candidate abnormal segment, determine the road operation benchmark from the road operation status time series data, use the road operation benchmark as a reference to perform deviation detection on the road operation status within the road analysis range, obtain the abnormal road operation segment, and determine the road range and duration of the abnormal road operation segment. S32. Obtain the corresponding road location and occurrence time from the vehicle abnormal events in the candidate abnormal segment, match the road operation abnormal segment with the vehicle abnormal event according to the road location and occurrence time, filter the vehicle abnormal events whose road location falls within the road range and whose occurrence time coincides with the continuous period, and form a matching event sequence according to the vehicle passage order. S33. Obtain the anomaly type and anomaly deviation direction from the vehicle anomaly events in the matching event sequence, select vehicle anomaly events with the same anomaly type and the same anomaly deviation direction, perform common deviation processing on the selected vehicle anomaly events, determine the common deviation interval of multiple vehicles, and confirm the road operation anomaly segment with the common deviation interval of multiple vehicles to form a road segment anomaly event; the road segment anomaly event retains the road range and duration of the road operation anomaly segment.

5. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 1, characterized in that, The formation of the vehicle-road anomaly determination result includes: S41. Based on the common deviation interval of the multiple vehicles and the abnormal deviation interval of the corresponding vehicle abnormal event, determine the road attribution deviation interval of each corresponding vehicle abnormal event. S42. Divide the abnormal deviation interval of the corresponding vehicle abnormal event according to the road attribution deviation interval, and form the remaining deviation interval for the part that does not fall into the road attribution deviation interval. S43. The abnormal deviations within the remaining deviation interval are determined as vehicle residual anomalies corresponding to the vehicle abnormal events, and a vehicle-road anomaly determination result is formed based on the road attribution deviation interval and the road segment abnormal event, combined with the vehicle residual anomalies; the vehicle residual anomalies retain the vehicle identity, road location and occurrence time of the corresponding vehicle abnormal event.

6. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 1, characterized in that, The formation of the real-time anomaly detection result includes: S51. According to the time sequence of road segment abnormal events and vehicle residual abnormalities in the vehicle-road abnormality determination results, continuously match the abnormal events of adjacent road segments that are continuous and connected in the road travel direction, and continuously match the vehicle residual abnormalities that belong to the same vehicle and occur in connected time periods, to obtain the continuous matching results of road segment abnormal events and vehicle residual abnormalities. S52. Based on the continuous matching results of the road segment abnormal events and the continuous matching results of the vehicle residual abnormalities, the road segment abnormal events and vehicle residual abnormalities that form continuous matching are continued and updated respectively, and the road segment abnormal events and vehicle residual abnormalities that do not form continuous matching are terminated and updated, so as to obtain the updated road segment abnormal events and vehicle residual abnormalities. S53. Rematch the updated vehicle residual anomalies with the updated road segment anomalies according to the road location and occurrence time. Form a rematch result for vehicle residual anomalies and road segment anomalies whose road locations fall within the corresponding road range and whose occurrence time coincides with the corresponding duration time. Retain the road segment anomalies and vehicle residual anomalies that have not formed a rematch. Generate a real-time anomaly detection result based on the rematch result and the retained anomalies.

7. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 3, characterized in that, The formation of the vehicle abnormal event sequence includes: S221. Based on the temporal position of each vehicle abnormal event in the vehicle operating status time-series data, determine the corresponding road position from the road segment matching result, assign the road position to the corresponding directed travel path according to the directed connection relationship in the road topology data, and determine the position order of each road position along the directed travel path. S222. Based on the positional order and the temporal position of each vehicle abnormal event, determine the vehicle passage order corresponding to each vehicle abnormal event. S223. Arrange the vehicle abnormal events according to the vehicle passage order to obtain the vehicle abnormal event sequence.

8. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 4, characterized in that, The formation of the common deviation range of the multiple vehicles includes: S331. Align the selected abnormal deviation intervals of the vehicle abnormal events with zero deviation as the common starting point along the same deviation direction to obtain the aligned abnormal deviation intervals. S332. Using the endpoints of each aligned abnormal deviation interval, divide the continuous deviation range along the deviation direction into multiple deviation sub-intervals, and count the number of vehicles covered by each deviation sub-interval. S333: Merge adjacent deviation sub-intervals where the number of covered vehicles reaches a preset vehicle number threshold along the deviation direction to form a common deviation interval for multiple vehicles.

9. The real-time anomaly detection method based on vehicle-road-cloud collaboration according to claim 5, characterized in that, The formation of the road attribution bias interval includes: S411. For each vehicle abnormal event corresponding to the common deviation interval of the multiple vehicles, align the common deviation interval of the multiple vehicles with the abnormal deviation interval of the vehicle abnormal event along the same deviation direction with zero deviation as the common starting point to obtain the interval alignment result. S412. Determine the maximum deviation amplitude along the deviation direction for the two deviation intervals in the interval alignment result, and determine the smaller of the two values ​​as the road attribution deviation amplitude. S413. Starting from zero deviation, extend along the deviation direction to the road attribution deviation amplitude to form the road attribution deviation interval of the vehicle abnormal event.

10. A real-time anomaly detection system based on vehicle-road-cloud collaboration, characterized in that, The system includes: The road segment matching unit is used to acquire vehicle trajectory time-series data, vehicle operation status time-series data, road topology data, and road operation status time-series data, and to determine and update the road segment matching results based on the vehicle trajectory time-series data and road topology data. The candidate abnormal segment forming unit is used to extract vehicle abnormal events and corresponding abnormal deviation intervals from the vehicle operation status time series data according to the road segment matching results, determine the vehicle passage order of each vehicle abnormal event, and merge the vehicle abnormal events according to the vehicle passage order to form candidate abnormal segments. The road segment anomaly confirmation unit is used to perform spatiotemporal matching of vehicle anomaly events in the candidate anomaly segment and the road operation status time series data, extract the common deviation interval of multiple vehicles from the matched vehicle anomaly events, and confirm the anomaly of the matched road operation status accordingly to form a road segment anomaly event. The vehicle-road anomaly determination unit is used to decompose the abnormal deviation interval of the corresponding vehicle abnormal event into the common deviation interval of the multi-vehicles, obtain the road attribution deviation interval and vehicle residual anomaly, and combine the road segment abnormal event to form the vehicle-road anomaly determination result. The real-time anomaly detection unit is used to continuously update the road segment anomaly events and vehicle residual anomalies according to the time sequence of the road segment anomaly events and vehicle residual anomalies in the vehicle-road anomaly determination result, and re-match the updated road segment anomaly events and vehicle residual anomalies to generate a real-time anomaly detection result.

Images