Positioning result evaluation method and device for vehicle and vehicle

Abstract

The embodiment of the application provides a positioning result evaluation method and device of a vehicle and the vehicle, wherein the method comprises the following steps: acquiring a first quantity of lanes sensed by the vehicle, and acquiring a second quantity of lanes on a target road where the vehicle is located from map information of the vehicle; in response to the first quantity being greater than the second quantity, determining a target lane set of the vehicle from the lanes on the roads around the target road; and based on the quantity of lanes in the target lane set and the quantity difference between the first quantity and the second quantity, evaluating the positioning result of the vehicle in the map information to obtain an evaluation result, wherein the evaluation result is used to represent the accuracy of the positioning result. The application solves the technical problem of low accuracy of the positioning result evaluation of the vehicle.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and more specifically, to a method, apparatus, and vehicle for evaluating vehicle positioning results. Background Technology

[0002] Currently, if the number of lanes sensed by the vehicle (e.g., visual perception) does not match the number of lanes in the map information (e.g., navigation map), the vehicle is directly judged to have failed to locate. However, the situations causing such discrepancies are complex and varied, and are not necessarily caused by vehicle positioning failure. Simply determining positioning failure directly can easily lead to frequent misjudgments, affecting user experience. Therefore, the technical problem of low accuracy in vehicle positioning result evaluation still exists.

[0003] There is currently no good solution to the above problems. Summary of the Invention

[0004] This application provides a method, apparatus, and vehicle for evaluating vehicle positioning results, in order to at least address the technical problem of low accuracy in evaluating vehicle positioning results.

[0005] According to one aspect of the embodiments of this application, a method for evaluating the positioning result of a vehicle is provided, wherein the method may include: obtaining a first number of lanes sensed by the vehicle, and obtaining a second number of lanes on a target road where the vehicle is located from the vehicle's map information; in response to the first number being greater than the second number, determining a target lane set of the vehicle from lanes on roads surrounding the target road; evaluating the positioning result of the vehicle in the map information based on the number of lanes in the target lane set and the difference between the first number and the second number, to obtain an evaluation result, wherein the evaluation result is used to represent the accuracy of the positioning result.

[0006] Furthermore, in response to the first quantity being greater than the second quantity, determining the target lane set of vehicles from the lanes on the roads surrounding the target road includes: in response to the first quantity being greater than the second quantity, performing an accumulation operation on the lanes on the roads surrounding the target road using the lane type to obtain the target lane set, wherein the lane type is used to characterize the positional relationship between the lane and the vehicle.

[0007] Furthermore, using the lane type, an accumulation operation is performed on the lanes on the roads surrounding the target road to obtain a target lane set. This includes: determining a first lane sequence and a second lane sequence from the lanes on the roads surrounding the target road according to the lane type, wherein the first lane sequence represents the number of lanes contained on the road branching off from the target road in a first direction, and the second lane sequence represents the number of lanes contained on the road branching off from the target road in a second direction; performing an accumulation operation on the number of lanes in the first lane sequence according to the order of the roads branching off in the first direction, and performing an accumulation operation on the number of lanes in the second lane sequence according to the order of the roads branching off in the second direction, to obtain the target lane set.

[0008] Furthermore, based on the number of lanes in the target lane cluster and the difference between the first and second numbers, the vehicle's positioning result in the map information is evaluated to obtain the evaluation result. This includes: calling a matching strategy adapted to the lane type of the target lane cluster, matching the number of lanes in the target lane cluster with the difference in number to obtain a matching result. Here, the lane type is used to characterize the positional relationship between the lane and the vehicle, and the matching result is used to indicate the degree of difference between the number of lanes in the target lane cluster and the difference in number. Based on the matching result, the evaluation result is determined, wherein the matching result and the evaluation result are positively correlated.

[0009] Furthermore, the matching strategy includes a first matching strategy, which represents a rule for matching the number of lanes in the target lane set with the difference in number on a road branching off from the target road in a first direction. The first matching strategy calls a matching strategy adapted to the lane type of the target lane set to match the number of lanes in the target lane set with the difference in number, obtaining a matching result. This includes: in response to the first matching strategy, determining the number of lanes in the target lane set in the first direction; and in response to the difference in number between the number of lanes in the first direction and the difference in number being less than a difference threshold, determining the matching result as a match between the number of lanes in the target lane set and the difference in number.

[0010] Further, the matching strategy includes a second matching strategy, which represents a rule for matching the number of lanes in the target lane set with the difference in number on roads branching from the target road to the first and second directions. The second matching strategy calls a matching strategy adapted to the lane type of the target lane set to match the number of lanes in the target lane set with the difference in number, obtaining a matching result. This includes: in response to the second matching strategy, traversing the number of lanes in the first direction and the number of lanes in the second direction of the target lane set; during the traversal, determining the sum between the number of lanes in the first direction and the number of lanes in the second direction traversed; in response to the difference between the sum and the difference in number being less than the minimum lane number error, determining the matching result as the number of lanes in the target lane set matching the difference in number; the method further includes: in response to the difference being greater than or equal to the minimum lane number error, returning to the execution of the following steps until the traversal of the number of lanes in the first direction and the number of lanes in the second direction of the target lane set is completed: traversing the number of lanes in the first direction and the number of lanes in the second direction of the target lane set.

[0011] Furthermore, based on the matching results, the evaluation results are determined, including: in response to the matching result that the number of lanes in the target lane cluster matches the number difference, the evaluation result is determined to be that the accuracy is greater than or equal to the accuracy threshold; in response to the matching result that the number of lanes in the target lane cluster does not match the number difference, the evaluation result is determined to be that the accuracy is less than the accuracy threshold.

[0012] Furthermore, the positioning result includes the sensing positioning result, and the method further includes: in response to the evaluation result that the accuracy is greater than or equal to the accuracy threshold, adjusting the current weight of the sensing positioning result to a first target weight, wherein the first target weight is greater than the current weight; in response to the evaluation result that the accuracy is less than the accuracy threshold, performing one of the following processing operations on the sensing positioning result: adjusting the weight of the sensing positioning result to a second target weight, wherein the second target weight is less than the current weight; and smoothing the sensing positioning result.

[0013] According to another aspect of the embodiments of this application, a vehicle positioning result evaluation device is also provided. The device may include: an acquisition module, configured to acquire a first number of lanes sensed by the vehicle, and acquire a second number of lanes on a target road where the vehicle is located from the vehicle's map information; a determination module, configured to determine a target lane set of the vehicle from lanes on roads surrounding the target road in response to the first number being greater than the second number; and an evaluation module, configured to evaluate the positioning result of the vehicle in the map information based on the number of lanes in the target lane set and the difference between the first number and the second number, and obtain an evaluation result, wherein the evaluation result is used to represent the accuracy of the positioning result.

[0014] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0015] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0016] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0017] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0018] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the methods in various embodiments of this application.

[0019] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0020] In this embodiment, if it is necessary to evaluate the vehicle's positioning results, a first number of lanes sensed by the vehicle can be obtained, or a second number of lanes on the target road where the vehicle is currently located can be obtained from the vehicle's map information. If the first number is greater than the second number, the vehicle's target lane set can be determined from vehicles on roads surrounding the target road. The vehicle's positioning results in the map information can be evaluated based on the number of lanes in the target lane set and the difference between the first and second numbers to obtain an evaluation result. In other words, in this embodiment of the application, when the first number sensed is greater than the second number in the map information, instead of simply determining that the positioning has failed, a matching set of target lanes is further extracted from the surrounding roads of the target road. Combined with the difference in the number, the difference between the number of sensed lanes and the number of lanes on the target road in the map information is quantitatively evaluated to determine whether it can be reasonably explained by the set of target lanes on the surrounding roads. This transforms the mismatch between the first and second numbers into a measurable accuracy evaluation of the positioning result, avoiding the misjudgment of determining positioning failure directly when the first and second numbers do not match. This achieves the technical effect of improving the accuracy of the vehicle positioning result evaluation and solves the technical problem of low accuracy of the vehicle positioning result evaluation. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0022] Figure 1 This is a flowchart of a vehicle positioning result evaluation method according to an embodiment of this application;

[0023] Figure 2 This is a schematic diagram of a lane matching device for a visual and navigation map according to an embodiment of this application;

[0024] Figure 3 This is a flowchart of a lane matching method for visual and navigation maps according to an embodiment of this application;

[0025] Figure 4 This is a flowchart of a sub-lane matching method according to an embodiment of this application;

[0026] Figure 5 This is a flowchart of a two-pointer search algorithm according to an embodiment of this application;

[0027] Figure 6 This is a schematic diagram of a vehicle positioning result evaluation device according to an embodiment of this application;

[0028] Figure 7This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] According to an embodiment of this application, a method for evaluating the positioning results of a vehicle is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0032] This embodiment provides a method for evaluating vehicle positioning results. Figure 1 This is a flowchart of a vehicle positioning result evaluation method according to an embodiment of this application, such as... Figure 1 As shown, the method may include the following steps.

[0033] Step S102: Obtain the first number of lanes sensed by the vehicle, and obtain the second number of lanes on the target road where the vehicle is located from the vehicle's map information.

[0034] In the technical solution provided by step S102 of this application embodiment, the first quantity can refer to the number of lanes around the vehicle sensed by the vehicle, and can be used to represent the real-time observation status of the road structure by the sensing system in the vehicle. For example, if the sensing system is a visual perception module, the quantity can be the total number of lanes visible in the current field of view detected by the visual perception module. The map information can refer to the road topology data stored in the vehicle's navigation map. For example, the map information can include the geometric features, lane distribution, and connection relationships of the roads around the vehicle. The target road can refer to the road the vehicle is currently traveling on, which is the main driving path corresponding to the vehicle's positioning point in the navigation map. For example, the target road can be the main road the vehicle is currently traveling on. The second quantity can refer to the number of main lanes preset in the navigation map for the target road, used to characterize the navigation map's description of the lane structure of the target road where the vehicle is currently located.

[0035] In this embodiment, if it is necessary to evaluate the vehicle's positioning results, the first number of lanes sensed by the vehicle can be obtained, or the second number of lanes on the target road where the vehicle is currently located can be obtained from the vehicle's map information.

[0036] Optionally, during vehicle operation, the onboard vision sensing module can identify and count visible lane line boundaries based on road images captured by the forward-facing camera using a deep learning model, and output the total number of consecutive lanes within the current field of view. This total number of lanes can be defined as the first quantity.

[0037] Optionally, during vehicle operation, the vehicle's positioning module can determine the vehicle's spatial location on the navigation map based on the global navigation satellite system and inertial navigation information, query the lane topology attributes of the target road in the map information of the navigation map, extract the number of lanes marked on the target road within a preset road segment, and determine the above number of lanes as a second quantity.

[0038] In this embodiment of the application, the above method forms comparable quantitative input pairs by uniformly collecting visual observation values ​​and map prior values, which lays the data foundation for the subsequent positioning evaluation mechanism based on difference analysis, ensuring that the process of evaluating the positioning results of the vehicle does not rely on fuzzy judgment, and improving the objectivity and reproducibility of the positioning result evaluation.

[0039] In step S104, in response to the first quantity being greater than the second quantity, the target lane set of the vehicle is determined from the lanes on the roads surrounding the target road.

[0040] In the technical solution provided in step S104 of this application embodiment, the target lane set can refer to the set of lane number sequences predefined by the navigation map on both sides of the target road where the vehicle is located, which may extend or diverge from the target road. The target lane set can be used to characterize the scalable lane topology of the target road under complex traffic structures. For example, the target lane set may include the lane number sequence of the left sub-path and the lane number sequence of the right sub-path, respectively used to record the lane segments arranged in the driving direction order of the continuous sub-paths (e.g., left-turn lanes, right-turn ramps, merging lanes, etc.) on the left and right sides of the target road. The target lane set does not include the target road itself, but covers the auxiliary paths branching off from the target road. Its function is to provide a possible matching source for the visually perceived excess lane number and support the structured interpretation of the difference in the number of lanes. The target lane set can also be called the sub-path lane cumulative sequence set.

[0041] In this embodiment, after obtaining the first quantity and the second quantity, the size relationship between the first quantity and the second quantity can be compared. If the first quantity is greater than the second quantity, the target lane set of the vehicle can be determined from the lanes on the roads surrounding the target road.

[0042] Optionally, in this embodiment, when the first number of visually perceived lanes exceeds the second number of target lanes in the navigation map, structured lane information is actively extracted from roads (sub-paths) surrounding the target road to construct a target lane set, which is used to explain the possible sources of the excess lane number. The above steps in this embodiment can be triggered when the first number exceeds the second number, aiming to distinguish between normal road structure transitions and positioning anomalies, avoid indiscriminate misjudgments, and provide topological basis for subsequent accurate evaluation.

[0043] Optionally, if the vehicle's location is located in a branching, merging, or lane-addition / decrease area of ​​the target road, the pre-stored sub-path topology data in the navigation map is activated. The lane number sequences of each connectable auxiliary path on the left and / or right side of the target road are extracted sequentially, and cumulative values ​​are calculated according to the driving direction, forming a left-side cumulative lane number sequence and a right-side cumulative lane number sequence. These two sequences together constitute the target lane set. It should be noted that the target lane set described in this embodiment can cover all roads branching off from the target road. Each element in the target lane set can be used to represent the cumulative number of lanes from the target road to the end of a certain road, characterizing the possible changes in the road structure in the lateral expansion dimension.

[0044] Optionally, the above steps can be performed when the first number is greater than the second number. When the first number is less than the second number, it indicates that there are lane omissions or occlusions in visual perception, which does not involve road structure expansion and does not require sub-path matching. When the first number equals the second number, it indicates that the visual perception matches the road structure in the navigation map, and the positioning is in a normal matching state, requiring no further analysis. Only when the first number is greater than the second number is there a possibility of an increase in the number of legal lanes due to branching, ramp access, or lane divergence. In this case, the target lane set can be used to verify whether the excess originates from a known topology structure on the map.

[0045] In the embodiments of this application, the above method achieves accurate modeling of dynamic changes in road structure by conditionally activating sub-path topology information. When necessary, it introduces complex calculations, which not only avoids meaningless redundant processing, but also ensures the ability to provide a structured explanation of the source of differences in the number of lanes in complex scenarios, significantly improving the scenario adaptability and resource efficiency of positioning evaluation.

[0046] Step S106: Based on the number of lanes in the target lane and the difference between the first number and the second number, evaluate the vehicle's positioning result in the map information to obtain the evaluation result.

[0047] In the technical solution provided by step S106 of this application embodiment, the quantity difference can refer to the numerical difference between the visually perceived first number of lanes and the second number of main lanes on the navigation map, which can be used to characterize the degree of deviation of the number of lanes between the current visual observation result and the description of the main road structure on the map. The above evaluation result can refer to the quantitative evaluation output generated based on the matching relationship between the cumulative lane number sequence in the target lane set and the quantity difference, which is used to indicate whether the vehicle's positioning result in the navigation map can be reasonably explained by the current road topology. The evaluation result can include a matching flag, the minimum matching error value, and the cumulative number of lanes in the left and right sub-paths corresponding to the matching. Its core function is to transform the binary positioning conclusion into a continuous confidence measure.

[0048] In this embodiment, after determining the target lane set from the lanes on the roads surrounding the target road, the positioning result of the vehicle in the map information can be evaluated based on the number of lanes in the target lane set and the difference between the first number and the second number, and the corresponding evaluation result can be obtained.

[0049] Optionally, the positioning result can refer to the spatial consistency estimate of the vehicle in the coordinate system of the navigation map obtained by the multi-sensor fusion algorithm, which is used to characterize the degree of matching between the vehicle's current position and the semantic structure of the map. In this embodiment of the application, the positioning result can be verified for confidence through evaluation results. For example, when the evaluation results show that the difference in quantity can be reasonably explained by the target lane set, the positioning result maintains a high confidence level; when it cannot be explained, the confidence level of the positioning result is lowered, thereby realizing dynamic evaluation and feedback of positioning accuracy.

[0050] Optionally, the quantity difference is used as the target matching value. The sum of all possible combinations in the left and right cumulative sequences of the target lane set is compared sequentially to find the value combination closest to this difference. When the sum of a combination matches the quantity difference, it indicates that the visually observed excess lanes can be fully explained by the known sub-path structure in the map. In this case, the semantic consistency of the positioning result is established, the evaluation result is a successful match, and the confidence level is high. When no matching combination exists, the combination with the smallest absolute error of the quantity difference can be found among all combinations. This minimum error is used as the core indicator of the evaluation result, indicating a slight deviation between the positioning result and the map structure, but still within an acceptable range of structural changes. The evaluation result is a partial match, and the confidence level is medium. When the minimum error exceeds a preset threshold, it indicates a significant conflict between visual observation and map topology. The semantic credibility of the positioning result fails, the evaluation result is a mismatch, and the confidence level is low.

[0051] It should be noted that the above evaluation process in the embodiments of this application does not need to rely on a single threshold judgment, but can construct a continuous evaluation function for positioning accuracy through structured search and error quantification.

[0052] Optionally, this step is not executed if the first number is less than or equal to the second number, because in this case, there is no structural expansion phenomenon where the number of lanes exceeds the number described by the main road, and there is no need to verify the sub-path matching. The evaluation of the localization result can be directly based on the consistency of the main road without introducing the target lane set. This step is triggered only when the first number is greater than the second number, ensuring that the algorithm resources are concentrated on high-risk and high-complexity boundary scenarios.

[0053] In this embodiment, the method elevates the evaluation of positioning accuracy from absolute coordinate alignment to road topology semantic consistency verification, thereby achieving a deeper understanding of positioning reliability in complex driving scenarios. It transforms what was originally considered an "error" in the number of lanes into interpretable and quantifiable structural matching evidence, distinguishing between "positioning failure" and "road structure changes," thus avoiding mishandling of traffic due to normal scenarios such as intersection forks and mergings.

[0054] In steps S102 to S106 of this embodiment, if it is necessary to evaluate the vehicle's positioning result, a first number of lanes sensed by the vehicle can be obtained, or a second number of lanes on the target road where the vehicle is currently located can be obtained from the vehicle's map information. If the first number is greater than the second number, the target lane set of the vehicle can be determined from the vehicles on the roads surrounding the target road. The positioning result of the vehicle in the map information can be evaluated based on the number of lanes in the target lane set and the difference between the first and second numbers to obtain the evaluation result. In other words, in this embodiment of the application, when the first number sensed is greater than the second number in the map information, instead of simply determining that the positioning has failed, a matching set of target lanes is further extracted from the surrounding roads of the target road. Combined with the difference in the number, the difference between the number of sensed lanes and the number of lanes on the target road in the map information is quantitatively evaluated to determine whether it can be reasonably explained by the set of target lanes on the surrounding roads. This transforms the mismatch between the first and second numbers into a measurable accuracy evaluation of the positioning result, avoiding the misjudgment of determining positioning failure directly when the first and second numbers do not match. This achieves the technical effect of improving the accuracy of the vehicle positioning result evaluation and solves the technical problem of low accuracy of the vehicle positioning result evaluation.

[0055] The embodiments of this application will be described in detail below with reference to the steps described above.

[0056] As an optional implementation, step S104, in response to the first quantity being greater than the second quantity, determines the target lane set of vehicles from the lanes on the roads surrounding the target road, including: in response to the first quantity being greater than the second quantity, performing an accumulation operation on the lanes on the roads surrounding the target road using the lane type to obtain the target lane set, wherein the lane type is used to characterize the positional relationship between the lane and the vehicle.

[0057] In this embodiment, during the process of determining the target lane set from the lanes on the roads surrounding the target road, if the first number is greater than the second number, the lane type can be used to perform an accumulation operation on the lanes on the roads surrounding the target road to obtain the target lane set. The lane type can be used as a semantic classification label to characterize the lateral spatial positional relationship of each lane around the target road relative to the currently traveling target road. Its core function is to distinguish the orientation of lanes in the vehicle's lateral coordinate system to support structured topological modeling. For example, lane types can include left-side lane types and right-side lane types. The left-side lane type can refer to an auxiliary lane located outside the left boundary of the main road and geometrically and topologically continuous with the main road. The right-side lane type can refer to an auxiliary lane located outside the right boundary of the main road and geometrically and topologically continuous with the main road. It should be noted that the lane types in this embodiment can be divided based on spatial relative position, with the aim of constructing an accumulative lateral expansion path index consistent with the vehicle's travel direction, ensuring that the construction of the target lane set has clear spatial semantic consistency.

[0058] Optionally, the aforementioned accumulation operation can refer to a mathematical processing procedure that sequentially accumulates the number of lanes in the sub-paths on both sides of the target road based on lane type. The purpose is to transform discrete sub-path lane information into a continuous, matchable sequence of lane increments to support a structured interpretation of the quantity differences. This accumulation operation can be performed independently on the left and right lane sequences, based on lane type. For example, starting from the end of the target road (e.g., the main road), the continuous sub-paths connected to the main road are traversed sequentially along the driving direction, and the number of lanes in each sub-path segment is accumulated segment by segment to form a sequentially arranged sequence of accumulated values.

[0059] For example, if the original number of lanes in the continuous sub-path on the left is [1,2,1], then after the accumulation operation, the sequence of accumulated lanes on the left is generated as [1,3,4], where each element represents the total number of lanes accumulated from the starting point of the main road to the end of the sub-path.

[0060] It should be noted that the above-mentioned accumulation operation in the embodiments of this application can be triggered when the lane type is the left lane type or the right lane type, ensuring that the accumulation process acts on the legal extended path connected to the main road topology, and excluding interference from non-contiguous, non-adjacent or unstructured paths.

[0061] Optionally, the accumulation operation can be a mathematical process of sequentially accumulating the number of lanes in each sub-path under the lane type on both sides of the target road. The purpose is to transform discrete sub-path lane information into a continuous, matchable sequence of lane increments to support a structured interpretation of the difference between the first and second counts. This accumulation operation can be performed independently on the left and right lane sequences, using lane type as the dividing dimension. During the accumulation operation, starting from the end of the target road, consecutive sub-paths topologically connected to the target road can be visited sequentially along the driving direction. The number of lanes in each sub-path is sequentially accumulated to form an increasing sequence of accumulated values. Each element in the accumulated value sequence can represent the total number of lanes accumulated from the starting point of the target road to the end of that sub-path. For example, if the original number of lanes in the left consecutive sub-paths is [1,2,1], then after the accumulation operation, the left accumulated lane count sequence [1,3,4] is generated, where the value 3 can be used to represent the total number of lanes in the first two sub-paths, and the value 4 can be used to represent the sum of the three sub-paths. Similarly, the right sub-path [2,1] is accumulated to generate the right cumulative lane number sequence [2,3].

[0062] In this embodiment, lane type is introduced as a spatial semantic classification standard, and an accumulation operation is performed based on the lane type to achieve accurate modeling of dynamic changes in road structure. The target lane set not only records the number of lanes in sub-paths but also encodes their spatial distribution and accumulated path relationship, enabling the distinction between "positioning errors" and "legitimate road structure changes." This mechanism endows the target lane set with clear topological interpretability, structural traceability, and computational repeatability, providing high-quality input for subsequent error matching and positioning evaluation. This significantly improves the system's positioning robustness and semantic consistency judgment capabilities in complex scenarios such as bifurcation, merging, and ramp access.

[0063] As an optional implementation, a target lane set is obtained by performing an accumulation operation on lanes on roads surrounding the target road using lane types. This includes: determining a first lane sequence and a second lane sequence from lanes on roads surrounding the target road according to lane types, wherein the first lane sequence represents the number of lanes contained on roads branching off from the target road in a first direction, and the second lane sequence represents the number of lanes contained on roads branching off from the target road in a second direction; performing an accumulation operation on the number of lanes in the first lane sequence according to the order of the roads branching off in the first direction, and performing an accumulation operation on the number of lanes in the second lane sequence according to the order of the roads branching off in the second direction, to obtain the target lane set.

[0064] In this embodiment, during the accumulation operation on lanes on roads surrounding the target road, a first lane sequence and a second lane sequence can be determined from the lanes on roads surrounding the target road according to lane type. The accumulation operation can be performed on the number of lanes in the first lane sequence according to the order of roads branching in the first direction. Alternatively, the accumulation operation can be performed on the number of lanes in the second lane sequence according to the order of roads branching in the second direction. The first lane sequence can refer to the original data sequence of the number of lanes contained in each sub-path branching to the left from the target road, arranged in order of travel direction. The first lane sequence can be used to represent the lateral extension structures that a vehicle may encounter on the left side of the main road. The first lane sequence can be extracted from predefined road topology data in a navigation map, where each element corresponds to an independent left-side auxiliary path that is geometrically and topologically continuous with the main road, and the elements are arranged in order of extension from the end of the main road to the end of the branch, excluding the main road itself. The first lane sequence can also be called the original lane number sequence of the left sub-path, the lane list of the left extension path, or the lane distribution sequence of the left branch.

[0065] Optionally, the second lane sequence can refer to the original data sequence of lane numbers in a continuous sub-path branching off to the right from the target road, arranged in order of travel direction. This second lane sequence can be used to characterize the lateral extension structures that vehicles may encounter on the right side of the main road. The second lane sequence can be extracted from the topology database of the navigation map, where each element corresponds to an independent right-hand auxiliary path that is geometrically and topologically continuous with the main road, and the elements are arranged in order of extension from the end of the main road to the end of the branch, excluding the main road itself. The second lane sequence can also be called the original lane number sequence of the right-hand sub-path, the lane list of the right-hand extension path, or the lane distribution sequence of the right-hand branch.

[0066] Optionally, the first direction can refer to the leftward spatial extension direction defined relative to the centerline of the target road's main road during vehicle travel. This is an inherent attribute of the road topology in the navigation map, used to identify the lateral spatial area to which a sub-path branching off from the main road to the left belongs. The first direction is consistent with the vehicle's direction of travel and does not change with the vehicle's heading angle. The second direction can refer to the rightward spatial extension direction defined relative to the centerline of the target road's main road during vehicle travel. This is also an inherent attribute of the road topology in the navigation map, used to identify the lateral spatial area to which a sub-path branching off from the main road to the right belongs. The second direction is consistent with the vehicle's direction of travel and does not change with the vehicle's heading angle. Essentially, it is an absolute spatial orientation definition based on the map coordinate system, used to distinguish the lane's location in the lateral coordinate system.

[0067] Optionally, based on predefined semantic labels of the road topology in the navigation map, connectable branching paths on the left and right sides of the target road in the driving direction are identified. According to the definitions of "left lane" and "right lane" in the lane types, geometrically and topologically continuous auxiliary paths connected to the main road are extracted, forming two independent lane number sequences. The first lane sequence only includes the number of lanes contained in continuous sub-paths branching to the left from the main road, with each element arranged in the order of extension from the end of the main road to the end of the branch, and includes a left lane segment directly connected to the main road without interruption. The second lane sequence includes the number of lanes contained in continuous sub-paths branching to the right from the main road, with each element also arranged in the order of extension from the end of the main road to the end of the branch, and includes a right lane segment directly connected to the main road without interruption. This process eliminates unstructured paths, discontinuous branching, cross-connections, or non-parallel extension paths, ensuring that each element in the first and second lane sequences corresponds to a clearly defined spatially belonging, topologically connected, and directionally consistent sub-path, thereby guaranteeing the semantic accuracy and computational reliability of subsequent cumulative operations.

[0068] Optionally, according to the road sequence of the first directional branch, an accumulation operation is performed on the number of lanes in the first lane sequence. This includes: taking the first lane sequence as input, starting from the first element of the sequence, and accumulating the number of lanes in each sub-path item by item along the driving direction to generate an increasing accumulation sequence, where each item represents the total number of left lanes accumulated from the end of the main road to the end of the current sub-path. For example, if the original data of the first lane sequence is [1,2,1], the accumulation process is as follows: the first item is 1, the second item is 1+2=3, the third item is 3+1=4, and finally, the left accumulated lane number sequence [1,3,4] is generated. This accumulation operation is triggered when the lane type is "left lane" and applies to consecutive adjacent sub-paths. If the path is interrupted or there are non-continuous branches, the accumulation is terminated, and only valid continuous segments are retained. The result of this accumulation operation is the left accumulated lane number sequence, where each element corresponds to a visually perceptible combination of laterally extended paths, providing a structured candidate set for subsequent matching with the number difference.

[0069] Optionally, taking the second lane sequence as input, starting from the first element of the second lane sequence, the number of lanes in each sub-path is accumulated sequentially along the driving direction to generate an increasing cumulative sequence, where each item represents the total number of right-hand lanes accumulated from the end of the main road to the end of the current sub-path. For example, if the original data of the second lane sequence is [2,1], the accumulation process is as follows: the first item is 2, the second item is 2+1=3, and finally the right-hand cumulative lane number sequence [2,3] is generated. This accumulation operation is triggered only when the lane type is "right-hand lane" and only applies to consecutive adjacent sub-paths. If the path is interrupted or there are non-continuous branches, the accumulation is terminated, and only valid continuous segments are retained. The result of this accumulation operation is the right-hand cumulative lane number sequence, where each element corresponds to a visually perceptible combination of laterally extended paths, providing a structured candidate set for subsequent matching with the number difference. The left-hand cumulative sequence and the right-hand cumulative sequence together constitute the target lane set, which is a two-dimensional, structured, and composable lane increment candidate set, supporting efficient matching through dual pointers or binary search.

[0070] In this embodiment, the method introduces a spatial classification mechanism for the first and second directions, combines structured extraction of the first and second lane sequences, and then generates a target lane set with semantic continuity through accumulation operations. This process does not rely on the absolute positioning accuracy of high-precision maps and can reuse the topological information of standard navigation maps and the relative quantity information of visual perception, building a highly robust lane-level semantic consistency verification capability with extremely low computational cost. This not only improves positioning fault tolerance in complex scenarios such as bifurcation, merging, and ramp access, but also enables gradient decision-making based on the matching error of the accumulated sequence, avoiding accidental takeover due to normal road structure changes, significantly enhancing the safety, continuity, and user experience of the autonomous driving system.

[0071] As an optional implementation, step S106 involves evaluating the vehicle's location in the map information based on the number of lanes in the target lane cluster and the difference between the first and second numbers, to obtain an evaluation result. This includes: invoking a matching strategy adapted to the lane type of the target lane cluster, matching the number of lanes in the target lane cluster with the difference in numbers to obtain a matching result, wherein the lane type is used to characterize the positional relationship between the lane and the vehicle, and the matching result is used to indicate the degree of difference between the number of lanes in the target lane cluster and the difference in numbers; and determining the evaluation result based on the matching result, wherein the matching result and the evaluation result are positively correlated.

[0072] In this embodiment, during the evaluation of the positioning results based on the number of vehicles in the target lane set and the difference in their numbers, a matching strategy adapted to the lane type of the target lane set can be invoked to match the number of lanes and their differences in numbers in the target lane set, obtaining a matching result. The evaluation result can be determined based on this matching result. The matching result can refer to the lane combination value determined by a structured search algorithm that is closest to the difference in numbers between the first and second numbers, within the left and right cumulative lane number sequences represented by the target lane set. This matching result can be used to quantify the degree of topological matching between the visually perceived number of lanes and the sub-path structure of the navigation map. The matching result can be a structured output consisting of three key elements: minimum lane number error, cumulative lane number of the left sub-path, and cumulative lane number of the right sub-path.

[0073] The minimum lane number error can be defined as the minimum absolute difference between the sum of all possible left and right sub-path combinations and the lane number difference. The magnitude of the minimum lane number error reflects the semantic deviation between the current visual observation and the map structure. The cumulative lane number of the left sub-path represents the total number of lanes contributed by the left sub-path in the matching combination. The cumulative lane number of the right sub-path represents the total number of lanes contributed by the right sub-path in the matching combination.

[0074] Optionally, a quantity difference is obtained, which can be the difference between the first number of lanes perceived visually and the second number of main lanes on the navigation map, representing the number of excess lanes not covered by the main road structure in the current observation. Based on the lane type of the lanes in the target lane set, it is determined whether the current road structure has the possibility of expansion to a left sub-path, a right sub-path, or both sub-paths. When only a left sub-path exists, the matching strategy calls a binary search algorithm to find the value closest to the quantity difference in the left cumulative lane count sequence, calculates the absolute error between this value and the quantity difference, and records the corresponding location index as the cumulative lane count of the left sub-path; the cumulative lane count of the right sub-path is zero. When a right-hand sub-path exists, the matching strategy calls a binary search algorithm, performing the same operation in the right-hand cumulative lane count sequence, recording the cumulative lane count of the right-hand sub-path, while the cumulative lane count of the left-hand sub-path is zero. When both left and right-hand sub-paths exist, the matching strategy calls a two-pointer linear scan algorithm, initializing the left pointer to point to the beginning of the left-hand cumulative sequence and the right pointer to point to the end of the right-hand cumulative sequence. Iteratively calculating the sum of the cumulative values ​​of the left and right sub-paths and comparing it with the difference in number, dynamically adjusting the pointer positions until all possible combinations are traversed, finally determining the left-right sub-path combination that minimizes the total error, and recording the minimum lane count error and the corresponding cumulative lane counts of the left and right sides. This matching strategy is selected based on the structural characteristics of the sub-path spatial distribution, ensuring that the algorithm can complete matching efficiently in different scenarios. Its output is a structured matching result, containing three core fields: minimum lane count error, cumulative lane count of the left-hand sub-path, and cumulative lane count of the right-hand sub-path.

[0075] Optionally, after determining the matching result, the minimum lane number error in the matching result can be analyzed. The minimum lane number error can be the minimum absolute deviation between each candidate combination and the number difference during the matching process, and can be used to reflect the degree of consistency between visual observation and map topology. According to the preset semantic classification rules, the minimum lane number error is mapped to the location reliability level. For example, when the minimum lane number error is zero, it indicates that the excess lanes perceived by vision can be fully explained by the map sub-path, and the evaluation result is "match successful", and the location reliability is marked as high; when the minimum lane number error is 1 or 2, it indicates that there is a slight structural deviation or perceived noise, but the overall topology can still be explained, and the evaluation result is "partial match", and the location reliability is marked as medium; when the minimum lane number error exceeds the preset threshold (such as 3 or higher), it indicates that there is a significant conflict between visual observation and map structure, which cannot be reasonably explained by legal sub-paths, and the evaluation result is "mismatch", and the location reliability is marked as low.

[0076] The evaluation results in this embodiment do not rely on hard judgment based on absolute thresholds. Instead, they are semantically graded based on continuous error values ​​to ensure that the system can respond differently to different levels of matching quality. This mapping process can follow the logic that matching results and evaluation results are positively correlated: the smaller the matching error, the higher the evaluation confidence; the larger the matching error, the lower the evaluation confidence, and the two have a monotonically positive functional relationship.

[0077] In this embodiment, the above method transforms the originally ambiguous problem of "inconsistent lane numbers" into a computable, interpretable, and decision-making task of positioning credibility assessment through a structured matching strategy and a semantic evaluation mechanism. Its core advantage lies in its ability to achieve semantic-level verification of dynamic scenarios such as bifurcation, merging, and ramp access without relying on the absolute accuracy of high-precision maps. It reuses the topological structure of standard navigation maps and visually perceived relative quantity information.

[0078] As an optional implementation, the matching strategy includes a first matching strategy, which represents a rule for matching the number of lanes in the target lane set with the difference in number on a road branching off from the target road in a first direction. The first matching strategy calls a matching strategy adapted to the lane type of the target lane set to match the number of lanes in the target lane set with the difference in number, obtaining a matching result. This includes: in response to the first matching strategy, determining the number of lanes in the target lane set in the first direction; and in response to the difference in number between the number of lanes in the first direction and the difference in number being less than a difference threshold, determining the matching result as a match between the number of lanes in the target lane set and the difference in number.

[0079] In this embodiment, during the process of matching the number of lanes and their differences in the target lane set using a matching strategy, if the matching strategy is the first matching strategy, the number of lanes in the first direction of the target lane set can be determined. If the difference between the number of lanes and their differences in the first direction is less than a difference threshold, the matching result can be determined as a match between the number of lanes and their differences in the target lane set. The first matching strategy can refer to a structured matching rule used to evaluate the consistency between the cumulative number of lanes in that direction in the target lane set and the difference between the visually perceived number of lanes and the number of main lanes on the map when there is a single-direction sub-path extension in the navigation map (i.e., only the left or only the right side has a connectable branch path). The first matching strategy is used to handle single-sided lane extension scenarios. Its core logic is: under the condition of containing a left-side sub-path or only a right-side sub-path, by searching for the element closest to the difference in the corresponding single-sided cumulative lane count sequence, it determines whether the difference between this element and the difference in the number is within an acceptable semantic tolerance range, thereby inferring whether the visually observed excess lanes can be reasonably explained by the single-sided sub-path structure.

[0080] Optionally, the first matching strategy mentioned above can also be called a single-lane matching strategy, a one-way sub-path matching rule, or a one-sided cumulative matching mechanism. Its purpose is to provide an efficient and low-computational-overhead structural matching scheme for road nodes that do not have the ability to expand to both sides.

[0081] Optionally, after constructing the target lane set, the system determines whether there is a situation where only one-sided sub-path expansion exists based on the topological characteristics of the current road environment. This includes situations where a left-side sub-path exists while the right-side sub-path is empty, or vice versa. In this case, the matching strategy is automatically identified as the first matching strategy. A cumulative lane count sequence corresponding to the first direction is extracted from the target lane set. This sequence is generated by cumulatively adding the original lane counts of all continuous sub-paths in that direction according to the driving direction in the navigation map. Each element represents the total number of lanes accumulated from the end of the main road to the end of the corresponding sub-path. In this cumulative sequence, a binary search is used to retrieve the candidate value closest to the difference in lane count. This candidate value represents the number of lanes in the first direction. This process is triggered only when a single-sided sub-path exists and does not involve any information about the other side of the path. The result is a definite, non-empty cumulative lane value, representing the maximum interpretable lane increment that can be provided by extending from the main road to the first direction in the current road structure. The generation of this value does not depend on the absolute coordinates of the vehicle's real-time location, but only on the topological relationships and lane number modeling of the map, ensuring its semantic consistency and environmental independence.

[0082] Optionally, the difference between the number of lanes in the first direction and the difference in number is calculated. This difference is determined as the difference degree, which is used to quantify whether the excess lanes observed in the current visual observation can be accommodated by the structure of the sub-path on this side. The difference degree is compared with a preset difference degree threshold, which is a semantic tolerance value set by the system based on typical perception errors and map resolution. Its value range is usually 0 to 2 lanes, used to distinguish between "structurally reasonable explanation" and "significant anomaly". When the difference degree is less than the difference degree threshold, it is determined that the number of lanes in the first direction can fully explain the difference in number, that is, the lane increment observed in the visual observation is within the interpretable range of the map topology, and the matching result is set as "match successful". When the difference degree is greater than or equal to the difference degree threshold, the system determines that the single sub-path cannot reasonably explain the observation increment, and the matching result is "mismatch", but the number of lanes in the first direction can be recorded as a minimum error candidate for subsequent evaluation.

[0083] In this embodiment, the method described above achieves efficient and robust verification of lane matching in single-sided bifurcation scenarios through a first matching strategy. It constructs structured matching logic with low computational complexity, avoiding redundant computations caused by blindly initiating bilateral combined searches in single-sided scenarios.

[0084] As an optional implementation, the matching strategy includes a second matching strategy, which represents a rule for matching the number of lanes in the target lane set with the difference in number on roads branching from the target road to the first and second directions. The second matching strategy calls a matching strategy adapted to the lane type of the target lane set to match the number of lanes in the target lane set with the difference in number, obtaining a matching result. This includes: in response to the second matching strategy, traversing the number of lanes in the first direction and the number of lanes in the second direction of the target lane set; during the traversal, determining the sum between the number of lanes in the first direction and the number of lanes in the second direction traversed; in response to the difference between the sum and the difference in number being less than the minimum lane number error, determining the matching result as the number of lanes in the target lane set matching the difference in number; the method further includes: in response to the difference being greater than or equal to the minimum lane number error, returning to the execution of the following steps until the traversal of the number of lanes in the first direction and the number of lanes in the second direction of the target lane set is completed: traversing the number of lanes in the first direction and the number of lanes in the second direction of the target lane set.

[0085] In this embodiment, during the process of matching the number of lanes in the target lane set with the difference between the number of lanes, if the matching strategy is the second matching strategy, the number of lanes in the first direction and the number of lanes in the second direction of the target lane set can be traversed. During the traversal, the sum of the number of lanes in the first direction and the number of lanes in the second direction that have been traversed can be determined. If the difference between the above sum and the difference between the number of lanes is less than the minimum lane number error, then the matching result can be determined as the number of lanes in the target lane set matching the difference between the number of lanes. If the above difference is greater than or equal to the minimum lane number error, then the process can return to continue traversing the number of lanes until the number of lanes in the first direction and the number of lanes in the second direction of the target lane set have been traversed.

[0086] The aforementioned second matching strategy refers to a structured search rule used to jointly match the difference between the cumulative number of lanes on both sides of the target lane set and the visually perceived number of lanes and the number of main lanes on the map when there are sub-path extensions in both directions on the navigation map. This second matching strategy can also be called the combined optimization matching of the two-sided sub-paths. Its core function is to traverse all possible matching combinations of the left and right cumulative lane number sequences, finding the matching scheme that makes the sum of the combinations closest to the difference in number, thereby achieving topological semantic interpretation of visually observed excess lanes at complex intersections (e.g., bidirectional forks, merging zones, and multi-exit interchanges).

[0087] Optionally, the sum of the number of lanes in the first direction and the number of lanes in the second direction can refer to the sum obtained by arithmetically adding the cumulative number of lanes on the left and the cumulative number of lanes on the right in a certain group of target lane sets during the execution of the second matching strategy. This sum can be used to represent the cumulative number of interpretable lanes that can be jointly provided by the main road branching to the left and right sides under a specific sub-path combination. This sum can also be called the total number of lanes in the combination, the cumulative lanes on both sides, or the joint contribution value of sub-paths, and its mathematical expression is sum=left_accum_cnt[i]+right_accum_cnt[j], where left_accum_cnt[i] can be used to represent the i-th element in the left cumulative sequence, and right_accum_cnt[j] can be used to represent the j-th element in the right cumulative sequence.

[0088] Optionally, the difference between the sum and the quantity difference can refer to the value obtained after calculating the absolute deviation between the sum and the quantity difference in the second matching strategy. This value is used to quantify the matching error between the current combination scheme and the observation requirements. The above difference can also be called the lane quantity error, mathematically expressed as error = abs(sum - target), where target can be used to represent the difference between the first number of visually perceived lanes and the second number of main lanes on the map, i.e., the quantity difference. This difference can be used to reflect the gap between the total number of lanes that the current left and right sub-path combination can provide and the number of "extra" lanes in actual visual observation.

[0089] Optionally, the aforementioned minimum lane number error can refer to the minimum value among all the differences recorded after traversing all possible combinations of left and right sub-paths during the execution of the second matching strategy. It is used to characterize the matching accuracy that can be achieved between visual observation and map topology under the current road structure, and can be represented by min_lane_cnt_error.

[0090] Optionally, if the current road environment exhibits bilateral sub-path expansion characteristics (i.e., both the left and right sub-path cumulative sequences in the navigation map are not empty), the matching strategy is activated as the second matching strategy. The left pointer is initialized to point to the start position of the left cumulative sequence, and the right pointer to the end position of the right cumulative sequence, forming the initial state of the two-pointer search. An iterative loop is then entered, sequentially selecting each element in the left and right cumulative sequences to form ordered pairs, achieving comprehensive coverage of all possible left and right sub-path combinations. This traversal process does not rely on the absolute coordinates of the vehicle's actual position, but only on the map's pre-defined topology and lane number modeling, ensuring semantic consistency and scene independence. The traversal order follows the two-pointer optimization strategy: the left pointer increments from the minimum to the maximum cumulative value, and the right pointer decrements from the maximum to the minimum cumulative value, ensuring the search process maintains integrity while achieving convergence.

[0091] Optionally, in each round of traversal, the cumulative number of lanes on the left pointed to by the current left pointer is read, denoted as left_accum_cnt[i]. This value represents the total number of lanes accumulated by i sub-paths continuously extending from the end of the main road to the first direction (such as the left). At the same time, the cumulative number of lanes on the right pointed to by the current right pointer is read, denoted as right_accum_cnt[j]. This value represents the total number of lanes accumulated by j sub-paths continuously extending from the end of the main road to the second direction (such as the right). The two are arithmetically added to obtain the combined total number of lanes, which is determined as the sum. This can be determined by the following mathematical expression: sum = left_accum_cnt[i] + right_accum_cnt[j].

[0092] Optionally, the absolute deviation between the current error and the difference in number is calculated. This deviation is called the difference, and it is mathematically expressed as error = abs(sum - target). The current error is compared with the recorded minimum lane number error min_lane_cnt_error. If the current error is less than min_lane_cnt_error, min_lane_cnt_error is updated to the current error, and the corresponding left_accum_cnt[i] and right_accum_cnt[j] are recorded as the final matching combination. When the error is zero, it indicates that the current combination matches the difference in number, and the matching is successful. The matching result is the most suitable match, and the subsequent traversal is terminated. If the error is greater than zero but less than the original minimum lane number error, the system still updates the optimal matching result, but does not terminate the traversal to ensure that the global optimal solution is found.

[0093] Optionally, if the currently calculated difference is not better than the recorded minimum lane number error, the matching result is not updated, but it is determined whether the current pointer state is still within the valid range. If the left pointer has not yet reached the end of the left cumulative sequence and the right pointer has not yet fallen below zero, the pointer position is dynamically adjusted according to the relationship between sum and target. For example, if sum is less than target, it means that the total number of lanes in the current combination is insufficient, and the left contribution should be increased, so the left pointer moves forward. If sum is greater than target, it means that the total number of lanes in the current combination exceeds the requirement, and the right contribution should be reduced, so the right pointer moves backward. This pointer movement logic ensures that the search path always converges in the direction of reducing error, avoiding invalid traversal. The closed-loop process of traversal, summation, comparison, and pointer adjustment is repeated until the left pointer exceeds the range of the left sequence or the right pointer falls below zero, completing the search for each valid matching combination.

[0094] In this embodiment, the above method achieves high-precision, low-latency matching verification of bifurcated road structures through a second matching strategy. By using dynamic traversal with dual pointers and error minimization updates, it not only avoids the false alarm problem of "equal number of lanes" judgment in related technologies, but also outputs quantifiable matching errors and suitable matching combinations, providing continuous and gradient confidence input for downstream positioning fusion modules.

[0095] As an optional implementation, the evaluation result is determined based on the matching result, including: in response to the matching result that the number of lanes in the target lane concentration matches the number difference, the evaluation result is determined to be that the accuracy is greater than or equal to the accuracy threshold; in response to the matching result that the number of lanes in the target lane concentration does not match the number difference, the evaluation result is determined to be that the accuracy is less than the accuracy threshold.

[0096] In this embodiment, during the process of determining the evaluation result based on the matching results, if the matching result shows that the number of lanes in the target lane cluster matches the difference in number, then the evaluation result can be determined to be that the accuracy is greater than or equal to the accuracy threshold. Conversely, if the matching result shows that the number of lanes in the target lane cluster does not match the difference in number, then the evaluation result can be determined to be that the accuracy is less than the accuracy threshold. The accuracy threshold refers to the semantic discrimination boundary value used to classify the location reliability level when evaluating the consistency between visual perception and navigation map lane topology based on the matching results. Essentially, it is a system-preset quantitative standard reflecting the lower limit of the acceptability of the positioning result.

[0097] Optionally, after completing the matching verification between the visually perceived number of lanes and the number of main lanes and sub-path combinations on the navigation map, if the judgment result is "matched," meaning that the visually observed lane increment can be completely and reasonably explained by the cumulative structure of the main road and at least one sub-path in the navigation map, and the minimum lane number error is zero or within an acceptable range, then the matching result is explicitly marked as "matched." At this time, the system enters the evaluation result generation logic, and according to the preset mapping rules, regards the "matched" state as a high semantic consistency event, determining that the current positioning state has sufficient road structure support. The system directly sets the positioning accuracy to be greater than or equal to the accuracy threshold, where the accuracy threshold is the confidence lower limit set by the system based on historical scene data and safety requirements, representing that the positioning result has a trustworthy structural consistency. This accuracy value is usually represented in the range of 0.8 to 1.0 within the system, corresponding to a high-confidence positioning state, indicating that the current lane position of the vehicle can be accurately restored by the map topology, and can be used to support functional modules with strict requirements for lateral accuracy, such as lane keeping, lane change assist, and path planning, without needing to reduce weight or trigger redundant verification. The accuracy value is not a fixed number, but is logically derived based on the semantic integrity reflected by the matching results. In essence, it transforms the qualitative judgment of "structural consistency" into the quantitative assessment of "location credibility", thereby realizing the semantic expression of the quality of the location results.

[0098] Optionally, when the matching verification result shows "no match," meaning that the visually perceived number of lanes cannot semantically agree with the combination of the main road and all optional sub-paths on the map, manifested as the minimum lane number error exceeding the system's tolerance range, or no sub-path combination can reasonably explain the observation increment, the system determines that the current positioning state has a structural conflict or perception anomaly. At this time, the system enters a conservative evaluation mode, explicitly setting the positioning accuracy to be less than an accuracy threshold. This threshold serves as the system's preset confidence threshold, typically set to 0.6 to 0.7, representing that the positioning result has lost reliable topological support. This decrease in accuracy does not mean that positioning has completely failed, but rather indicates that when... The previous results are uncertain, so their weight needs to be reduced in the fusion positioning framework, or the enhancement correction mechanism of other sensors (such as GNSS, IMU, wheel speedometer) needs to be triggered to prevent the risk of lateral drift caused by outdated maps, lane line occlusion, or visual misdetection. This mechanism ensures that when the system faces structural anomalies, it will not blindly accept visual or map information, but will achieve a safe transition through confidence degradation, providing a basis for decision-making for takeover or deceleration strategies. The determination that the accuracy is less than the accuracy threshold is a conservative assessment made under the premise that the system confirms "no reasonable topological interpretation". Its purpose is to avoid positioning drift caused by semantic inconsistency, thereby establishing an active protection mechanism at the functional safety level.

[0099] In this embodiment, the method described above achieves a systematic and quantifiable transformation from structural semantic matching to positioning quality assessment by mapping the matching results to an accuracy threshold. This process transforms the original binary "match / non-match" judgment into a continuous positioning accuracy score, enabling the autonomous driving system to dynamically adjust its dependence on visual positioning results based on the semantic integrity of the road structure, avoiding the "one-size-fits-all" positioning failure handling in related technologies. When a match is successful, the accuracy is greater than or equal to the accuracy threshold, allowing for high-reliability operation; when a match is unsuccessful, the accuracy is less than the accuracy threshold, allowing the system to smoothly degrade, ensuring functional continuity and improving safety redundancy.

[0100] As an optional implementation, the positioning result includes a sensing positioning result, and the method further includes: in response to an evaluation result that the accuracy is greater than or equal to an accuracy threshold, adjusting the current weight of the sensing positioning result to a first target weight, wherein the first target weight is greater than the current weight; in response to an evaluation result that the accuracy is less than the accuracy threshold, performing one of the following processing operations on the sensing positioning result: adjusting the weight of the sensing positioning result to a second target weight, wherein the second target weight is less than the current weight; and smoothing the sensing positioning result.

[0101] In this embodiment, if the evaluation result shows an accuracy greater than or equal to an accuracy threshold, the current weight of the sensing evaluation result can be adjusted to the first target weight. Conversely, if the evaluation result shows an accuracy less than the accuracy threshold, the weight of the sensing positioning result can be adjusted to the second target weight. The sensing positioning result can also be smoothed. Here, the sensing positioning result refers to the relative position and heading information of the vehicle in the current road coordinate system, calculated by the vehicle's visual perception system based on static elements in the road environment (such as lane lines, curbs, traffic signs, etc.). Essentially, it is a relative positioning output based on visual semantic features. The aforementioned sensing positioning result can also be called a visual positioning result.

[0102] Optionally, the aforementioned first target weight can refer to the priority coefficient assigned to the sensing and positioning result for participating in multi-sensor fusion positioning calculation when the evaluation result shows an accuracy greater than or equal to an accuracy threshold. This first target weight is higher than the current weight of the sensing and positioning result under normal conditions. The aforementioned first target weight can also be called a high confidence weight, an enhanced confidence coefficient, or an optimized confidence factor. The aforementioned second target weight can also refer to the reduced priority coefficient assigned to the sensing and positioning result for participating in multi-sensor fusion positioning calculation when the evaluation result shows an accuracy less than an accuracy threshold. This second target weight is lower than the current weight of the sensing and positioning result under normal conditions.

[0103] Optionally, the aforementioned smoothing process can refer to the temporal filtering and state constraint operations performed on the sensing and positioning results when the evaluation result shows an accuracy less than the accuracy threshold. Essentially, it compensates for and suppresses current abnormal outputs using historical state information to reduce the risk of instantaneous jitter and abrupt changes in the positioning results. This smoothing process can be for trajectory stability correction, positioning noise suppression, or state continuity enhancement.

[0104] Optionally, when the evaluation result of the current positioning status is determined to be that the accuracy is greater than or equal to the accuracy threshold, it means that the number of lanes perceived by vision and the topology of the navigation map have achieved semantic consistency, and the sensing positioning result has sufficient structural support and environmental interpretability. At this time, the high confidence enhancement process is entered, and the current weight of the sensing positioning result in the fusion algorithm is increased to the first target weight. The above adjustment is not based on the original sensor noise level, but on the topological consistency confirmed by semantic matching verification. Essentially, it increases its contribution ratio in the fusion model, and at this moment, it prioritizes the use of visual positioning information for lateral position correction. The introduction of the first target weight allows the high-precision, high-dynamic-response visual positioning result to play a full role, thereby improving the positioning accuracy and response speed of the vehicle in scenarios with clear lane lines and well-defined structures, and realizing the proactive optimization of positioning performance. The above first target weight can be between 0.8 and 1.0, which is higher than the default weight value (current weight) of the sensing positioning result under normal conditions.

[0105] Optionally, when the evaluation result of the current positioning status is determined to be less than the accuracy threshold, it means that there is a semantic inconsistency between the number of lanes perceived visually and the map topology, and the sensing positioning result may have uncertainties such as false detection, missed detection, or map not being updated. At this time, the system enters the low confidence suppression process, reducing the current weight of the sensing positioning result in the fusion algorithm to the second target weight. This weight is a system-preset suppression confidence factor used to characterize low semantic confidence, and its value range can be between 0.1 and 0.4, significantly lower than the default weight and the first target weight. This adjustment reduces the influence of the visual positioning result in the fusion calculation, preventing it from dominating the positioning output, thereby reducing the risk of global positioning drift caused by local perception errors. The setting of the second target weight does not deny the physical validity of visual information, but rather acknowledges its insufficient semantic interpretability. By reducing the weight, a conservative strategy of "preserving information but weakening dominance" is achieved, ensuring that the system can still utilize the local information of the visual output, while handing over the dominance to the Global Navigation Satellite System (GNSS). Systems, or simply GNSS, inertial measurement units, or wheel speedometers, which possess global or kinematic stability, enable safer fusion decisions.

[0106] Optionally, if the current positioning state is determined to have semantic inconsistencies but is not completely invalid, to avoid instantaneous jumps or oscillations in the positioning output, a smoothing operation is performed while reducing the weights. This operation is essentially a dynamic state correction mechanism that introduces temporal continuity constraints. The smoothing process uses historical state information to filter and compensate the current sensing and positioning results. For example, it can be performed using first-order low-pass filtering, Kalman filtering, kinematic model-based prediction correction, or trajectory optimization algorithms based on velocity and acceleration constraints. The above process does not change the semantic meaning of the sensing and positioning results, but rather suppresses high-frequency noise and abnormal pulses to make parameters such as vehicle lateral offset and heading angle show a more continuous trend that conforms to the laws of physical motion. The purpose of the smoothing process is to eliminate positioning jitter caused by brief occlusion, sudden changes in illumination, or local false detections without completely discarding visual information, so that the system can still maintain a stable vehicle control response in the transition zone of structural inconsistencies, thereby improving ride comfort and avoiding frequent system takeovers. This processing, together with the weight reduction, constitutes a dual protection mechanism of "weight reduction + buffering" to ensure that the system still has functional continuity in a low-confidence state.

[0107] In the embodiments of this application, when semantic consistency is high, the above method releases the high-precision potential of visual positioning by increasing the weight to the first target weight; when semantic consistency is low, positioning risks are actively avoided by decreasing the weight to the second target weight or performing smoothing processing; this mechanism does not rely on additional sensor hardware and can achieve adaptive optimization of positioning performance based on software algorithms and logical judgments.

[0108] The technical solutions of the embodiments of this application will be illustrated below with reference to preferred embodiments.

[0109] Currently, in autonomous driving systems, achieving high-precision and robust vehicle positioning is fundamental to ensuring driving safety and functional implementation. Mainstream solutions typically employ multi-sensor fusion positioning technology, with the fusion of visual perception and navigation maps being a key technological approach.

[0110] Navigation map positioning: Navigation maps, also known as standard-precision maps, provide information such as road-level topological connections, number of lanes, geometry, and traffic rules. Using sensors such as the Global Positioning System (GPS), inertial measurement units (INS), and wheel speed sensors, vehicles can be matched to a specific road on the map. The advantage of this method is that it provides global, semantic location information.

[0111] Visual perception and positioning: By identifying static elements on the road such as lane lines, curbs, and traffic signs using onboard cameras, the lateral offset and heading angle of the vehicle relative to these elements can be calculated. Visual positioning can provide extremely high relative accuracy (down to the centimeter level) in a short time, which is key to achieving lane keeping and lane-level lateral control.

[0112] However, the solutions of the aforementioned technologies have the following significant limitations when facing complex driving scenarios in the real world.

[0113] Insufficient capability to handle lane number changes: Existing methods rely on a binary judgment of "visual lane number == map lane number." However, when a vehicle is about to enter an intersection, merging zone, or the start / end of a bus lane, the visually perceived number of lanes will naturally be more or less than the number of lanes on the main road in the map. In this case, a simple equality judgment will incorrectly conclude that "localization has failed," when in fact this is just a normal transition in road structure.

[0114] Lack of quantitative assessment and fine-grained matching: When judging mismatches, related technologies can only provide a "yes" or "no" conclusion, without providing quantitative information such as "the degree of mismatch" or "which sub-path is most likely to be matched." This prevents downstream systems from making tiered decisions based on the severity of the mismatch (such as reducing usage weight or selectively ignoring it), forcing them to adopt a conservative "one-size-fits-all" strategy, frequently triggering unnecessary functional restrictions or takeovers, severely impacting the continuity of the autonomous driving system and the user experience.

[0115] In summary, there is a prominent contradiction in the relevant technologies: on the one hand, it is necessary to integrate vision and maps to improve positioning reliability; on the other hand, the existing fusion verification methods are too simple and cannot effectively cope with the common scenario of dynamic changes in road structure, resulting in poor system robustness and high false alarm rate.

[0116] In this application embodiment, the aim is to address the problem in autonomous driving systems where visual positioning information and standard navigation map information are difficult to accurately match under complex road structures (such as intersections, ramps, and forks / merges), leading to low positioning reliability and erroneous system takeover. Specifically, the prior art has the following drawbacks.

[0117] Insufficient accuracy of navigation maps: Standard navigation maps typically only provide road-level topological and geometric information, and their absolute positioning accuracy is low, which cannot meet the needs of lane-level lateral control for autonomous driving. In particular, they are prone to lateral positioning deviations at complex intersections, which can lead to unnecessary safety takeovers.

[0118] Incomplete visual positioning information: Although visual perception (such as lane line detection) can provide high-precision lateral relative positioning, it often lacks sufficient and uniquely identifiable references in the longitudinal direction, resulting in fuzzy or insufficient accuracy in longitudinal positioning.

[0119] Simple matching has poor robustness: Existing simple matching methods (such as directly comparing whether the number of lanes is equal) are prone to mismatch or matching failure in scenarios where the road structure changes (such as the number of lanes increases or decreases). They cannot provide continuous and reliable positioning verification results and cannot handle complex scenarios such as bifurcations and merging.

[0120] Therefore, the technical problem to be solved by the embodiments of this application is: how to provide a matching verification method that can accurately and efficiently verify whether the lane number information perceived by vision is consistent with the lane topology information of the navigation map, and can handle the dynamic change scenario of lane number, so as to improve the robustness and reliability of fusion positioning.

[0121] The methods of the embodiments of this application will be further illustrated below.

[0122] Figure 2 This is a schematic diagram of a lane matching device for a visual and navigation map according to an embodiment of this application, such as... Figure 2 As shown, the device 20 may include an acquisition module 202, a first judgment module 204, a second judgment module 206, a sub-path matching module 208, and a result output module 210. The acquisition module 202 acquires the visually perceived lane number information at the current moment, as well as the main lane number information and the lane number information of the corresponding road segment in the navigation map. The first judgment module 204 determines whether the visually perceived lane number is equal to the main lane number in the navigation map; if they are equal, it outputs a successful match. The second judgment module 206 calculates the difference between the visually perceived lane number and the main lane number if the visually perceived lane number is greater than the main lane number in the navigation map, using this difference as the number of supplementary lanes to be matched. The sub-path matching module 208, based on the left and right sub-path lane number sequences provided in the navigation map, searches for combinations of one or more sub-paths through cumulative calculation and optimized search, minimizing the error between the cumulative lane number and the supplementary lane number. The result output module 210 outputs the matching result, which includes a matching flag, the minimum lane number error, and the number of matched left and right sub-paths.

[0123] The optimized search of the sub-path matching module 208 employs a two-pointer algorithm, starting from the head of the left sub-path cumulative sequence and the tail of the right sub-path cumulative sequence, respectively, to find the combination that makes the total number of lanes closest to the target difference in linear time complexity. When a single-sided sub-path exists, a binary search algorithm is used to quickly locate the value closest to the target difference in the cumulative sequence.

[0124] The method described in this application, by introducing and utilizing sub-path information from the navigation map, can effectively handle complex scenarios such as road forks, merging, and lane number increases or decreases. It returns not only a binary judgment of "match" or "mismatch," but also minimized lane number error and specific sub-path matching schemes, providing richer and more quantifiable confidence information for the downstream positioning fusion module. The core matching algorithm is ingeniously designed, using efficient algorithms such as two pointers and binary search to control the time complexity to O(n), ensuring the algorithm's real-time performance. Actual testing shows that a matching verification can be completed within 0.1 milliseconds, with extremely low computational resource consumption on the vehicle computing platform. Through high-precision vision-map matching verification, potential risks of positioning inconsistencies can be detected in a timely manner, providing a crucial layer of redundant verification for the autonomous driving system. This helps reduce lateral positioning drift caused by inaccurate maps or visual misdetections, thereby effectively reducing unnecessary takeovers triggered in complex intersections and other scenarios, improving driving safety and passenger experience. The above method can be implemented based on software algorithms, without relying on additional expensive sensors or high-precision map data. It reuses existing visual perception modules and standard navigation maps, significantly improving the overall performance of the positioning system at extremely low marginal cost.

[0125] Figure 3 This is a flowchart of a lane matching method for visual and navigation maps according to an embodiment of this application, such as... Figure 3 As shown, the method may include the following steps.

[0126] Step S301: Obtain the number of lanes perceived visually and the lane information on the map.

[0127] In this embodiment, the visual lane variable information (VisionLaneVarInfo) and the navigation map lane variable information (SdVarInfoWithSub) are obtained at the current moment. VisionLaneVarInfo comes from the visual perception model, and SdVarInfoWithSub comes from the map data management module. The aforementioned visual lane variable information (VisionLaneVarInfo) includes the number of lanes detected by the visual perception module at the current moment (lane_count). This information is output in real time by the forward-facing camera combined with a deep learning model.

[0128] Optionally, SdVarInfoWithSub contains lane information of the current location on the main navigation path, extracted from a standard navigation map:

[0129] lane_count, the number of lanes on the main path (the main road in the current direction of travel); left_subpath_counts, a list of integers representing the sequence of lane counts for each subpath that may branch off on the left (such as left-turn lanes and left-turn ramps); right_subpath_counts, a list of integers representing the sequence of lane counts for each subpath that may branch off on the right (such as right-turn lanes and right-turn ramps).

[0130] Step S302: Determine the matching type.

[0131] In this embodiment, if the number of visually perceived lanes equals the number of lanes on the map, the matching type is a perfect match. If the number of visually perceived lanes is less than the number of lanes on the map, the matching type is insufficient visual lanes. If the number of visually perceived lanes is greater than the number of lanes on the map, then step S303 can be executed.

[0132] Step S303, Sub-lane matching algorithm.

[0133] In this embodiment, a set of continuous sub-path combinations is found by systematically searching the predefined left and right sub-path lane count sequences in the navigation map, such that the cumulative lane count matches the visually observed excess as closely as possible. The algorithm constructs arrays of cumulative lane counts for the left and right sub-paths, employs efficient search strategies such as two-pointer or binary search, and traverses all possible left and right sub-path combinations within linear or logarithmic time complexity. It calculates the minimum absolute error between the calculated combination and the target value, and outputs the cumulative values ​​of the left and right sub-paths and the final error of the optimal matching combination.

[0134] Figure 4 This is a flowchart of a sub-lane matching method according to an embodiment of this application, such as... Figure 4 As shown, the method may include the following steps.

[0135] Step S401: Determine the number of sub-lanes on the map.

[0136] In this embodiment, the VisionLaneVarInfo and SdVarInfoWithSub at the current moment are obtained.

[0137] Step S402: Construct the cumulative lane array.

[0138] In this embodiment, the number of visual lanes is compared with the number of main lanes on the map. If they are equal, a match is considered successful. If the number of visual lanes is less than the number of main lanes on the map, a mismatch is considered, with the error being the difference between the visual and map lane counts. In this case, the visual system missed detecting a lane. If the number of visual lanes is greater than the number of main lanes on the map, a sub-lane supplementary matching algorithm is initiated.

[0139] Optionally, it can be determined as a mismatch. In this case, the error is the difference between the number of lanes in the visual representation and the map. Construct a lane accumulation array.

[0140] For example, the original data of the left sub-path is [1,2,1].

[0141] Cumulative calculation process:

[0142] Step 1: [1] ← First element;

[0143] Step 2: [1,3]←1+2=3;

[0144] Step 3: [1,3,4]←3+1=4;

[0145] Final: left_accum_cnt=[1,3,4].

[0146] For another example, the original data of the right subpath is [2,1].

[0147] Cumulative calculation process:

[0148] Step 1: [2] ← First element;

[0149] Step 2: [2,3]←2+1=3;

[0150] Final: right_accum_cnt=[2,3].

[0151] Optionally, if there exists a value in left_accum_cnt and right_accum_cnt that is equal to possible_sub_lane_cnt, it indicates that the visual lane can be matched by a set of continuous sub-paths on one side, and the match is considered successful.

[0152] Step S403: Is the sub-path bilateral?

[0153] In this embodiment, if the sub-path is bilateral, step S404 can be executed; otherwise, if the sub-path is unilateral, step S405 can be executed.

[0154] Step S404: Execute the two-sided matching strategy.

[0155] In this embodiment, when sub-paths exist on both sides, this application uses an efficient two-pointer algorithm to search for the optimal combination. Figure 5 (As shown).

[0156] Step S405: Execute the one-sided matching strategy.

[0157] In this embodiment, when only one side of the dual-sided sub-path exists, dual-sided combination matching is not possible, and single-sided lane matching is used. In this case, the value closest to `possible_sub_lane_cnt` is searched in the single-sided cumulative sequence. A binary search is performed in the cumulative sequence to find the first iterator not less than `possible_sub_lane_cnt`. The absolute errors between the value pointed to by this iterator and the value pointed to by the previous iterator (if it exists) and the target value are calculated. The smaller error is taken as `min_lane_cnt_error`, and it is determined as a mismatch (`is_matched=false`), but a minimum error value `min_lane_cnt_error` is provided.

[0158] Figure 5 This is a flowchart of a two-pointer search algorithm according to an embodiment of this application, such as... Figure 5 As shown, the method may include the following steps.

[0159] Step S501: Begin bilateral search.

[0160] In this embodiment, the purpose of initiating the dual-pointer search algorithm is to efficiently find a combination of left and right sub-paths through bidirectional collaborative exploration between the accumulated lane sequence of the left sub-path and the accumulated lane sequence of the right sub-path, so that the sum of them is closest to the difference between the visually perceived number of lanes and the number of lanes on the main road of the map (i.e., the target difference), thereby achieving accurate matching of complex road structure changes.

[0161] Step S502: Initialize pointer i=0, j=right_size-1.

[0162] In this embodiment, the starting positions of two search pointers are set: the left pointer i points to the first element (minimum cumulative value) of the left cumulative sequence, and the right pointer j points to the last element (maximum cumulative value) of the right cumulative sequence. This ensures that the search starts from the extreme combination of the two ends, providing a structural basis for subsequent linear convergence.

[0163] Step S503, initialize the best record: Left_cnt=0, right_cnt=0, min_lane_cnt_error=INT_MAX.

[0164] In this embodiment, an initial state is set for storing the subsequent optimal solution. The current optimal left and right sub-path combination is assigned an empty value (0), and the minimum error is initialized to the maximum integer value to ensure that any subsequent valid error value can be updated, thus avoiding the omission of the optimal solution due to interference from the initial value.

[0165] Step S504, i<left_sizeANDj> =0?

[0166] In this embodiment, if i < left_size AND j >= 0 are satisfied, step S505 can be executed; otherwise, step S515 can be executed.

[0167] Optionally, it is judged whether it is the first iteration of the search loop, which is used to distinguish the processing logics of the initial state and subsequent iterations, ensure that the first combined calculation is correctly executed when first entering, and avoid repeated or omitted initialization actions.

[0168] Step S505, calculate the current combination: Sum = left_acc[i] + right_acc[j].

[0169] In this embodiment, adding the two cumulative lane values pointed to by the current left pointer and right pointer to form the total lane number of the current combination, which is used as a candidate matching result for comparison with the target difference, is the core operation of each iteration of the algorithm.

[0170] Step S506, error < min_lane_cnt_error?

[0171] In this embodiment, if error < min_lane_cnt_error is satisfied, step S507 can be executed; otherwise, step S508 is executed.

[0172] Step S507, update the best record: min_lane_cnt_error = error, left_cnt = left_acc[i], right_cnt = right_acc[j].

[0173] In this embodiment, storing the current error value and the corresponding cumulative values of the left and right sub-paths into the best record as a new matching benchmark, providing a reference basis for subsequent comparisons, and ensuring that the output of the algorithm is globally optimal rather than locally optimal.

[0174] Step S508, maintain the current best record.

[0175] In this embodiment, it is executed when the current error is not better than the historical optimum, without any update, maintaining the existing optimum solution, ensuring the stability and continuity of the search process, and avoiding interference with the matching result due to fluctuations.

[0176] Step S509, error = 0?

[0177] In this embodiment, if error = 0 is satisfied, step S510 can be executed; otherwise, step S511 can be executed.

[0178] Step S510, return the exact matching result, is_matched = true, error = 0.

[0179] In this embodiment, when a perfect match is detected, the search is immediately terminated and a match success flag and a zero-error result are output, indicating that the current road structure change can be perfectly explained by the map sub-path combination and no further optimization is required. The system can directly confirm a high-confidence positioning.

[0180] Step S511, compare sum with target.

[0181] In this embodiment, if sum == target, step S514 can be executed; if sum < target, step S512 can be executed; if sum > target, step S513 can be executed.

[0182] Step S512, increment the left pointer i++.

[0183] In this embodiment, it is executed when the total number of lanes in the current combination is less than the target value, indicating that the contribution of the left side is insufficient and the cumulative number of lanes of the left sub-path needs to be expanded. Therefore, the left pointer is moved to the right to select a larger cumulative value on the left to approach the target. If sum < target, it means that the total number of lanes is insufficient and the contribution of the left side should be increased, so i++.

[0184] Step S513, decrement the right pointer j--.

[0185] In this embodiment, it is executed when the total number of lanes in the current combination is greater than the target value, indicating that the contribution of the right side is excessive and the cumulative number of lanes of the right sub-path needs to be reduced. Therefore, the right pointer is moved to the left to select a smaller cumulative value on the right to approach the target. If sum > target, it means that the total number of lanes is excessive and the contribution of the right side should be reduced, so j--.

[0186] Step S514, an exact match has been processed.

[0187] In this embodiment, when it is marked that the optimal matching process has been completed when error = 0, it indicates that no further pointer movement or error comparison is required in the future and the search can be safely terminated.

[0188] Step S515, return the approximate matching result, is_matched = false, min_lane_cnt_error = the best error value.

[0189] In this embodiment, loop end processing: If a perfect match with an error of 0 is not found, the loop is terminated. In the result returned at this time, is_matched = false, but min_lane_cnt_error stores the minimum error found, and left_subpath_cnt and right_subpath_cnt store the cumulative number of lanes of the left and right sub-paths corresponding to this minimum error.

[0190] Optionally, downstream systems can use the output matching results to make decisions: if `is_matched` is true, the visual-map matching verification is successful, and the visual positioning result can be given a higher confidence weight. If `is_matched` is false, but `min_lane_cnt_error` is small (e.g., error of 1), it indicates that the system may be in a transition zone where the number of lanes changes or that there is a slight perception bias. In this case, the visual positioning result can be used with reduced weight or smoothed out, rather than being discarded directly. If `min_lane_cnt_error` is large, it indicates a serious mismatch, and the result should be discarded directly.

[0191] In summary, through the above-described implementation methods, this application achieves refined and quantitative matching verification of visual and map lane number information. Its technical effects and performance advantages are as follows: Comprehensive scene coverage: By explicitly modeling and searching map sub-paths, this method can correctly handle complex intersection scenarios with lane number changes caused by bifurcation and merging, significantly improving the robustness and scene adaptability of the matching verification. Quantitative output: `min_lane_cnt_error` provides a continuous confidence metric for downstream systems, enabling smoother and more intelligent decision-making and avoiding the "either / or" problem of traditional binary judgments. Extremely high computational efficiency: The core algorithm of this embodiment has a time complexity of O(L+R), where L and R are the number of left and right sub-paths, respectively. Single-sided search uses O(log N) binary search, and double-sided combined search uses O(L+R) two-pointer linear scanning. This allows the entire matching verification process to be completed in sub-milliseconds (<0.1ms) even at complex intersections with many sub-paths, meeting the stringent real-time requirements of autonomous driving systems. It is simple and reliable, with clear algorithm logic. Based on standard data structures and algorithm implementation, it is easy to integrate into existing autonomous driving software stacks and has high engineering practical value.

[0192] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0193] According to another aspect of the embodiments of this application, corresponding to the embodiments of the above-described vehicle positioning result evaluation method, this specification also provides a vehicle positioning result evaluation device.

[0194] Figure 6This is a schematic diagram of a vehicle positioning result evaluation device according to an embodiment of this application, as shown below. Figure 6 As shown, the vehicle positioning result evaluation device 60 may include: an acquisition module 602, a determination module 604, and an evaluation module 606. The acquisition module 602 is used to acquire a first number of lanes sensed by the vehicle, and a second number of lanes on the target road where the vehicle is located, obtained from the vehicle's map information. The determination module 604 is used to determine a target lane set for the vehicle from lanes on roads surrounding the target road in response to the first number being greater than the second number. The evaluation module 606 is used to evaluate the vehicle's positioning result in the map information based on the number of lanes in the target lane set and the difference between the first and second numbers, obtaining an evaluation result, wherein the evaluation result represents the accuracy of the positioning result.

[0195] Embodiments of this application also provide a vehicle, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods described in various embodiments of this application when it runs.

[0196] Figure 7 This is a schematic diagram of an electronic device according to an embodiment of this application, such as... Figure 7 As shown, the electronic device 70 includes: a memory 701 storing an executable program; and a processor 702 for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0197] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0198] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0199] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0200] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this application.

[0201] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0202] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0203] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0204] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0205] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0206] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for evaluating vehicle positioning results, characterized in that, include: Obtain a first number of lanes sensed by the vehicle, and obtain a second number of lanes on the target road where the vehicle is located from the vehicle's map information; In response to the first quantity being greater than the second quantity, a target lane set for the vehicle is determined from the lanes on the roads surrounding the target road; Based on the number of lanes in the target lane and the difference between the first number and the second number, the positioning result of the vehicle in the map information is evaluated to obtain an evaluation result, wherein the evaluation result is used to represent the accuracy of the positioning result.

2. The method according to claim 1, characterized in that, In response to the first quantity being greater than the second quantity, determining the target lane set of the vehicle from lanes on roads surrounding the target road includes: In response to the first quantity being greater than the second quantity, an accumulation operation is performed on lanes on roads surrounding the target road using the lane type of the lane to obtain the target lane set, wherein the lane type is used to characterize the positional relationship between the lane and the vehicle.

3. The method according to claim 2, characterized in that, The step of using the lane type of the lane to perform an accumulation operation on lanes on roads surrounding the target road to obtain the target lane set includes: According to the lane type, a first lane sequence and a second lane sequence are determined from the lanes on the roads surrounding the target road, wherein the first lane sequence is used to represent the number of lanes contained on the road branching off from the target road in a first direction, and the second lane sequence is used to represent the number of lanes contained on the road branching off from the target road in a second direction. The accumulation operation is performed on the number of lanes in the first lane sequence according to the order of the roads branching in the first direction, and the accumulation operation is performed on the number of lanes in the second lane sequence according to the order of the roads branching in the second direction, to obtain the target lane set.

4. The method according to claim 1, characterized in that, The location result of the vehicle in the map information is evaluated based on the number of lanes concentrated in the target lane and the difference between the first number and the second number, and the evaluation result is obtained, including: A matching strategy adapted to the lane type of the target lane group is invoked to match the number of lanes in the target lane group with the number difference to obtain a matching result. The lane type is used to characterize the positional relationship between the lane and the vehicle, and the matching result is used to indicate the degree of difference between the number of lanes in the target lane group and the number difference. Based on the matching results, the evaluation results are determined, wherein the matching results and the evaluation results are positively correlated.

5. The method according to claim 4, characterized in that, The matching strategy includes a first matching strategy, which represents a rule for matching the number of lanes in the target lane cluster on roads branching off in a first direction from the target road with the difference in the number of lanes. The step of invoking a matching strategy adapted to the lane type of the target lane cluster, matching the number of lanes in the target lane cluster with the difference in the number of lanes, and obtaining a matching result includes: In response to the matching strategy being the first matching strategy, the number of lanes in the first direction in the target lane set is determined; In response to the number of lanes in the first direction, if the difference between the number and the difference is less than a difference threshold, the matching result is determined to be the number of lanes in the target lane set that matches the difference between the number and the difference.

6. The method according to claim 4, characterized in that, The matching strategy includes a second matching strategy, which represents a rule for matching the number of lanes in the target lane cluster with the difference in number on roads branching from the target road into the first and second directions. The step of invoking a matching strategy adapted to the lane type of the target lane cluster to match the number of lanes in the target lane cluster with the difference in number to obtain a matching result includes: In response to the matching strategy being the second matching strategy, the number of lanes in the first direction and the number of lanes in the second direction in the target lane set are traversed. During the traversal, the sum between the number of lanes in the first direction traversed and the number of lanes in the second direction traversed is determined; In response to the difference between the sum and the quantity difference being less than the minimum lane number error, the matching result is determined to be the number of lanes in the target lane set, which matches the quantity difference. The method further includes: In response to the difference being greater than or equal to the minimum lane number error, return to the starting step of the following steps until the number of lanes in the first direction and the number of lanes in the second direction in the target lane set have been traversed: The number of lanes in the first direction and the number of lanes in the second direction in the target lane set are traversed.

7. The method according to claim 4, characterized in that, Determining the evaluation result based on the matching result includes: In response to the matching result being that the number of lanes in the target lane concentration matches the difference in number, the evaluation result is determined to be that the accuracy is greater than or equal to the accuracy threshold; In response to the matching result that the number of lanes in the target lane concentration does not match the number difference, the evaluation result is determined to be that the accuracy is less than the accuracy threshold.

8. The method according to any one of claims 1 to 7, characterized in that, The positioning result includes the sensing positioning result, and the method further includes: In response to the evaluation result that the accuracy is greater than or equal to the accuracy threshold, the current weight of the sensing and positioning result is adjusted to a first target weight, wherein the first target weight is greater than the current weight; In response to the evaluation result that the accuracy is less than the accuracy threshold, one of the following processing operations is performed on the sensing and positioning result: The weight of the sensing and positioning result is adjusted to a second target weight, wherein the second target weight is less than the current weight; The sensing and positioning results are then smoothed.

9. A vehicle positioning result evaluation device, characterized in that, include: The acquisition module is used to acquire a first number of lanes sensed by the vehicle, and to acquire a second number of lanes on the target road where the vehicle is located from the vehicle's map information. A determining module is configured to, in response to the first quantity being greater than the second quantity, determine the target lane set of the vehicle from the lanes on roads surrounding the target road; The evaluation module is used to evaluate the positioning result of the vehicle in the map information based on the number of lanes in the target lane and the difference between the first number and the second number, and to obtain an evaluation result, wherein the evaluation result is used to represent the accuracy of the positioning result.

10. A vehicle, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 8.

Images