An edge cooperative perception-based parameter anomaly identification method and system for internet of vehicles

Abstract

This invention discloses a method and system for identifying abnormal vehicle parameters in the Internet of Vehicles (IoV) based on edge collaborative perception, belonging to the technical field of anomaly identification. It can identify whether there are problems such as unclear identity transfer, broken time windows, spatial coordinate mismatch, or inconsistent lane topology mapping when the local judgment results of multiple edge nodes are spliced ​​across regions. Furthermore, it determines whether a continuous vehicle state chain can be formed based on the closure result of the cross-edge joint, thereby determining the vehicle parameter anomaly identification result and the source of the anomaly. This allows for the differentiation between normal cross-regional driving processes such as tunnel exit passage, ramp diversion, passage through elevated obstruction areas, and turning at complex intersections, and abnormal vehicle parameters caused by local perception differences, edge node splicing distortion, and continuous evolution. This reduces misjudgments of normal states and detects positioning drift, abnormal state reporting, or abnormal timestamps that are masked by reasonable local judgments.

Description

Technical Field

[0001] This invention relates to the field of anomaly recognition technology, specifically to a method and system for anomaly recognition of vehicle network parameters based on edge collaborative perception. Background Technology

[0002] Existing edge-coordinated perception-based methods for identifying anomalies in vehicle-to-everything (V2X) parameters typically involve deploying roadside units, cameras, millimeter-wave radar, lidar, and edge computing nodes within the target road area. This is combined with operational parameters uploaded by the vehicle itself, such as speed, acceleration, braking status, steering status, location coordinates, lane information, communication latency, and message timestamps, forming a data perception system involving the vehicle, roadside, and edge nodes. In practice, edge nodes first correlate vehicle-uploaded data, roadside perception data, and surrounding vehicle-coordinated perception data according to vehicle identification, collection time, and spatial location. Then, within the same time window, they perform consistency checks on the vehicle's self-reported status, roadside observation status, and surrounding vehicle feedback status. For example, they determine whether the vehicle's reported location matches the camera's identified location, whether the vehicle's reported speed matches the radar's measured speed, whether the vehicle's trajectory conforms to the road lane relationship, and whether there are significant delays, packet loss, or timestamp anomalies in communication parameters. Based on this, edge nodes determine whether the target vehicle exhibits location anomalies, speed anomalies, communication anomalies, status reporting anomalies, or suspected perception faults, according to the degree of deviation, duration, and anomaly type among the multi-source data. Because this type of method moves the anomaly identification process to the edge closer to the vehicle and road scene, and uses multi-source data to corroborate each other, it can reduce misjudgments caused by errors in a single data source compared to methods that rely solely on single-vehicle data or centralized cloud analysis, thereby improving the real-time performance and reliability of anomaly identification of vehicle network parameters.

[0003] However, the above approach still has shortcomings in scenarios where vehicles travel continuously and cross the coverage areas of multiple edge nodes. Specifically, existing methods focus more on whether the multi-source data within the coverage area of ​​a single edge node is consistent. That is, as long as the vehicle's self-reported data, the observation results of roadside equipment, and the perception results of surrounding vehicles can corroborate each other within the local time window, the vehicle state judgment result formed by that edge node is considered reliable. However, when a vehicle passes continuously through areas such as tunnel exits, ramp divergences, elevated road obstruction areas, and complex intersections, it may be perceived by multiple edge nodes sequentially or simultaneously. Different edge nodes form vehicle state judgment results based on their respective local perspectives. In this case, the judgment of each edge node may be reasonable within its local scope, but when multiple local judgment results are spliced ​​across regions, conflicts or breaks in the global state chain may occur due to reasons such as unclear vehicle identity, incomplete time window connection, inconsistent spatial coordinates or lane topology mapping, and different interpretations of road scene changes by different nodes. For example, one node might determine that the vehicle is still in the right lane of the tunnel, another node might determine that the vehicle has entered the ramp merging area, and yet another node might determine that the vehicle is at risk of position deviation. While each of these conclusions is supported by data individually, when pieced together, the system struggles to determine whether they represent a normal, continuous driving process, a localized perceptual difference, or a continuously evolving parameter anomaly. Therefore, existing edge-coordinated perception-based vehicle network parameter anomaly identification methods still lack a mechanism for cross-node continuity verification and splicing consistency analysis of local judgments formed by multiple edge nodes. This can easily lead to normal cross-regional driving states being misjudged as abnormal, or genuine anomalies being masked by multiple reasonable local judgments. Summary of the Invention

[0004] The purpose of this invention is to solve the problems mentioned in the background art above, and to propose a method and system for identifying abnormal vehicle network parameters based on edge collaborative perception.

[0005] A first aspect of this invention provides a method for identifying anomalies in vehicle network parameters based on edge collaborative perception, the method comprising: Obtain vehicle local state fragments formed by multiple edge nodes, and arrange the multiple vehicle local state fragments into a cross-edge state fragment chain based on vehicle identification, time sequence and road topology. Extract state seam information between two adjacent local state segments from the cross-edge state segment chain, and construct the cross-edge seam closure result based on the state seam information; Based on the closure result of the cross-edge joint, determine whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain; The abnormal vehicle parameter identification results are determined based on the judgment results of the continuous vehicle state chain, and the sources of abnormality are distinguished.

[0006] Optionally, the steps for arranging multiple vehicle local state fragments into a cross-edge state fragment chain based on vehicle identification, time sequence, and road topology are as follows: Obtain vehicle local state segments generated by each edge node within the corresponding coverage area. Each vehicle local state segment includes vehicle identifier, edge node identifier, segment start time, segment end time, segment start position, segment end position, segment start road unit, segment end road unit, and local state judgment result. Vehicle local state segments with the same vehicle identifier are grouped into the same segment set, and the vehicle local state segments in the same segment set are initially sorted according to the segment start time from earliest to latest to obtain the same vehicle state segment sequence. Based on the preset road topology map, read the road unit at the end of the previous segment and the road unit at the beginning of the next segment corresponding to any two adjacent local state segments of the same vehicle in the vehicle state segment sequence, and determine whether there is a direct connection relationship or a reachable connection relationship formed through the intermediate road unit between the road unit at the end of the previous segment and the road unit at the beginning of the next segment. If there is a direct or reachable connection between the road unit at the end of the previous segment and the road unit at the beginning of the next segment, then the sequential arrangement of two adjacent vehicle local state segments in the same vehicle state segment sequence is preserved. If there is no direct or reachable connection between the road unit at the end of the previous segment and the road unit at the beginning of the next segment, the next segment is removed from the current vehicle state segment sequence. Based on the road unit at the beginning of the next segment, a new candidate preceding segment is searched in the other vehicle local state segments corresponding to the same vehicle identifier. The road unit at the end of the previous segment has a direct or reachable connection with the road unit at the beginning of the next segment, and the end time of the previous segment is earlier than the start time of the next segment. Connect the next segment to the candidate preceding segment, and update the vehicle state segment sequence according to the connected order, until the ending road unit and the starting road unit of two adjacent vehicle local state segments in the vehicle state segment sequence both satisfy the direct connectivity relationship or the reachable connectivity relationship. The updated vehicle state fragment sequence is identified as the cross-edge state fragment chain corresponding to the vehicle identifier.

[0007] Optionally, the steps for extracting state seam information between two adjacent local state segments from the cross-edge state segment chain, and constructing the cross-edge seam closure result based on the state seam information are as follows: Extract state seam information between two adjacent local state segments from the cross-edge state segment chain, calculate the bidirectional state backfilling encounter index and the vehicle identity trail bearing index, add the bidirectional state backfilling encounter index and the vehicle identity trail bearing index to obtain the seam closure index, and use the seam closure index as state seam information to construct the cross-edge seam closure result.

[0008] Optionally, the calculation steps for the bidirectional state backfill encounter index are as follows: For adjacent local state segments in a cross-edge state segment chain, read the end position of the previous local state segment and the start position of the next local state segment. Based on the preset road topology map, determine whether there is a legal joint road corridor between the end position of the previous local state segment and the start position of the next local state segment. If there is no legal joint road corridor, the bidirectional state backfilling encounter index corresponding to the previous local state segment and the next local state segment is determined to be zero. If a legal joint road corridor exists, calculate the road passage length between the end position of the previous local state segment and the start position of the next local state segment along the legal joint road corridor. Read the last two state points from the end of the previous local state segment, calculate the road travel distance between the last two state points along the legal joint road corridor, and determine the road travel distance as the forward backfill length. Read the first two state points from the beginning of the next local state segment, calculate the road travel distance between the first two state points along the legal joint road corridor, and determine the road travel distance as the backfill length. Add the forward backfill length to the backward backfill length to obtain the bidirectional backfill coverage length; Calculate the absolute value of the difference between the road passage length and the two-way backfill coverage length, and determine the absolute value of the difference as the two-way backfill encounter difference; The difference in encounters between the two-way backfills is added to the road length to obtain the normalized benchmark value for encounters; Divide the bidirectional backfill encounter difference by the encounter normalization benchmark value to obtain the encounter deviation ratio; Subtract the encounter deviation ratio from 1 to obtain the bidirectional state backfill encounter index corresponding to the previous local state segment and the next local state segment.

[0009] Optionally, the calculation steps for the vehicle identity trail continuity index are as follows: For adjacent local state segments and adjacent local state segments in a cross-edge state segment chain, extract the surrounding vehicles that are observed at the same time as the target vehicle from the end of the previous local state segment to obtain the end-accompanying vehicle set. Extract the surrounding vehicles that are observed at the same time as the target vehicle from the beginning of the next local state segment to obtain the set of vehicles accompanying the beginning. Extract the surrounding vehicles with the same vehicle identifier in the end-accompanying vehicle set and the beginning-accompanying vehicle set to obtain the shared accompanying vehicle set; Count the number of vehicles in the shared accompanying vehicle set, and count the number of vehicles in the vehicle set obtained by merging the end accompanying vehicle set and the beginning accompanying vehicle set and removing duplicate vehicle identifiers. The number of vehicles in the shared accompanying vehicle set is divided by the number of vehicles in the vehicle set obtained after merging and removing duplicate vehicle identifiers, to obtain the accompanying vehicle retention value; For each shared accompanying vehicle in the shared accompanying vehicle set, calculate the relative distance of the shared accompanying vehicle to the target vehicle at the end of the previous local state segment and the relative distance of the shared accompanying vehicle to the target vehicle at the beginning of the next local state segment. Select any two different vehicles from the set of shared accompanying vehicles to form a unique vehicle pair, and count the total number of unique vehicle pairs. For each non-repeating vehicle pair, compare the relative distances of the two shared accompanying vehicles in the non-repeating vehicle pair to the target vehicle at the end of the previous local state segment, and compare the relative distances of the two shared accompanying vehicles in the non-repeating vehicle pair to the target vehicle at the beginning of the next local state segment. When the relative distance between the same non-repeating vehicle pair at the end of a previous local state segment is consistent with the relative distance between the pairs at the beginning of a subsequent local state segment, the non-repeating vehicle pair is determined as a vehicle pair whose distance order remains consistent. The number of vehicle pairs that maintain the same order of distance is counted, and the number of vehicle pairs that maintain the same order of distance is divided by the total number of non-repeating vehicle pairs to obtain the relative formation preservation value. Multiplying the accompanying vehicle retention value by the relative formation retention value yields the vehicle identity trail continuity index corresponding to the previous local state segment and the next local state segment.

[0010] Optionally, the step of determining whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain based on the closure result of the cross-edge joint is as follows: Read the seam closure index corresponding to each two adjacent local state segments in the cross-edge state segment chain, and use the seam closure index as the cross-edge seam closure result of the corresponding state seam; Compare the joint closure index corresponding to each state joint with the preset closure threshold respectively; If the joint closure index corresponding to any state joint is less than the preset closure threshold, then the state joint is marked as a broken joint, and it is determined that the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain. If the joint closure index corresponding to all state joints is not less than the preset closure threshold, then the cross-edge state segment chain is determined as a continuous vehicle state chain, and it is determined that the local judgment results of multiple edge nodes can form a continuous vehicle state chain.

[0011] Optionally, the steps for determining the vehicle parameter anomaly identification result based on the judgment result of the continuous vehicle state chain and distinguishing the source of the anomaly are as follows: If the local judgment results of multiple edge nodes can form a continuous vehicle state chain, then the local state judgment results corresponding to each local state segment of the vehicle in the continuous vehicle state chain will be used. If the local state judgment results corresponding to each local state segment of a vehicle in a continuous vehicle state chain are all normal, then the abnormal vehicle parameter identification result of the target vehicle is determined to be without abnormality. If there is a vehicle local state segment in the continuous vehicle state chain whose local state judgment result is an abnormal state, then extract the abnormal parameter type corresponding to the abnormal state and determine the abnormal parameter type as the vehicle parameter anomaly identification result of the target vehicle. If the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain, they are marked as a broken joint state joint, and the previous local state fragment and the next local state fragment corresponding to the broken joint are obtained. Obtain the bidirectional backfill encounter index and the vehicle identity trail bearing index corresponding to the fracture joint. If the bidirectional state backfilling encounter index is less than the preset backfilling threshold and the vehicle identity trail acceptance index is not less than the preset identity threshold, then the source of the anomaly will be determined as the vehicle parameter cross-edge motion acceptance anomaly, and the vehicle parameter anomaly identification result of the target vehicle will be determined as position parameter anomaly, speed parameter anomaly or timestamp parameter anomaly. If the vehicle identity trail continuity index is less than the preset identity threshold, the source of the anomaly will be determined as vehicle identity continuity anomaly, and the vehicle parameter anomaly identification result of the target vehicle will be determined as target aliasing anomaly or fragment splicing anomaly. If the bidirectional state backfill encounter index is not less than the preset backfill threshold and the vehicle identity trace acceptance index is not less than the preset identity threshold, but the joint closure index corresponding to the fracture joint is less than the preset closure threshold, then the source of the anomaly will be determined as the local judgment anomaly of the edge node, and the vehicle parameter anomaly identification result of the target vehicle will be determined as the edge node perception deviation anomaly. Output the results of vehicle parameter anomaly identification and the corresponding anomaly source.

[0012] A second aspect of this invention provides a vehicle network parameter anomaly identification system based on edge collaborative perception, the system comprising: The edge state module acquires vehicle local state fragments formed by multiple edge nodes, and arranges these fragments into a cross-edge state fragment chain based on vehicle identification, time sequence, and road topology. The seam closure module extracts state seam information between two adjacent local state segments from the cross-edge state segment chain, and constructs the cross-edge seam closure result based on the state seam information; The state chain module determines whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain based on the closure result of the cross-edge seam. The anomaly identification module determines the anomaly identification result of vehicle parameters based on the judgment result of the continuous vehicle state chain, and distinguishes the source of the anomaly.

[0013] The beneficial effects of this invention are: This invention proposes a method and system for identifying anomalies in vehicle network parameters based on edge collaborative perception. It identifies issues such as unclear identity transfer, broken time windows, mismatched spatial coordinates, or inconsistent lane topology mapping when multiple edge node local judgment results are spliced ​​across regions. Based on this, it determines whether multiple local judgment results can form a continuous vehicle state chain according to the closure result of the cross-edge seam, and determines the vehicle parameter anomaly identification result and its source. This method can distinguish normal cross-regional continuous driving processes such as tunnel exit passage, ramp diversion, passage through elevated obstruction areas, or turning at complex intersections from vehicle parameter anomalies caused by local perception differences, edge node splicing distortion, and continuous evolution. It avoids global state chain conflicts caused by directly assuming state credibility based solely on the consistency of multi-source data within a single edge node, reducing the misjudgment of normal cross-regional driving states as abnormal. Simultaneously, it can detect persistent parameter anomalies such as positioning drift, abnormal state reporting, or abnormal communication timestamps that are reasonably masked by multiple local judgments, thereby improving the accuracy, continuity, and interpretability of vehicle network parameter anomaly identification in scenarios with overlapping multi-edge coverage and cross-edge switching. Attached Figure Description

[0014] Figure 1 The flowchart illustrates a method for identifying abnormal vehicle network parameters based on edge collaborative perception, as provided in an embodiment of the present invention. Detailed Implementation

[0015] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0016] This invention provides a method for identifying anomalies in vehicle network parameters based on edge collaborative perception. See also... Figure 1 , Figure 1 A flowchart illustrating a method for identifying anomalies in vehicle network parameters based on edge collaborative perception, provided in an embodiment of the present invention. The method includes the following steps: S1, obtain vehicle local state fragments formed by multiple edge nodes respectively, and arrange the multiple vehicle local state fragments into a cross-edge state fragment chain according to vehicle identification, time sequence and road topology relationship; S2, extract the state seam information between two adjacent local state segments from the cross-edge state segment chain, and construct the cross-edge seam closure result based on the state seam information; S3, based on the closure result of the cross-edge joint, determine whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain; S4 determines the abnormal vehicle parameter identification result based on the judgment result of the continuous vehicle state chain, and distinguishes the source of the abnormality.

[0017] This invention provides a method for identifying anomalies in vehicle network parameters based on edge collaborative perception. The method first acquires local vehicle state segments formed by multiple edge nodes, then arranges these segments into a cross-edge state segment chain based on vehicle identification, temporal sequence, and road topology. This ensures that segmented judgment results formed by vehicles within different edge node coverage areas are no longer used in isolation, but are organized into a holistic state chain reflecting the continuous driving process of the vehicle. State seam information between adjacent local state segments is extracted from the cross-edge state segment chain, and a cross-edge seam closure result is constructed based on this information. This verifies whether the end state of the previous edge node and the start state of the next edge node satisfy temporal continuity, spatial reachability, lane topology connectivity, and state semantic continuity. This identifies whether unclear identity continuity, broken time windows, or spatial misalignment exist when multiple edge node local judgment results are spliced ​​across regions. Issues such as mismatched labels or inconsistent lane topology mapping are addressed. Based on this, the system determines whether multiple local judgment results can form a continuous vehicle state chain according to the closure results of cross-edge seams. This helps identify the abnormal vehicle parameters and their sources. The system can distinguish normal cross-regional continuous driving processes, such as tunnel exit passage, ramp diversion, passage through elevated obstruction areas, or turning at complex intersections, from local perception differences, edge node splicing distortion, and continuously evolving vehicle parameter anomalies. This avoids global state chain conflicts caused by directly determining the reliability of a state based solely on the consistency of multi-source data within a single edge node, reducing the possibility of normal cross-regional driving states being misjudged as abnormal. It can also detect persistent parameter anomalies such as positioning drift, abnormal state reporting, or abnormal communication timestamps that are reasonably masked by multiple local judgments. This improves the accuracy, continuity, and interpretability of vehicle network parameter anomaly identification in scenarios with overlapping multi-edge coverage and cross-edge switching.

[0018] In one embodiment, S1, the step of obtaining vehicle local state fragments formed by multiple edge nodes and arranging the multiple vehicle local state fragments into a cross-edge state fragment chain based on vehicle identification, time sequence, and road topology is as follows: Obtain vehicle local state segments generated by each edge node within the corresponding coverage area. Each vehicle local state segment includes vehicle identifier, edge node identifier, segment start time, segment end time, segment start position, segment end position, segment start road unit, segment end road unit, and local state judgment result. Vehicle local state segments with the same vehicle identifier are grouped into the same segment set, and the vehicle local state segments in the same segment set are initially sorted according to the segment start time from earliest to latest to obtain the same vehicle state segment sequence. Based on the preset road topology map, read the road unit at the end of the previous segment and the road unit at the beginning of the next segment corresponding to any two adjacent local state segments of the same vehicle in the vehicle state segment sequence, and determine whether there is a direct connection relationship or a reachable connection relationship formed through the intermediate road unit between the road unit at the end of the previous segment and the road unit at the beginning of the next segment. If there is a direct or reachable connection between the road unit at the end of the previous segment and the road unit at the beginning of the next segment, then the sequential arrangement of two adjacent vehicle local state segments in the same vehicle state segment sequence is preserved. If there is no direct or reachable connection between the road unit at the end of the previous segment and the road unit at the beginning of the next segment, the next segment is removed from the current vehicle state segment sequence. Based on the road unit at the beginning of the next segment, a new candidate preceding segment is searched in the other vehicle local state segments corresponding to the same vehicle identifier. The road unit at the end of the previous segment has a direct or reachable connection with the road unit at the beginning of the next segment, and the end time of the previous segment is earlier than the start time of the next segment. Connect the next segment to the candidate preceding segment, and update the vehicle state segment sequence according to the connected order, until the ending road unit and the starting road unit of two adjacent vehicle local state segments in the vehicle state segment sequence both satisfy the direct connectivity relationship or the reachable connectivity relationship. The updated vehicle state fragment sequence is identified as the cross-edge state fragment chain corresponding to the vehicle identifier.

[0019] It should be noted that after each edge node completes local perception within its coverage area, it will generate corresponding vehicle local state fragments. For example, a fragment records the starting position, ending position, starting road unit, ending road unit, and local judgment result of vehicle A within the coverage area of ​​edge node E1 from the start time to the end time. First, local state fragments belonging to the same vehicle are grouped into the same fragment set according to the vehicle identifier to avoid state fragments of different vehicles from being mixed into the same chain. Then, the fragments are initially sorted from early to late according to their start time to form a sequence of state fragments of the same vehicle. Since chronological ordering alone may result in spatial inconsistencies, for example, vehicle A's previous segment might end in the right lane of a tunnel, while its next segment might begin at an entrance to an auxiliary road that is not connected to that lane. Therefore, a pre-defined road topology map is introduced to verify the connectivity of the "ending road unit of the previous segment" and the "starting road unit of the next segment" for adjacent segments. If the two segments are directly connected in the road topology or can be connected through an intermediate road unit, it means that the two segments can be connected in spatial path, and their sequential order can be retained. If the two segments are not directly connected or reachable, it means that the sequential relationship obtained by chronological ordering alone is unreasonable. The next segment is removed from the current sequence, and a more suitable candidate preceding segment is searched among other segments corresponding to the same vehicle identifier. This candidate preceding segment must be earlier than the next segment in time and be able to connect to the starting road unit of the next segment in the road topology. Then, the next segment is connected to the candidate preceding segment, and the sequence is updated. By continuously processing adjacent segments using this "initial temporal arrangement—road topology verification—reconnection of disconnected segments" process, each pair of adjacent segments in the same vehicle state segment sequence ultimately satisfies both temporal sequence and road reachability. For example, when vehicle A passes through the tunnel node, tunnel exit node, and ramp node in sequence, it is not directly spliced ​​together simply because the reported times of the three nodes are close. Instead, it is confirmed whether the "right lane of the tunnel" can connect to the "exit diversion area," and whether the "exit diversion area" can connect to the "ramp entrance." Only when these road units can be continuously connected is it determined as a cross-edge state segment chain for vehicle A. Through this processing, it is possible to avoid erroneously splicing local state segments of the same vehicle in different edge nodes onto discontinuous road areas, providing a reliable basis for subsequent extraction of state seam information and determination of whether the vehicle state chain is continuous.

[0020] In one embodiment, S2, the step of extracting state seam information between two adjacent local state segments from the cross-edge state segment chain, and constructing the cross-edge seam closure result based on the state seam information is as follows: Extract state seam information between two adjacent local state segments from the cross-edge state segment chain, calculate the bidirectional state backfilling encounter index and the vehicle identity trail bearing index, add the bidirectional state backfilling encounter index and the vehicle identity trail bearing index to obtain the seam closure index, and use the seam closure index as state seam information to construct the cross-edge seam closure result.

[0021] In one implementation, the calculation steps for the bidirectional state backfilling encounter index are as follows: For adjacent local state segments in a cross-edge state segment chain, read the end position of the previous local state segment and the start position of the next local state segment. Based on the preset road topology map, determine whether there is a legal joint road corridor between the end position of the previous local state segment and the start position of the next local state segment. If there is no legal joint road corridor, the bidirectional state backfilling encounter index corresponding to the previous local state segment and the next local state segment is determined to be zero. If a legal joint road corridor exists, calculate the road passage length between the end position of the previous local state segment and the start position of the next local state segment along the legal joint road corridor. Read the last two state points from the end of the previous local state segment, calculate the road travel distance between the last two state points along the legal joint road corridor, and determine the road travel distance as the forward backfill length. Read the first two state points from the beginning of the next local state segment, calculate the road travel distance between the first two state points along the legal joint road corridor, and determine the road travel distance as the backfill length. Add the forward backfill length to the backward backfill length to obtain the bidirectional backfill coverage length; Calculate the absolute value of the difference between the road passage length and the two-way backfill coverage length, and determine the absolute value of the difference as the two-way backfill encounter difference; The difference in encounters between the two-way backfills is added to the road length to obtain the normalized benchmark value for encounters; Divide the bidirectional backfill encounter difference by the encounter normalization benchmark value to obtain the encounter deviation ratio; Subtract the encounter deviation ratio from 1 to obtain the bidirectional state backfill encounter index corresponding to the previous local state segment and the next local state segment.

[0022] It should be noted that, to illustrate the above steps, for example, if vehicle A's previous segment ends in the right lane of the tunnel exit and the next segment begins in the ramp divergence area, and the preset road topology map shows that there is a legal joint road corridor between the two segments with a road length of 20 meters; if the last two state points at the end of the previous segment show that the vehicle has advanced 9 meters towards the ramp, and the first two state points at the beginning of the next segment can cover 10 meters by pushing back in the opposite direction, then the bidirectional backfill coverage length is 19 meters, only 1 meter different from 20 meters, indicating that the two segments can basically meet naturally at the joint, and the bidirectional state backfill meeting index is high; conversely, if the forward backfill length is only 2 meters and the backward backfill length is only 3 meters, totaling only 5 meters, which is significantly different from the 20-meter joint distance, it indicates that the movement trends at the end of the previous segment and the beginning of the next segment are difficult to jointly fill the joint area. Although the two segments may be adjacent in time, the state connection is insufficient, and the bidirectional state backfill meeting index is low. This step avoids directly judging whether the joint is closed by simply using the distance or time difference between the endpoints of two segments, through the method of "road topology accessibility judgment - backfilling of movement on both sides of the boundary - matching the backfill coverage length with the length of the joint road". This makes the joint closure judgment more reflective of the real continuous driving relationship of vehicles in the cross-edge area.

[0023] The bidirectional state backfill encounter index is used to measure the degree to which two adjacent local state segments can be naturally connected at the joint. Specifically, it doesn't simply look at how far apart the end position of the previous segment and the beginning position of the next segment are. Instead, it considers whether the two segments can jointly cover the intermediate road corridor after the previous segment continues to extend backward according to the movement trend of its end and the next segment pushes back forward according to the movement trend of its beginning. If they can basically cover it, it means that although the two segments come from different edge nodes, their vehicle states can be connected in terms of movement direction, road position, and travel path. Therefore, the larger the bidirectional state backfilling encounter index, the easier it is for the joint to close, and the larger the joint closure index is. The more likely the local judgment results of multiple edge nodes are to form a continuous vehicle state chain. Conversely, if the index is smaller, it means that the previous segment cannot extend backward to the position of the next segment, and the next segment cannot push back forward to connect to the position of the previous segment. There is a clear break between the two. In this case, even if the vehicle identification is the same and the time is adjacent, there may be segment splicing errors, edge node recognition deviations, or abnormal vehicle parameters. Therefore, the possibility of multiple local judgment results forming a continuous vehicle state chain is smaller. For example, if the previous edge node determines that the vehicle ends at the tunnel exit and the next edge node determines that the vehicle begins at the ramp entrance, if the length of the backward extension of the previous segment and the length of the forward push of the next segment can basically cover the road distance between the tunnel exit and the ramp entrance, it means that the vehicle is likely to enter the ramp normally and continuously. If the sum of the extension lengths on both sides still cannot clearly cover the middle road distance, it means that there is a lack of continuous connection between the two judgment results, and they cannot be directly identified as the same continuous vehicle state chain.

[0024] The advantage of calculating the bidirectional backfill encounter index using the above method is that it doesn't simply rely on the straight-line distance, time interval, or positional difference between the endpoints of two segments to determine whether the joint is closed. Instead, it first confirms whether there is a legal joint road corridor between the two segments that conforms to the road topology. Then, it uses the nearest motion state at the end of the previous segment and the nearest motion state at the beginning of the next segment to perform bidirectional backfilling. Finally, it determines whether the backfill lengths on both sides can jointly cover the joint distance in the middle. Therefore, it is more in line with the actual law that vehicles can only travel continuously along road structures such as lanes, ramps, and intersections in real roads. This method can avoid incorrectly judging two segments that are very close in straight line but inaccessible on the road as continuous, and it can also avoid incorrectly judging two segments that are far apart on the road but are naturally connected along ramps or divergence zones as broken. At the same time, by using "forward extension" and "backward push" together in the judgment, it can simultaneously use the end information of the previous edge node and the beginning information of the next edge node, rather than unilaterally relying on the result of a single edge node, thus more accurately reflecting whether two local state segments have a real motion connection relationship at the joint. The bidirectional state backfilling encounter index calculated in this way better reflects the continuity between cross-edge state segments, and can provide a more reliable basis for subsequent judgments on whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain.

[0025] In one implementation method, the calculation steps for the vehicle identity trail continuity index are as follows: For adjacent local state segments and adjacent local state segments in a cross-edge state segment chain, extract the surrounding vehicles that are observed at the same time as the target vehicle from the end of the previous local state segment to obtain the end-accompanying vehicle set. Extract the surrounding vehicles that are observed at the same time as the target vehicle from the beginning of the next local state segment to obtain the set of vehicles accompanying the beginning. Extract the surrounding vehicles with the same vehicle identifier in the end-accompanying vehicle set and the beginning-accompanying vehicle set to obtain the shared accompanying vehicle set; Count the number of vehicles in the shared accompanying vehicle set, and count the number of vehicles in the vehicle set obtained by merging the end accompanying vehicle set and the beginning accompanying vehicle set and removing duplicate vehicle identifiers. The number of vehicles in the shared accompanying vehicle set is divided by the number of vehicles in the vehicle set obtained after merging and removing duplicate vehicle identifiers, to obtain the accompanying vehicle retention value; For each shared accompanying vehicle in the shared accompanying vehicle set, calculate the relative distance of the shared accompanying vehicle to the target vehicle at the end of the previous local state segment and the relative distance of the shared accompanying vehicle to the target vehicle at the beginning of the next local state segment. Select any two different vehicles from the set of shared accompanying vehicles to form a unique vehicle pair, and count the total number of unique vehicle pairs. For each non-repeating vehicle pair, compare the relative distances of the two shared accompanying vehicles in the non-repeating vehicle pair to the target vehicle at the end of the previous local state segment, and compare the relative distances of the two shared accompanying vehicles in the non-repeating vehicle pair to the target vehicle at the beginning of the next local state segment. When the relative distance between the same non-repeating vehicle pair at the end of a previous local state segment is consistent with the relative distance between the pairs at the beginning of a subsequent local state segment, the non-repeating vehicle pair is determined as a vehicle pair whose distance order remains consistent. The number of vehicle pairs that maintain the same order of distance is counted, and the number of vehicle pairs that maintain the same order of distance is divided by the total number of non-repeating vehicle pairs to obtain the relative formation preservation value. Multiplying the accompanying vehicle retention value by the relative formation retention value yields the vehicle identity trail continuity index corresponding to the previous local state segment and the next local state segment.

[0026] It should be noted that the data involved in the calculation of the vehicle identity trail continuity index mainly comes from the observation results of the target vehicle and its surrounding vehicles, which are synchronously saved by each edge node when generating local state segments of the vehicle. Specifically, each edge node identifies the target vehicle and its surrounding vehicles within its coverage area through vehicle-reported information received by roadside cameras, millimeter-wave radar, lidar, roadside units, and collaborative perception information of surrounding vehicles. For each identified vehicle, the edge node records the vehicle identifier, collection time, spatial location, road unit, and corresponding edge node identifier. When forming a previous local state segment, a preset number of state points at the end of the segment or an observation window near the end time of the segment are selected. Vehicles within this window that are simultaneously observed by the same edge node or adjacent edge nodes and are located within a preset neighborhood of the target vehicle are identified as the trailing vehicles, and their vehicle identifiers and positions relative to the target vehicle are recorded. When forming a subsequent local state segment, a preset number of state points at the beginning of the segment or an observation window near the beginning time of the segment are selected, and the trailing vehicles at the beginning are identified in the same way, and their vehicle identifiers and positions relative to the target vehicle are recorded. Subsequently, the set of accompanying vehicles at the end and the set of accompanying vehicles at the beginning are matched based on vehicle identification. Vehicles with the same identification are considered shared accompanying vehicles. The relative distance between a shared accompanying vehicle and the target vehicle is calculated based on their position coordinates at the same observation time or within the same observation window. For example, the relative position is obtained by subtracting the target vehicle's position coordinates from the shared accompanying vehicle's position coordinates, and then the road travel distance or planar distance between them is calculated as the relative distance. Therefore, the set of accompanying vehicles at the end, the set of accompanying vehicles at the beginning, the set of shared accompanying vehicles, the number of shared accompanying vehicles, the number of accompanying vehicles after merging and deduplication, and the relative distance before and after the seam, all required for the vehicle identity wake continuity index, can be directly obtained from the vehicle observation data collected and structured at the edge nodes.

[0027] The vehicle identity trail continuity index measures the degree to which two adjacent local state segments belong to the continuous state of the same target vehicle before and after a seam. Here, "identity trail" does not simply refer to vehicle identifiers, but rather to the relative relationships between the target vehicle and its surrounding vehicles at the end of one segment and the beginning of the next. For example, it considers which vehicles are around the target vehicle before the seam, which are closer to the target vehicle, and which are farther away. If most of the accompanying vehicles are preserved before and after the seam, and the order of their proximity relative to the target vehicle is generally consistent, it indicates that the relative positional relationship of the target vehicle in the traffic flow has not been significantly broken. The two local state segments are more likely to be continuous observations of the same vehicle. Therefore, the larger the vehicle identity trail continuity index, the more reliable the identity continuity. The seam closure index... The larger the index, the greater the likelihood that the local judgment results of multiple edge nodes will form a continuous vehicle state chain. Conversely, if the target vehicle is surrounded by vehicles B, C, and D before the seam, but becomes vehicles E, F, and G after the seam, or if vehicle B is closer to the target vehicle than vehicle C before the seam, but the proximity relationship is significantly reversed after the seam, it indicates that although the same vehicle identifier may have been used in the preceding and following segments, the traffic relationship around the target vehicle is no longer continuous. There may be issues such as incorrect vehicle identity continuation, target overlap, or edge node re-identification deviation. In this case, the vehicle identity trail continuation index is small, the seam closure index should also be reduced, and the likelihood that multiple local judgment results will form a continuous vehicle state chain is smaller. For example, when vehicle A enters the ramp from the tunnel exit, if the end of the previous node observes that A is in front of B and behind C, and the beginning of the next node still observes that A is in front of B and behind C, and the distance between the two remains basically unchanged, then it means that the vehicle A identified by the next node is most likely the same vehicle A in the previous node. If the vehicles around the so-called vehicle A in the next node change completely, then it cannot be concluded that the two segments can be spliced ​​continuously just because the vehicle identification is the same.

[0028] It should be noted that the advantage of calculating the vehicle identity trail continuity index using the above method lies in the fact that by incorporating the retention of accompanying vehicles around the target vehicle and whether the order of these accompanying vehicles relative to the target vehicle remains consistent, it is equivalent to using the target vehicle's "relative position trace" in the traffic flow to verify identity continuity. The benefit of this is that even if there are slight differences in the coordinate system, observation angle, or local recognition results of two consecutive edge nodes, as long as the accompanying vehicles around the target vehicle are basically consistent and the relative formation does not change significantly, it can be determined that the two segments have a high degree of identity continuity credibility. Conversely, if the vehicle identification is the same but the relationship between the surrounding vehicles suddenly changes, it can promptly detect potential target aliasing, incorrect splicing, or re-identification deviations. Therefore, this calculation method is more reliable than simply comparing vehicle IDs, location distances, or temporal continuity. It can determine whether consecutive local state segments belong to the continuous state of the same vehicle from the perspective of whether the surrounding traffic relationship continues, providing a more stable identity basis for seam closure judgment.

[0029] In one embodiment, S3, the step of determining whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain based on the closure result of the cross-edge seam is as follows: Read the seam closure index corresponding to each two adjacent local state segments in the cross-edge state segment chain, and use the seam closure index as the cross-edge seam closure result of the corresponding state seam; Compare the joint closure index corresponding to each state joint with the preset closure threshold respectively; If the joint closure index corresponding to any state joint is less than the preset closure threshold, then the state joint is marked as a broken joint, and it is determined that the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain. If the joint closure index corresponding to all state joints is not less than the preset closure threshold, then the cross-edge state segment chain is determined as a continuous vehicle state chain, and it is determined that the local judgment results of multiple edge nodes can form a continuous vehicle state chain.

[0030] It should be noted that the purpose of this step is to further transform the quantitative results of whether each "seam is closed" obtained in the previous step into a judgment result on whether the entire cross-edge state segment chain is continuous. Specifically, the process first reads the seam closure index corresponding to each pair of adjacent local state segments in the cross-edge state segment chain. For example, there is a seam closure index between segment 1 and segment 2, and another between segment 2 and segment 3. These seam closure indices are then used as the cross-edge seam closure results for the corresponding state seams. Subsequently, each seam closure index is compared with a preset closure threshold. This preset closure threshold represents the minimum degree of closure required for two adjacent local state segments to be considered as naturally connected. For example, the threshold can be set to 0.7. When the seam closure index of a certain state seam is lower than this threshold, it indicates that the bidirectional state backfilling encounter relationship, vehicle identity trail connection relationship, etc. at the seam are insufficient to support the natural connection of the two segments. The state seam is then marked as a broken seam, and it is determined that the local judgment results of multiple edge nodes cannot form a complete continuous vehicle state chain. This is because the cross-edge state segment chain is essentially formed by connecting multiple local state segments sequentially. If any adjacent segment breaks, it means that the entire chain cannot pass continuously at that position, and all local judgment results cannot be directly regarded as the continuous driving process of the same vehicle. Conversely, if the joint closure index corresponding to all state joints in the cross-edge state segment chain is not less than the preset closure threshold, it means that the minimum closure requirement has been met between each pair of adjacent local state segments. The previous segment can form a continuous connection with the next segment in terms of road traffic relationship, motion extension relationship and identity inheritance relationship. Then, the cross-edge state segment chain is determined as a continuous vehicle state chain, and it is determined that the local judgment results of multiple edge nodes can form a continuous vehicle state chain. For example, vehicle A passes through the tunnel node, the tunnel exit node, and the ramp node in sequence, resulting in two state joints: the joint closure index between the tunnel segment and the tunnel exit segment is 0.82, and the joint closure index between the tunnel exit segment and the ramp segment is 0.79. If the preset closure threshold is 0.7, then both joints meet the requirements, and it can be considered that the judgment results of the three edge nodes can be connected to form a continuous driving state chain of vehicle A. However, if the closure index of one of the joints is only 0.45, it indicates that there is a significant break at the joint, which may be caused by an incorrect identification of the target vehicle, unreasonable road topology connection, or local judgment deviation of a certain edge node. In this case, these local segments cannot be considered as a continuous vehicle state chain.

[0031] In one embodiment, S4, the step of determining the vehicle parameter anomaly identification result based on the judgment result of the continuous vehicle state chain and distinguishing the source of the anomaly is as follows: If the local judgment results of multiple edge nodes can form a continuous vehicle state chain, then the local state judgment results corresponding to each local state segment of the vehicle in the continuous vehicle state chain will be used. If the local state judgment results corresponding to each local state segment of a vehicle in a continuous vehicle state chain are all normal, then the abnormal vehicle parameter identification result of the target vehicle is determined to be without abnormality. If there is a vehicle local state segment in the continuous vehicle state chain whose local state judgment result is an abnormal state, then extract the abnormal parameter type corresponding to the abnormal state and determine the abnormal parameter type as the vehicle parameter anomaly identification result of the target vehicle. If the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain, they are marked as a broken joint state joint, and the previous local state fragment and the next local state fragment corresponding to the broken joint are obtained. Obtain the bidirectional backfill encounter index and the vehicle identity trail bearing index corresponding to the fracture joint. If the bidirectional state backfilling encounter index is less than the preset backfilling threshold and the vehicle identity trail acceptance index is not less than the preset identity threshold, then the source of the anomaly will be determined as the vehicle parameter cross-edge motion acceptance anomaly, and the vehicle parameter anomaly identification result of the target vehicle will be determined as position parameter anomaly, speed parameter anomaly or timestamp parameter anomaly. If the vehicle identity trail continuity index is less than the preset identity threshold, the source of the anomaly will be determined as vehicle identity continuity anomaly, and the vehicle parameter anomaly identification result of the target vehicle will be determined as target aliasing anomaly or fragment splicing anomaly. If the bidirectional state backfill encounter index is not less than the preset backfill threshold and the vehicle identity trace acceptance index is not less than the preset identity threshold, but the joint closure index corresponding to the fracture joint is less than the preset closure threshold, then the source of the anomaly will be determined as the local judgment anomaly of the edge node, and the vehicle parameter anomaly identification result of the target vehicle will be determined as the edge node perception deviation anomaly. Output the results of vehicle parameter anomaly identification and the corresponding anomaly source.

[0032] It should be noted that this step is used to further output whether the vehicle is abnormal and the source of the abnormality after determining whether the state chain is continuous. Specifically, if the local judgment results of multiple edge nodes can form a continuous vehicle state chain, it means that the preceding and following segments can be successfully connected in terms of road topology, motion joints, and vehicle identity. At this time, the local state judgment results of each vehicle local state segment are directly read. If all segments are in a normal state, it is determined that the target vehicle is not abnormal. If there are abnormal state segments, the abnormal parameter type corresponding to the segment is extracted, such as abnormal position, abnormal speed, abnormal lane recognition, or abnormal timestamp, and used as the vehicle parameter abnormality identification result. If the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain, it indicates that the problem mainly occurs at the fracture joint. Obtain the bidirectional state backfilling encounter index and vehicle identity trail continuity index corresponding to the fracture joint: If the bidirectional state backfilling encounter index is low but the vehicle identity trail continuity index meets the standard, it means that the preceding and following segments are likely still the same vehicle, but the motion state cannot be connected naturally. Therefore, the source of the anomaly is the abnormal motion continuity of vehicle parameters across the edge, which can be identified as anomaly of position parameters, speed parameters, or timestamp parameters. If the vehicle identity trail continuity index is low, it means that the preceding and following segments may not be continuous observation results of the same vehicle. Therefore, the source of the anomaly is the abnormal motion continuity of vehicle identity, which can be identified as anomaly of target aliasing or segment splicing. If both the bidirectional state backfilling encounter index and the vehicle identity trail continuity index meet the standard, but the joint closure index is still lower than the preset closure threshold, it indicates that the motion continuity and identity continuity problems themselves are not obvious. It is more likely caused by a local perception or judgment deviation of a certain edge node. Therefore, it is determined to be a local judgment anomaly of the edge node. This step allows us to distinguish between "abnormal vehicle parameters", "abnormal vehicle identity splicing", and "abnormal edge node perception deviation", rather than simply giving a conclusion of abnormality or normality.

[0033] Based on the same inventive concept, this invention also provides a vehicle network parameter anomaly identification system based on edge collaborative perception, including: The edge state module acquires vehicle local state fragments formed by multiple edge nodes, and arranges these fragments into a cross-edge state fragment chain based on vehicle identification, time sequence, and road topology. The seam closure module extracts state seam information between two adjacent local state segments from the cross-edge state segment chain, and constructs the cross-edge seam closure result based on the state seam information; The state chain module determines whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain based on the closure result of the cross-edge seam. The anomaly identification module determines the anomaly identification result of vehicle parameters based on the judgment result of the continuous vehicle state chain, and distinguishes the source of the anomaly.

[0034] This invention provides a vehicle network parameter anomaly identification system based on edge collaborative perception. It identifies issues such as unclear identity transfer, broken time windows, spatial coordinate mismatch, or inconsistent lane topology mapping when multiple edge node local judgment results are spliced ​​across regions. Based on this, it determines whether multiple local judgment results can form a continuous vehicle state chain according to the cross-edge seam closure result, and thereby identifies the vehicle parameter anomaly identification result and its source. This system can distinguish normal cross-regional continuous driving processes such as tunnel exit passage, ramp diversion, passage through elevated obstruction areas, or turning at complex intersections from vehicle parameter anomalies caused by local perception differences, edge node splicing distortion, and continuous evolution. It avoids global state chain conflicts caused by directly determining state credibility based solely on the consistency of multi-source data within a single edge node, reducing the misjudgment of normal cross-regional driving states as abnormal. Simultaneously, it can detect persistent parameter anomalies such as positioning drift, abnormal state reporting, or abnormal communication timestamps that are reasonably masked by multiple local judgments. This improves the accuracy, continuity, and interpretability of vehicle network parameter anomaly identification in scenarios with overlapping multi-edge coverage and cross-edge switching.

[0035] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention should still fall within the scope of the claims of the present invention.

Claims

1. A method for identifying anomalies in vehicle network parameters based on edge collaborative perception, characterized in that, Includes the following steps: Obtain vehicle local state fragments formed by multiple edge nodes, and arrange the multiple vehicle local state fragments into a cross-edge state fragment chain based on vehicle identification, time sequence and road topology. Extract state seam information between two adjacent local state segments from the cross-edge state segment chain, and construct the cross-edge seam closure result based on the state seam information; Based on the closure result of the cross-edge joint, determine whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain; The abnormal vehicle parameter identification results are determined based on the judgment results of the continuous vehicle state chain, and the sources of abnormality are distinguished.

2. The method for identifying abnormal vehicle network parameters based on edge collaborative perception according to claim 1, characterized in that, The steps for arranging multiple vehicle local state fragments into a cross-edge state fragment chain based on vehicle identification, time sequence, and road topology are as follows: Obtain vehicle local state segments generated by each edge node within the corresponding coverage area. Each vehicle local state segment includes vehicle identifier, edge node identifier, segment start time, segment end time, segment start position, segment end position, segment start road unit, segment end road unit, and local state judgment result. Vehicle local state segments with the same vehicle identifier are grouped into the same segment set, and the vehicle local state segments in the same segment set are initially sorted according to the segment start time from earliest to latest to obtain the same vehicle state segment sequence. Based on the preset road topology map, read the road unit at the end of the previous segment and the road unit at the beginning of the next segment corresponding to any two adjacent local state segments of the same vehicle in the vehicle state segment sequence, and determine whether there is a direct connection relationship or a reachable connection relationship formed through the intermediate road unit between the road unit at the end of the previous segment and the road unit at the beginning of the next segment. If there is a direct or reachable connection between the road unit at the end of the previous segment and the road unit at the beginning of the next segment, then the sequential arrangement of two adjacent vehicle local state segments in the same vehicle state segment sequence is preserved. If there is no direct or reachable connection between the road unit at the end of the previous segment and the road unit at the beginning of the next segment, the next segment is removed from the current vehicle state segment sequence. Based on the road unit at the beginning of the next segment, a new candidate preceding segment is searched in the other vehicle local state segments corresponding to the same vehicle identifier. The road unit at the end of the previous segment has a direct or reachable connection with the road unit at the beginning of the next segment, and the end time of the previous segment is earlier than the start time of the next segment. Connect the next segment to the candidate preceding segment, and update the vehicle state segment sequence according to the connected order, until the ending road unit and the starting road unit of two adjacent vehicle local state segments in the vehicle state segment sequence both satisfy the direct connectivity relationship or the reachable connectivity relationship. The updated vehicle state fragment sequence is identified as the cross-edge state fragment chain corresponding to the vehicle identifier.

3. The method for identifying abnormal vehicle network parameters based on edge collaborative perception according to claim 1, characterized in that, The steps for extracting state seam information between two adjacent local state segments from the cross-edge state segment chain, and constructing the cross-edge seam closure result based on the state seam information are as follows: Extract state seam information between two adjacent local state segments from the cross-edge state segment chain, calculate the bidirectional state backfilling encounter index and the vehicle identity trail bearing index, add the bidirectional state backfilling encounter index and the vehicle identity trail bearing index to obtain the seam closure index, and use the seam closure index as state seam information to construct the cross-edge seam closure result.

4. The method for identifying abnormal vehicle network parameters based on edge collaborative perception according to claim 3, characterized in that, The steps for calculating the encounter index of bidirectional state backfilling are as follows: For adjacent local state segments in a cross-edge state segment chain, read the end position of the previous local state segment and the start position of the next local state segment. Based on the preset road topology map, determine whether there is a legal joint road corridor between the end position of the previous local state segment and the start position of the next local state segment. If there is no legal joint road corridor, the bidirectional state backfilling encounter index corresponding to the previous local state segment and the next local state segment is determined to be zero. If a legal joint road corridor exists, calculate the road passage length between the end position of the previous local state segment and the start position of the next local state segment along the legal joint road corridor. Read the last two state points from the end of the previous local state segment, calculate the road travel distance between the last two state points along the legal joint road corridor, and determine the road travel distance as the forward backfill length. Read the first two state points from the beginning of the next local state segment, calculate the road travel distance between the first two state points along the legal joint road corridor, and determine the road travel distance as the backfill length. Add the forward backfill length to the backward backfill length to obtain the bidirectional backfill coverage length; Calculate the absolute value of the difference between the road passage length and the two-way backfill coverage length, and determine the absolute value of the difference as the two-way backfill encounter difference; The difference in encounters between the two-way backfills is added to the road length to obtain the normalized benchmark value for encounters; Divide the bidirectional backfill encounter difference by the encounter normalization benchmark value to obtain the encounter deviation ratio; Subtract the encounter deviation ratio from 1 to obtain the bidirectional state backfill encounter index corresponding to the previous local state segment and the next local state segment.

5. The method for identifying abnormal vehicle network parameters based on edge collaborative perception according to claim 3, characterized in that, The calculation steps for the vehicle identification trail continuity index are as follows: For adjacent local state segments and adjacent local state segments in a cross-edge state segment chain, extract the surrounding vehicles that are observed at the same time as the target vehicle from the end of the previous local state segment to obtain the end-accompanying vehicle set. Extract the surrounding vehicles that are observed at the same time as the target vehicle from the beginning of the next local state segment to obtain the set of vehicles accompanying the beginning. Extract the surrounding vehicles with the same vehicle identifier in the end-accompanying vehicle set and the beginning-accompanying vehicle set to obtain the shared accompanying vehicle set; Count the number of vehicles in the shared accompanying vehicle set, and count the number of vehicles in the vehicle set obtained by merging the end accompanying vehicle set and the beginning accompanying vehicle set and removing duplicate vehicle identifiers. The number of vehicles in the shared accompanying vehicle set is divided by the number of vehicles in the vehicle set obtained after merging and removing duplicate vehicle identifiers, to obtain the accompanying vehicle retention value; For each shared accompanying vehicle in the shared accompanying vehicle set, calculate the relative distance of the shared accompanying vehicle to the target vehicle at the end of the previous local state segment and the relative distance of the shared accompanying vehicle to the target vehicle at the beginning of the next local state segment. Select any two different vehicles from the set of shared accompanying vehicles to form a unique vehicle pair, and count the total number of unique vehicle pairs. For each non-repeating vehicle pair, compare the relative distances of the two shared accompanying vehicles in the non-repeating vehicle pair to the target vehicle at the end of the previous local state segment, and compare the relative distances of the two shared accompanying vehicles in the non-repeating vehicle pair to the target vehicle at the beginning of the next local state segment. When the relative distance between the same non-repeating vehicle pair at the end of a previous local state segment is consistent with the relative distance between the pairs at the beginning of a subsequent local state segment, the non-repeating vehicle pair is determined as a vehicle pair whose distance order remains consistent. The number of vehicle pairs that maintain the same order of distance is counted, and the number of vehicle pairs that maintain the same order of distance is divided by the total number of non-repeating vehicle pairs to obtain the relative formation preservation value. Multiplying the accompanying vehicle retention value by the relative formation retention value yields the vehicle identity trail continuity index corresponding to the previous local state segment and the next local state segment.

6. The method for identifying abnormal vehicle network parameters based on edge collaborative perception according to claim 1, characterized in that, The steps for determining whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain based on the closure result of the cross-edge joint are as follows: Read the seam closure index corresponding to each two adjacent local state segments in the cross-edge state segment chain, and use the seam closure index as the cross-edge seam closure result of the corresponding state seam; Compare the joint closure index corresponding to each state joint with the preset closure threshold respectively; If the joint closure index corresponding to any state joint is less than the preset closure threshold, then the state joint is marked as a broken joint, and it is determined that the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain. If the joint closure index corresponding to all state joints is not less than the preset closure threshold, then the cross-edge state segment chain is determined as a continuous vehicle state chain, and it is determined that the local judgment results of multiple edge nodes can form a continuous vehicle state chain.

7. The method for identifying abnormal vehicle network parameters based on edge collaborative perception according to claim 1, characterized in that, The steps to distinguish the source of the anomaly are as follows: If the local judgment results of multiple edge nodes can form a continuous vehicle state chain, then the local state judgment results corresponding to each local state segment of the vehicle in the continuous vehicle state chain will be used. If the local state judgment results corresponding to each local state segment of a vehicle in a continuous vehicle state chain are all normal, then the abnormal vehicle parameter identification result of the target vehicle is determined to be without abnormality. If there is a vehicle local state segment in the continuous vehicle state chain whose local state judgment result is an abnormal state, then extract the abnormal parameter type corresponding to the abnormal state and determine the abnormal parameter type as the vehicle parameter anomaly identification result of the target vehicle. If the local judgment results of multiple edge nodes cannot form a continuous vehicle state chain, they are marked as a broken joint state joint, and the previous local state fragment and the next local state fragment corresponding to the broken joint are obtained. Obtain the bidirectional backfill encounter index and the vehicle identity trail bearing index corresponding to the fracture joint. If the bidirectional state backfilling encounter index is less than the preset backfilling threshold and the vehicle identity trail acceptance index is not less than the preset identity threshold, then the source of the anomaly will be determined as the vehicle parameter cross-edge motion acceptance anomaly, and the vehicle parameter anomaly identification result of the target vehicle will be determined as position parameter anomaly, speed parameter anomaly or timestamp parameter anomaly. If the vehicle identity trail continuity index is less than the preset identity threshold, the source of the anomaly will be determined as vehicle identity continuity anomaly, and the vehicle parameter anomaly identification result of the target vehicle will be determined as target aliasing anomaly or fragment splicing anomaly. If the bidirectional state backfill encounter index is not less than the preset backfill threshold and the vehicle identity trace acceptance index is not less than the preset identity threshold, but the joint closure index corresponding to the fracture joint is less than the preset closure threshold, then the source of the anomaly will be determined as the local judgment anomaly of the edge node, and the vehicle parameter anomaly identification result of the target vehicle will be determined as the edge node perception deviation anomaly. Output the results of vehicle parameter anomaly identification and the corresponding anomaly source.

8. A vehicle network parameter anomaly identification system based on edge collaborative perception, used to implement the vehicle network parameter anomaly identification method based on edge collaborative perception as described in any one of claims 1-7, characterized in that, The system includes: The edge state module acquires vehicle local state fragments formed by multiple edge nodes, and arranges these fragments into a cross-edge state fragment chain based on vehicle identification, time sequence, and road topology. The seam closure module extracts state seam information between two adjacent local state segments from the cross-edge state segment chain, and constructs the cross-edge seam closure result based on the state seam information; The state chain module determines whether the local judgment results of multiple edge nodes can form a continuous vehicle state chain based on the closure result of the cross-edge seam. The anomaly identification module determines the anomaly identification result of vehicle parameters based on the judgment result of the continuous vehicle state chain, and distinguishes the source of the anomaly.

Images