Simulation parameter calibration methods, apparatus, equipment and storage media

By generating structured scene features and semantic vectors, calculating similarity, and verifying simulation parameters, the problem of insufficient timeliness and accuracy in the evaluation of highway control strategies is solved. This enables the rapid generation of adaptable simulation parameters and improves the evaluation effect of emergency control strategies.

CN123069658APending Publication Date: 2026-11-10ZHEJIANG EXPRESSWAY INFO ENG TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611221513.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-12
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies lack timeliness and accuracy in evaluating emergency response strategies for highways. Black-box optimization and iterative parameter calibration are time-consuming and difficult to respond quickly. Fixed parameters based on general or expert experience have weak cross-scenario generalization ability, and the reliability of simulation results is insufficient.

Method used

By acquiring multi-source trajectory observation data, structured scene features and scene semantic vectors are generated. The structured similarity and semantic similarity between the current scene and the preset simulation vector library are calculated. Similar historical scenes and their multi-vehicle simulation parameter sets are selected and verified. Finally, target simulation parameters are generated.

Benefits of technology

In the simulation of emergency control strategies for highways, the simulation parameters can be quickly output to shorten the parameter preparation cycle, improve the timeliness and accuracy of the control strategy evaluation, and ensure the adaptability of the simulation parameters to the current road network, traffic and event scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069658A_ABST
    Figure CN123069658A_ABST
Patent Text Reader

Abstract

This application discloses a simulation parameter calibration method, apparatus, device, and storage medium, relating to the field of intelligent traffic management technology. The simulation parameter calibration method of this application generates structured scene features and scene semantic vectors based on multi-source trajectory observation data of the current scene. By calculating structured similarity and semantic similarity, it filters out matching similar historical scenes and their corresponding multi-vehicle model simulation parameter sets from a preset simulation vector library. After verification, target simulation parameters are generated. In this way, in the scenario of highway emergency control strategy simulation, usable simulation parameters can be quickly output, ensuring the timeliness of control strategy evaluation. Simultaneously, through a two-dimensional matching and verification mechanism of structured numerical and semantic descriptions, the adaptability of simulation parameters to the current scene can be improved, enhancing the accuracy of control strategy evaluation results. Thus, it solves the problems of insufficient timeliness and accuracy in highway control strategy evaluation in existing technologies.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of intelligent traffic management technology, and in particular to simulation parameter calibration methods, devices, equipment, and storage media. Background Technology

[0002] With the development of intelligent traffic management technology, the scientific formulation and selection of emergency response strategies for highways heavily rely on microscopic traffic simulation for pre-assessment of strategy implementation effectiveness. Simulation analysis allows for the quantitative evaluation of the traffic efficiency and safety benefits of various control schemes, providing managers with a basis for decision-making.

[0003] To ensure simulation fitting accuracy, some techniques employ black-box optimization parameter calibration methods, adjusting model parameters through repeated simulation iterations. While this achieves high calibration accuracy, parameter preparation is time-consuming and cannot meet the rapid response requirements of sudden scenarios such as traffic accidents and road construction. Other techniques use general parameter sets or fixed parameters based on expert experience, enabling rapid simulation initiation, but they struggle to adapt to differences in road alignment and traffic composition across different road sections, exhibiting weak cross-scenario generalization ability and insufficient reliability of simulation results. Therefore, improving the timeliness and accuracy of highway management strategy evaluation has become an urgent problem to be solved. Summary of the Invention

[0004] This application provides a simulation parameter calibration method, apparatus, device, and storage medium to at least address the issues of insufficient timeliness and accuracy in highway control strategy assessment in related technologies.

[0005] This application provides a simulation parameter calibration method, including: Acquire multi-source trajectory observation data for the current scene; the multi-source trajectory observation data includes real-time vehicle trajectory data and historical vehicle trajectory data; Based on multi-source trajectory observation data, structured scene features and scene semantic vectors are generated; Based on structured scene features and scene semantic vectors, the structured similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library are calculated; where each historical scene corresponds to a multi-vehicle simulation parameter set; Based on the structural and semantic similarity between the current scenario and multiple historical scenarios in the preset simulation vector library, determine at least one similar historical scenario and the multi-vehicle simulation parameter set corresponding to the similar historical scenario; Verify multiple similar historical scenarios and the simulation parameter sets of multiple vehicle models corresponding to similar historical scenarios; Use at least one similar historical scene that passes the verification as the target historical scene; Target simulation parameters are generated based on the target historical scenario and the multi-vehicle simulation parameter set corresponding to the target historical scenario.

[0006] In one optional implementation, structured scene features and scene semantic vectors are generated based on multi-source trajectory observation data, including: Based on multi-source trajectory observation data, structured scene features are generated; among them, structured scene features include average speed, speed standard deviation, flow rate, density, average headway, lane change rate, proportion of large vehicles, observation time window, and congestion index. Based on multi-source trajectory observation data, determine the road network type, congestion status, vehicle type composition, and event type of the current scenario; The semantic vector of the scene is determined based on the road network type, congestion status, vehicle type composition, and event type of the current scene.

[0007] In one optional implementation, based on structured scene features and scene semantic vectors, the structural similarity and semantic similarity between the current scene and multiple historical scenes in a preset simulation vector library are calculated, including: Based on the structured scene features, the numerical similarity between the current scene and multiple historical scenes in the preset simulation vector library is calculated to obtain the structured similarity. Based on the scene semantic vector, the semantic similarity is calculated between the current scene and multiple historical scenes in the preset simulation vector library.

[0008] In one optional implementation, based on the structural and semantic similarity between the current scene and multiple historical scenes in a preset simulation vector library, at least one similar historical scene and a multi-vehicle simulation parameter set corresponding to the similar historical scene are determined, including: The structural similarity and semantic similarity are weighted and fused to obtain the matching score for each historical scene; Based on the ranking of matching scores, at least one similar historical scenario is selected from the preset simulation vector library, and the simulation parameter set of multiple vehicle models corresponding to the similar historical scenario is obtained.

[0009] In one optional implementation, the simulation parameter sets of multiple similar historical scenarios and corresponding vehicle models are validated, including: If the proportion of effective trajectory points, the missing rate of structured features, and the proportion of effective trajectory time windows meet the corresponding preset thresholds, determine whether the matching score of the top-ranked similar historical scene is not lower than the preset minimum similarity threshold. If the matching score of the first-ranked similar historical scene is not lower than the preset minimum similarity threshold, determine whether the difference in matching scores between the first and second-ranked similar historical scenes is not lower than the minimum interval threshold. If the difference in matching scores between the first and second most similar historical scenarios is not lower than the minimum interval threshold, the first most similar historical scenario is determined to pass the verification.

[0010] In one optional implementation, target simulation parameters are generated based on the target historical scene and the multi-vehicle simulation parameter set corresponding to the target historical scene, including: If the verification passes, perform the first fast simulation verification on the multi-vehicle simulation parameter set corresponding to the target historical scenario; If the first rapid simulation verification passes, the multi-vehicle simulation parameter set corresponding to the target historical scenario is determined as the target simulation parameter.

[0011] In one alternative implementation, the method further includes: If the matching score of the top-ranked similar historical scene is lower than the preset minimum similarity threshold, or the difference in matching scores between the top and second-ranked similar historical scenes is lower than the minimum interval threshold, a fast search is performed within the preset boundary using the multi-vehicle simulation parameter set corresponding to the top-ranked similar historical scene as the baseline to obtain the search results; wherein, the search results include at least one multi-vehicle simulation parameter set to be verified. Perform a second fast simulation verification on at least one set of simulation parameters for multiple vehicle models to be verified; If any set of simulation parameters for multiple vehicle models to be verified passes the second fast simulation verification, the set of simulation parameters for multiple vehicle models to be verified is determined as the target simulation parameters.

[0012] This application also provides a simulation parameter calibration device, including: The acquisition module is used to acquire multi-source trajectory observation data of the current scene; the multi-source trajectory observation data includes real-time vehicle trajectory data and historical vehicle trajectory data. The generation module is used to generate structured scene features and scene semantic vectors based on multi-source trajectory observation data; The calculation module is used to calculate the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library based on the structured scene features and scene semantic vectors; where each historical scene corresponds to a multi-vehicle simulation parameter set; The determination module is used to determine at least one similar historical scene and the multi-vehicle simulation parameter set corresponding to the similar historical scene based on the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library; The verification module is used to verify multiple similar historical scenarios and the simulation parameter sets of multiple vehicle models corresponding to similar historical scenarios; The scene determination module is used to select at least one similar historical scene that has passed the verification as the target historical scene. The parameter generation module is used to generate target simulation parameters based on the target historical scenario and the multi-vehicle simulation parameter set corresponding to the target historical scenario.

[0013] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described simulation parameter calibration methods.

[0014] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above-described simulation parameter calibration methods.

[0015] This application also provides a computer program product, including computer instructions for causing a computer to implement the steps of any of the above-described simulation parameter calibration methods.

[0016] In some embodiments of this application, structured scene features and scene semantic vectors are generated based on multi-source trajectory observation data of the current scene. By calculating structured similarity and semantic similarity, matching similar historical scenes and their corresponding multi-vehicle model simulation parameter sets are selected from a preset simulation vector library. After verification, target simulation parameters adapted to the current scene are generated. In this way, in the scenario of highway emergency control strategy simulation, usable simulation parameters can be quickly output, shortening the parameter preparation cycle and ensuring the timeliness of control strategy evaluation. Simultaneously, through a two-dimensional matching and verification mechanism of structured numerical and semantic descriptions, the adaptability of simulation parameters to the current road network, traffic, and event scenario can be improved, enhancing the accuracy of control strategy evaluation results. Thus, the problems of insufficient timeliness and accuracy in highway control strategy evaluation in the prior art can be solved. Attached Figure Description

[0017] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of a simulation parameter calibration method provided for some embodiments of this application; Figure 2 A schematic diagram of a simulation parameter calibration device provided for some embodiments of this application; Figure 3 A schematic diagram of the modules of an electronic device provided for some embodiments of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0020] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0021] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] In related technologies, the assessment and simulation of highway emergency response strategies often relies on micro-level traffic simulation to support decision-making. To ensure simulation accuracy, one approach uses black-box optimization and iterative parameter calibration, which offers high accuracy but time-consuming parameter preparation, making it difficult to meet the rapid response requirements of emergency scenarios. Another approach uses general or expert-based fixed parameters, enabling rapid simulation initiation but exhibiting weak cross-scenario generalization ability and insufficient reliability of the simulation results. Therefore, improving the timeliness and accuracy of highway control strategy assessment has become an urgent problem to be solved.

[0023] In view of this, this application provides a simulation parameter calibration method that can solve the above problems. The simulation parameter calibration method can be applied to electronic devices. Electronic devices may include, but are not limited to, tablet computers, laptops, desktop computers, servers, etc. (See also...) Figure 1 This is a flowchart illustrating a simulation parameter calibration method provided in some embodiments of this application. Figure 1 In this simulation parameter calibration method, the following steps are included: Step S101: Obtain multi-source trajectory observation data for the current scene.

[0024] Specifically, multi-source trajectory observation data refers to a collection of vehicle trajectory records from various sensing channels, including real-time vehicle trajectory data and historical vehicle trajectory data. This can include millimeter-wave radar points, vehicle trajectories recognized by video AI (Artificial Intelligence), floating car GPS (Global Positioning System) / BeiDou positioning, and roadside radar-visual integrated detection data. Real-time vehicle trajectory data refers to short-term sliding window vehicle movement records collected during the event response phase, including fields such as latitude and longitude, speed, direction of travel, vehicle type, and timestamp. Historical vehicle trajectory data refers to historical vehicle movement records pre-stored in a database, containing full trajectory data from past complete event cycles and traffic flow observation data under normal traffic conditions.

[0025] Optionally, the multi-source trajectory observation data only retains the trajectories within the simulation range and does not simulate the entire high-speed network. Instead, it takes the event space points as the endpoints, takes a certain distance upstream of the road network, and then performs the simulation on this road network.

[0026] Step S102: Generate structured scene features and scene semantic vectors based on multi-source trajectory observation data.

[0027] Specifically, structured scene features refer to a set of numerical indicators used to quantitatively describe the current operational status of traffic scenes. Each indicator can be directly calculated from trajectory data.

[0028] Specifically, scene semantic vectors refer to vectorized semantic codes used to represent the attribute features of the current scene, which can characterize the essential features of the scene at the attribute level.

[0029] In this embodiment, multi-source trajectory observation data is used as input. First, the structured scene features and scene semantic vectors are standardized, cleaned, and converted to a unified format. Then, features are extracted and encoded from the quantized numerical dimension and the attribute semantic dimension, respectively, to obtain the structured scene features and scene semantic vectors.

[0030] Step S103: Calculate the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library based on the structured scene features and scene semantic vectors.

[0031] Specifically, the preset simulation vector library refers to a pre-built and stored historical scene resource library. The library includes multiple historical scenes covering different road network types, different event types, and different traffic states. Each historical scene stores its scene identifier (ID), timestamp, road network type, structured scene features, scene semantic vector, simulation model configuration, and multi-vehicle simulation parameter set.

[0032] Specifically, the multi-vehicle simulation parameter set refers to the set of traffic simulation model parameters adapted to the traffic characteristics of corresponding historical scenarios, mainly including car-following behavior parameters and lane-changing decision parameters. Among them, the car-following behavior parameters include time reaction threshold, maximum acceleration, maximum deceleration, minimum headway, driver imperfection factor, and expected speed factor; the lane-changing decision parameters include strategic lane changing, cooperative lane changing, and speed gain lane-changing weight.

[0033] Specifically, structural similarity refers to the numerical indicator of how close the current scene and the historical scene are in terms of traffic operation quantitative indicators; semantic similarity refers to the numerical indicator of how well the current scene and the historical scene match in terms of semantic dimension.

[0034] In this embodiment, structured scene features and scene semantic vectors are used as inputs. All historical scenes stored in the preset simulation vector library are compared, and similarity calculations are performed along two independent dimensions: the structured similarity of the numerical dimension is calculated based on the structured scene features, and the semantic similarity of the attribute dimension is calculated based on the scene semantic vectors.

[0035] Step S104: Based on the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library, determine at least one similar historical scene and the multi-vehicle simulation parameter set corresponding to the similar historical scene.

[0036] Specifically, similar historical scenarios refer to historical scenarios selected from a pre-defined simulation vector library that have high structural and semantic similarity to the current scenario.

[0037] In this embodiment, based on the structural and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library, at least one more matching historical scene is selected from the preset simulation vector library as a similar historical scene, and the multi-vehicle simulation parameter set bound to each similar historical scene is retrieved.

[0038] Step S105: Verify multiple similar historical scenarios and the simulation parameter sets of multiple vehicle models corresponding to similar historical scenarios.

[0039] In this embodiment, three types of verifications are performed on multiple similar historical scenarios and the multi-vehicle simulation parameter sets corresponding to similar historical scenarios: data quality, matching threshold, and discrimination interval. Similar historical scenarios and the multi-vehicle simulation parameter sets corresponding to similar historical scenarios that pass all verifications can proceed to the subsequent parameter generation process.

[0040] Step S106: Select at least one similar historical scene that has passed the verification as the target historical scene.

[0041] Specifically, the target historical scenario refers to the historical scenario that has the highest compatibility with the current scenario after being screened through multi-dimensional confidence gating verification.

[0042] In this embodiment, based on the above verification results, similar historical scenarios that meet all preset gating conditions, i.e., pass the verification, are designated as target historical scenarios.

[0043] Step S107: Generate target simulation parameters based on the target historical scenario and the multi-vehicle simulation parameter set corresponding to the target historical scenario.

[0044] Specifically, the target simulation parameters refer to the set of configuration parameters of the traffic simulation model that are adapted to the traffic characteristics of the current highway scenario.

[0045] In this embodiment, the multi-vehicle simulation parameter set corresponding to the target historical scenario with the highest matching degree is selected as the target simulation parameter and output to the microscopic traffic simulation platform for use as configuration parameters.

[0046] In summary, in the technical solutions of some embodiments of this application, structured scene features and scene semantic vectors are generated based on multi-source trajectory observation data of the current scene. By calculating structured similarity and semantic similarity, matching similar historical scenes and their corresponding multi-vehicle model simulation parameter sets are selected from a preset simulation vector library. After verification, target simulation parameters adapted to the current scene are generated. Thus, in the scenario of highway emergency control strategy simulation, usable simulation parameters can be quickly output, shortening the parameter preparation cycle and ensuring the timeliness of control strategy evaluation. Simultaneously, through a two-dimensional matching and verification mechanism of structured numerical and semantic descriptions, the adaptability of simulation parameters to the current road network, traffic, and event scenario can be improved, enhancing the accuracy of control strategy evaluation results. This solves the problem of insufficient timeliness and accuracy in highway control strategy evaluation in the prior art.

[0047] In some embodiments, step S102, generating structured scene features and scene semantic vectors based on multi-source trajectory observation data, includes: Step S1021: Generate structured scene features based on multi-source trajectory observation data; wherein, the structured scene features include average speed, speed standard deviation, flow rate, density, average headway, lane change rate, proportion of large vehicles, observation time window, and congestion index.

[0048] Step S1022: Based on the multi-source trajectory observation data, determine the road network type, congestion status, vehicle type composition, and event type of the current scenario.

[0049] Step S1023: Determine the scene semantic vector based on the current road network type, congestion status, vehicle type composition, and event type.

[0050] Specifically, structured scene features include, but are not limited to, the following features: 1) Average speed: The average speed of all vehicles on the target road segment during the observation period. It is used to quantify the overall traffic flow level. The higher the value, the smoother the traffic flow.

[0051] 2) Speed ​​standard deviation: The degree of dispersion of vehicle speed during the observation period. It is used to reflect the dispersion and disturbance of traffic flow speed. The larger the value, the more violent the speed fluctuation and the more unstable the traffic flow operation.

[0052] 3) Traffic flow: The total number of vehicles passing through the target road section within a unit of time, used to quantify the traffic load intensity of the road section.

[0053] 4) Density: The number of vehicles distributed within a unit length of road segment, used to quantify the degree of road congestion.

[0054] 5) Average headway: The average time interval between two adjacent vehicles passing through the same section during the observation period, used to reflect the following driving status of the traffic flow. The smaller the value, the closer the traffic flow is.

[0055] 6) Lane change rate: The frequency of lane change behavior of vehicles on the target road segment per unit time, used to quantify the degree of lateral disturbance of traffic flow.

[0056] 7) Large vehicle ratio: The proportion of large vehicles to the total number of vehicles passing through during the observation period, used to characterize the vehicle type structure of the road segment.

[0057] 8) Observation time window: The time span of trajectory data used to extract scene features, which determines the time coverage of feature data.

[0058] 9) Congestion Index: A quantitative score calculated by combining multiple indicators such as vehicle speed, traffic flow, and density. It is used to intuitively represent the overall congestion level in the current scenario. The higher the value, the more severe the congestion.

[0059] Specifically, the characteristics of the structured scene are shown in Table 1.

[0060] Table 1. Structured Scene Features

[0061] Specifically, road network type refers to the type of road infrastructure to which the current scenario belongs, such as main road sections, ramp merging areas, and toll station sections; congestion status refers to the overall traffic operation level of the current scenario, such as smooth traffic, slow traffic, congestion, or severe congestion; vehicle type composition refers to the composition structure of different types of vehicles in the current scenario, such as the proportion of small passenger cars, large trucks, and buses; event type refers to the core event category that causes changes in traffic status, such as traffic accidents, road construction, and severe weather.

[0062] In this embodiment, firstly, nine traffic operation quantitative indicators are calculated based on multi-source trajectory observation data to generate multi-dimensional structured scene features covering average speed, speed standard deviation, flow rate, density, average headway, lane change rate, proportion of large vehicles, observation time window, and congestion index. Secondly, four types of semantic attributes—road network type, congestion status, vehicle type composition, and event type—are identified and extracted from the trajectory data to clarify the essential characteristics of the scene. Finally, the extracted semantic attributes are vectorized and encoded to generate scene semantic vectors that can be used for similarity calculation.

[0063] In the above embodiments, the generation logic of structured scene features and scene semantic vectors is clarified, making the scene feature extraction process more standardized and comprehensive, and providing a reliable feature basis for subsequent matching of historical scenes and corresponding simulation parameters.

[0064] In some embodiments, step S103, calculating the structural similarity and semantic similarity between the current scene and multiple historical scenes in a preset simulation vector library based on structured scene features and scene semantic vectors, includes: Step S1031: Based on the structured scene features, perform numerical similarity calculations on the current scene and multiple historical scenes in the preset simulation vector library to obtain the structured similarity.

[0065] Step S1032: Based on the scene semantic vector, perform semantic similarity calculation on the current scene and multiple historical scenes in the preset simulation vector library to obtain semantic similarity.

[0066] Optionally, let the feature vector of the query scenario be... The first in the preset simulation vector library The feature vectors of each historical scene are For the first The features are first standardized using z-score, as shown in expression (1). Then, the weighted Euclidean distance is calculated based on the standardized features, as shown in expression (2). The final structured similarity is defined as shown in expression (3).

[0067] (1) in, Indicates the current query scenario. The dimensionless values ​​of the dimensional features after z-score normalization. Indicates the current query scenario. The specific numerical values ​​of the dimensional structured features, This indicates that all historical samples in the preset simulation vector library are in the [number]th [period]. Mean value on dimensional structured features This indicates that all historical samples in the preset simulation vector library are in the [number]th [period]. Standard deviation of dimensional structural features Indicates the first The first historical scene The dimensionless values ​​of the dimensional features after z-score normalization. Indicates the first The historical scene in The specific numerical values ​​of the dimensional structured features.

[0068] (2) in, Indicates the current scene and the first The weighted Euclidean distance between historical scenes Indicates the first Preset weighting coefficients for dimensional structured features.

[0069] (3) in, Indicates the current scene and the first The structured similarity of historical scenes.

[0070] Optionally, let the semantic vector of the current scene be... The first in the preset simulation vector library The semantic vector of each historical scene is First, L2 normalization is performed on the two vectors to make the vector magnitudes both 1. Then, the semantic similarity is calculated using the cosine similarity algorithm. The calculation formula is shown in expression (4).

[0071] (4) in, Indicates semantic similarity. The semantic vector representing the current scene. Indicates the first in the preset simulation vector library Semantic vectors of historical scenes, This represents the L2 norm of a vector.

[0072] In this embodiment, firstly, based on the structured scene features of the current scene and each historical scene, the weighted Euclidean distance is calculated after z-score standardization, and the structured similarity corresponding to each historical scene is obtained, thus completing the similarity quantification in the numerical dimension; then, based on the scene semantic vectors of the current scene and each historical scene, the semantic similarity corresponding to each historical scene is calculated through cosine similarity after L2 normalization, thus completing the similarity quantification in the semantic dimension.

[0073] In the above embodiments, the calculation logic of structured similarity and semantic similarity are clarified. By independently calculating in two dimensions, the numerical features and semantic attributes of the scene are fully covered, making the similarity calculation process more standardized and providing strong support for improving the timeliness and accuracy of highway management strategy evaluation.

[0074] In some embodiments, step S104, which determines at least one similar historical scene and a set of multi-vehicle simulation parameters corresponding to the similar historical scene based on the structural and semantic similarity between the current scene and multiple historical scenes in a preset simulation vector library, includes: Step S1041: Weighted fusion of structured similarity and semantic similarity to obtain the matching score for each historical scene.

[0075] Step S1042: Based on the ranking of matching scores, at least one similar historical scenario is selected from the preset simulation vector library, and the multi-vehicle simulation parameter set corresponding to the similar historical scenario is obtained.

[0076] Specifically, the matching score is a comprehensive evaluation index that quantifies the overall fit between the current scene and the historical scene. It integrates matching information from two dimensions: numerical operation status and attribute semantic features. The formula for calculating the matching score is shown in expression (5).

[0077] (5) in, This represents the matching score for a single historical scene. The weighting coefficients representing structured similarity. This indicates the structural similarity between the current scene and the corresponding historical scene. Weight coefficients representing semantic similarity It represents the semantic similarity between the current scene and the corresponding historical scene.

[0078] In this embodiment, firstly, for each historical scene in the preset simulation vector library, the corresponding structured similarity and semantic similarity are weighted and summed according to the corresponding weight coefficients to calculate the comprehensive mixed matching score of each historical scene; then, all historical scenes are sorted in descending order of comprehensive matching score from high to low, and at least one candidate scene with the highest ranking is selected as a similar historical scene, and the multi-vehicle simulation parameter set corresponding to each similar historical scene is retrieved simultaneously.

[0079] In the above embodiments, the weighted fusion calculation formula and sorting and filtering logic of the two-dimensional similarity are clarified, making the screening process of similar scenarios more standardized and objective, and providing strong support for improving the timeliness and accuracy of highway control strategy evaluation.

[0080] In some embodiments, step S105, verifying multiple similar historical scenarios and the multi-vehicle simulation parameter sets corresponding to similar historical scenarios, includes: Step S1051: If the proportion of effective trajectory points, the missing rate of structured features, and the proportion of effective trajectory time windows meet the corresponding preset thresholds, determine whether the matching score of the top-ranked similar historical scene is not lower than the preset minimum similarity threshold.

[0081] Step S1052: If the matching score of the first-ranked similar historical scene is not lower than the preset minimum similarity threshold, determine whether the difference in matching scores between the first and second-ranked similar historical scenes is not lower than the minimum interval threshold.

[0082] Step S1053: If the difference in matching scores between the first and second most similar historical scenes is not lower than the minimum interval threshold, the first most similar historical scene is determined to have passed the verification.

[0083] Specifically, the effective trajectory percentage refers to the proportion of trajectory points in multi-source trajectory observation data where the velocity and acceleration do not exhibit abnormal jumps exceeding physical limits, out of the total number of trajectory points. The structured feature missing rate refers to the proportion of dimensions in the structured scene features that have missing values. The effective trajectory time window percentage refers to the ratio of the duration of a continuous effective time period after deduplication and breakpoint removal to the total trajectory duration.

[0084] Specifically, the minimum similarity threshold refers to the minimum matching score required for the top-ranked similar historical scene. It can be set to 0.35 to ensure that the top-ranked similar historical scene has a basic degree of matching with the current scene.

[0085] Specifically, the minimum interval threshold refers to the minimum difference in the mixed matching scores required between the top and second-ranked similar historical scenarios. It can be set to 0.02 to ensure that the matching advantage of the top-ranked similar historical scenario is sufficiently significant.

[0086] In this embodiment, firstly, the basic quality of the trajectory data of the current scene is verified to confirm that the proportion of effective trajectory points, the missing rate of structured features, and the proportion of effective trajectory time windows all meet the corresponding preset thresholds. On the basis of the trajectory feature quality meeting the standards, the matching score of the top-ranked similar historical scene is further judged to see if it is not lower than the preset minimum similarity threshold, thus confirming that the top-ranked similar historical scene has a basic matching degree. If the top-ranked similar historical scene meets the minimum similarity requirement, the difference in matching scores between the top and second-ranked similar historical scenes is calculated, and it is judged whether the difference is not lower than the minimum interval threshold, thus confirming that the top-ranked similar historical scene has sufficient distinguishing advantage. When all three judgment conditions are met, the top-ranked similar historical scene is finally determined to have passed the verification.

[0087] In the above embodiments, the confidence gating verification logic based on data quality, basic matching degree and candidate discrimination degree is clarified, the judgment criteria for similar scenarios and corresponding parameters are standardized, and the output simulation parameters have reliable scenario adaptability.

[0088] In some embodiments, step S107, generating target simulation parameters based on the target historical scene and the multi-vehicle simulation parameter set corresponding to the target historical scene, includes: Step S1071: If the verification passes, perform the first fast simulation verification on the multi-vehicle simulation parameter set corresponding to the target historical scenario.

[0089] Step S1072: If the first rapid simulation verification passes, the multi-vehicle simulation parameter set corresponding to the target historical scenario is determined as the target simulation parameter.

[0090] Specifically, the first rapid simulation verification, also known as short-window simulation rapid verification, is a lightweight simulation effect verification method. For example, by running the SUMO microscopic traffic simulation platform for 180 seconds, the traffic operation characteristics output by the simulation are compared with the real observation characteristics of the current scene, which can quickly verify the scene adaptation effect of the parameters.

[0091] In this embodiment, firstly, provided that the target historical scene passes the confidence gating verification, the multi-vehicle simulation parameter set corresponding to the target historical scene is imported into the simulation platform, the first fast simulation verification is started, the traffic operation characteristics output by the simulation operation are obtained and compared with the real observation characteristics of the current scene; if it is determined that the normalized feature error does not exceed the preset threshold, that is, the first fast simulation verification is passed, and the multi-vehicle simulation parameter set is directly determined as the target simulation parameters adapted to the current scene.

[0092] In the above embodiments, the rapid simulation verification rules and the confirmation logic of the target simulation parameters after the confidence gating is passed are clarified, further controlling the scenario adaptation accuracy of the parameters, making the output process of the target simulation parameters more reliable and standardized, and providing support for improving the timeliness and accuracy of highway management strategy evaluation.

[0093] In some embodiments, the method of this application further includes: Step a1: If the matching score of the top-ranked similar historical scene is lower than the preset minimum similarity threshold or the difference in matching scores between the top and second-ranked similar historical scenes is lower than the minimum interval threshold, a fast search is performed within the preset boundary using the multi-vehicle simulation parameter set corresponding to the top-ranked similar historical scene as the baseline to obtain the search results; wherein, the search results include at least one multi-vehicle simulation parameter set to be verified.

[0094] Step a2: Perform a second fast simulation verification on at least one set of simulation parameters for multiple vehicle models to be verified.

[0095] Step a3: If any set of simulation parameters for multiple vehicle models to be verified passes the second fast simulation verification, the set of simulation parameters for multiple vehicle models to be verified is determined as the target simulation parameters.

[0096] Optionally, bounded random fast search is the local parameter optimization method adopted in this embodiment. Taking the multi-vehicle simulation parameter set corresponding to the first-ranked similar historical scene as the baseline center, random sampling and mutation are performed within the preset value boundaries of each parameter dimension. Up to 12 rounds of short-window simulation fast deduction are performed to output the simulation parameters with the best fitting effect.

[0097] Specifically, the preset boundary refers to the range of values ​​for parameter search. Based on the values ​​of each dimension of the baseline parameters, upper and lower limits are set according to a fluctuation ratio of ±10% to ensure that the parameters generated by the search always meet the legal value requirements of the traffic engineering physical constraints and simulation platform.

[0098] Specifically, the second rapid simulation verification adopts a short-window simulation verification mechanism, which runs the micro-traffic simulation platform for 180 seconds as the simulation duration. The traffic operation characteristics output by the simulation are compared with the real observation characteristics of the current scene. The scene adaptability of the parameters is determined by normalizing the feature error. If the error does not exceed the preset threshold, the verification is deemed to have passed.

[0099] In this embodiment, firstly, when the matching score of the top-ranked similar historical scene is lower than the preset minimum similarity threshold, or the difference in matching scores between the top and second-ranked scenes is lower than the minimum interval threshold, but the trajectory feature quality meets the standard, a bounded random fast search is performed within the preset parameter value boundaries using the multi-vehicle simulation parameter set corresponding to the top-ranked similar historical scene as the baseline, generating a search result containing at least one set of multi-vehicle simulation parameter sets to be verified; secondly, a second fast simulation verification is performed sequentially on each set of multi-vehicle simulation parameter sets to be verified obtained from the search; finally, if the normalized feature error of any set of multi-vehicle simulation parameter sets to be verified passes the second fast simulation verification, then that set of parameters is determined as the target simulation parameters adapted to the current scene.

[0100] Optionally, if the feature quality is insufficient or no matching results are found in the fast search, the simulation parameters are not matched with the preset simulation vector library. Instead, the simulation parameters are directly set to the preset baseline parameter set and a fast simulation verification is started. At the same time, the reason for the downgrade is recorded for subsequent manual review.

[0101] In the above embodiments, the logic for rapid parameter search and verification in scenarios where confidence gating is not fully passed is clarified, which effectively improves the parameter adaptation effect in scenarios with insufficient matching degree, and takes into account the timeliness and accuracy of highway management strategy evaluation.

[0102] In some optional embodiments, after the emergency response is completed, real trajectory data and corresponding simulation output data within the event period are collected. First, spatial alignment between the trajectory and the road network is achieved through spatial index construction, initial screening of candidate road segments, and geometric projection matching. Then, average speed and average traffic flow are aggregated and calculated using road segments and minutes as joint dimensions to obtain a spatiotemporally aligned road segment-minute level comparison dataset. Next, based on the road segment-minute level spatiotemporally aligned comparison data, five loss components are constructed, namely, point error, time consistency, spatial morphology, hotspot segment weighting, and coverage constraint. These are then weighted and fused to form a multi-dimensional constraint tensor objective function. Iterative optimization is then carried out using the TPE (Tree-Structured Parzen Estimator) Bayesian optimization framework. In each round, candidate parameter sampling, simulation running, objective function calculation, and surrogate model updating are performed sequentially to gradually narrow the search range of the optimal parameters. Finally, after the iteration, the simulation parameters that minimize the value of the tensor objective function are output as the optimal result for offline calibration.

[0103] Specifically, the formula for calculating the tensor objective function is shown in expression (6).

[0104] (6) in, Represents the tensor objective function. Indicates the positional error. Indicates time consistency loss, Indicates the loss of spatial form. Indicates the weighted loss of hotspot segments. Indicates coverage penalty loss. , , , , The weights are respectively for location error, time consistency loss, spatial morphology loss, hotspot segment weighted loss, and coverage penalty loss.

[0105] Specifically, the location error is the deviation between the simulated and observed values ​​of velocity and flow rate for each effective spatiotemporal unit. The formula for calculating the location error is shown in expression (7).

[0106] (7)

[0107]

[0108] in, Indicates the positional error. Indicates the speed loss value. Indicates the value of traffic loss. This represents the Huber loss function with piecewise thresholds, subscript... This represents the segmented threshold parameter. This represents the simulated average velocity. This represents the actual observed average velocity. This indicates the smoothing term in the velocity calculation. Indicates the relative error of speed. This represents the actual observed average flow rate. This represents the simulated average flow rate. This indicates the smoothing term for flow calculation.

[0109] Specifically, the time consistency loss is a consistency penalty imposed on the speed change of adjacent minute indices on the same road segment in order to constrain the evolution of simulation results in the time dimension. The calculation formula of the time consistency loss is shown in expression (8).

[0110] (8) in, Indicates time consistency loss, This indicates the total number of valid road segments involved in the loss calculation. Indicates the first The target road section Indicates the first Total effective simulation minutes for each road segment Indicates minute index, Indicates the first The road section in The average speed of the simulation over minutes, Indicates the first The road section in The average speed of the simulation over minutes, Indicates the first The road section in The actual average speed observed over minutes Indicates the first The road section in The actual average speed observed over minutes.

[0111] Specifically, spatial morphology loss is used to characterize the spatial propagation pattern of congestion. Within the same minute, the speed sequences of all road segments are z-score standardized. After eliminating the absolute numerical differences, the morphological differences of the standardized sequences are compared. The formula for calculating spatial morphology loss is shown in expression (9).

[0112] (9) in, Indicates the loss of spatial form. This represents the total number of valid minutes involved in the spatial morphology calculation. Represents a minute-based time index. This indicates the total number of simulated road segments. Indicates the first The target road section Indicates the first The road section in The average velocity over minutes, after z-score normalization. Indicates the first The road section in The average velocity over minutes is the z-score normalized value.

[0113] Specifically, the hot spot weighted loss refers to the fitting priority used to strengthen the control of road sections. Based on the actual observation data, the road sections with the highest average flow or the lowest average speed are divided into hot spot subsets H, and the location error of the subset is given a higher weight. The calculation formula of the hot spot weighted loss is shown in expression (10).

[0114] (10)

[0115] in, Indicates the weighted loss of hotspot segments. This represents the total number of valid simulation minutes. Represents a minute-based time index. This indicates the total number of simulated road segments. Indicates the first The target road section Indicates the first The road section in The weighting coefficient corresponding to each minute Indicates the first The road section in The point error corresponding to each minute.

[0116] Specifically, the coverage penalty loss is used to prevent overfitting during the optimization process due to the scarcity of effective comparison data. The coverage is defined as the proportion of effective comparison units to the total theoretical units. ,when Below the preset target coverage threshold Additional penalties are applied at times, and the calculation formula for the coverage penalty loss is shown in expression (11).

[0117] (11) in, Indicates coverage penalty loss. This represents the coverage penalty coefficient. This represents the preset target coverage threshold. This indicates the effective comparison unit coverage.

[0118] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0119] Corresponding to the simulation parameter calibration method, this application also provides a simulation parameter calibration device. (See also...) Figure 2 This is a schematic diagram of a simulation parameter calibration device provided in some embodiments of this application. Figure 2 The simulation parameter calibration device includes: The acquisition module 201 is used to acquire multi-source trajectory observation data of the current scene; wherein, the multi-source trajectory observation data includes real-time vehicle trajectory data and historical vehicle trajectory data.

[0120] The generation module 202 is used to generate structured scene features and scene semantic vectors based on multi-source trajectory observation data.

[0121] The calculation module 203 is used to calculate the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library based on the structured scene features and scene semantic vectors; wherein each historical scene corresponds to a multi-vehicle simulation parameter set.

[0122] The determination module 204 is used to determine at least one similar historical scene and a set of multi-vehicle simulation parameters corresponding to the similar historical scene based on the structured similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library.

[0123] The verification module 205 is used to verify multiple similar historical scenarios and the simulation parameter sets of multiple vehicle models corresponding to similar historical scenarios.

[0124] The scene determination module 206 is used to select at least one similar historical scene that has passed the verification as the target historical scene.

[0125] The parameter generation module 207 is used to generate target simulation parameters based on the target historical scene and the multi-vehicle simulation parameter set corresponding to the target historical scene.

[0126] In some embodiments, the generation module 202 is specifically used for: Based on multi-source trajectory observation data, structured scene features are generated; among them, the structured scene features include average speed, speed standard deviation, flow rate, density, average headway, lane change rate, proportion of large vehicles, observation time window, and congestion index.

[0127] Based on multi-source trajectory observation data, the road network type, congestion status, vehicle type composition, and event type of the current scenario are determined.

[0128] The semantic vector of the scene is determined based on the road network type, congestion status, vehicle type composition, and event type of the current scene.

[0129] In some embodiments, the calculation module 203 is specifically used for: Based on the structured scene features, the numerical similarity between the current scene and multiple historical scenes in the preset simulation vector library is calculated to obtain the structured similarity.

[0130] Based on the scene semantic vector, the semantic similarity is calculated between the current scene and multiple historical scenes in the preset simulation vector library.

[0131] In some embodiments, the determining module 204 is specifically used for: We perform weighted fusion of structural similarity and semantic similarity to obtain the matching score for each historical scene.

[0132] Based on the ranking of matching scores, at least one similar historical scenario is selected from the preset simulation vector library, and the simulation parameter set of multiple vehicle models corresponding to the similar historical scenario is obtained.

[0133] In some embodiments, the verification module 205 is specifically used for: If the proportion of effective trajectory points, the missing rate of structured features, and the proportion of effective trajectory time windows meet the corresponding preset thresholds, determine whether the matching score of the top-ranked similar historical scene is not lower than the preset minimum similarity threshold.

[0134] If the matching score of the top-ranked similar historical scene is not lower than the preset minimum similarity threshold, determine whether the difference in matching scores between the top and second-ranked similar historical scenes is not lower than the minimum interval threshold.

[0135] If the difference in matching scores between the first and second most similar historical scenarios is not lower than the minimum interval threshold, the first most similar historical scenario is determined to pass the verification.

[0136] In some embodiments, the parameter generation module 207 is specifically used for: If the verification passes, perform the first fast simulation verification on the multi-vehicle simulation parameter set corresponding to the target historical scenario.

[0137] If the first rapid simulation verification passes, the multi-vehicle simulation parameter set corresponding to the target historical scenario is determined as the target simulation parameter.

[0138] In some embodiments, the parameter generation module 207 is further configured to: If the matching score of the top-ranked similar historical scene is lower than the preset minimum similarity threshold, or the difference in matching scores between the top and second-ranked similar historical scenes is lower than the minimum interval threshold, a fast search is performed within the preset boundary using the multi-vehicle simulation parameter set corresponding to the top-ranked similar historical scene as the baseline to obtain the search results; wherein, the search results include at least one multi-vehicle simulation parameter set to be verified.

[0139] Perform a second fast simulation verification on at least one set of simulation parameters for multiple vehicle models to be verified.

[0140] If any set of simulation parameters for multiple vehicle models to be verified passes the second fast simulation verification, the set of simulation parameters for multiple vehicle models to be verified is determined as the target simulation parameters.

[0141] For a description of the features in the embodiment corresponding to the simulation parameter calibration device, please refer to the relevant description in the embodiment corresponding to the sample data processing method, which will not be repeated here.

[0142] See also Figure 3 The embodiments of this application also provide an electronic device, including a memory 10 and a processor 20, wherein the memory 10 stores a computer program and the processor 20 is configured to run the computer program to perform the steps in any of the above-described simulation parameter calibration method embodiments.

[0143] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described simulation parameter calibration method embodiments at runtime.

[0144] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0145] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described simulation parameter calibration method embodiments.

[0146] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above-described simulation parameter calibration method embodiments.

[0147] Any of the components, modules, units, parts, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuitry), manual processing, or any combination thereof. Alternatively or additionally, any functionality described herein can be executed at least in part by one or more hardware logic components, such as, but not limited to, a central processing unit (CPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an application-specific standard product (ASSP), a system-on-a-chip (SoC), a complex programmable logic device (CPLD), a microprocessor (MCU), etc. The terms "system," "computing device," or "apparatus" as used herein encompass various means, devices, and machines for processing data, including, for example, one or more programmable processors, computers, SoCs, or combinations thereof. The apparatus may also include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or one or more combinations thereof. The aforementioned computer program (also known as a program, software, software application, app, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for a computing environment.

[0148] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0149] The simulation parameter calibration method, apparatus, device, and storage medium provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to help understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A simulation parameter calibration method, characterized in that, The method includes: Acquire multi-source trajectory observation data for the current scene; wherein, the multi-source trajectory observation data includes real-time vehicle trajectory data and historical vehicle trajectory data; Based on the multi-source trajectory observation data, structured scene features and scene semantic vectors are generated; Based on the structured scene features and scene semantic vectors, the structured similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library are calculated; wherein, each historical scene corresponds to a multi-vehicle simulation parameter set; Based on the structural and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library, at least one similar historical scene and a multi-vehicle simulation parameter set corresponding to the similar historical scene are determined. Verify the multiple similar historical scenarios and the multi-vehicle simulation parameter sets corresponding to the similar historical scenarios; Use at least one similar historical scene that passes the verification as the target historical scene; Target simulation parameters are generated based on the target historical scenario and the multi-vehicle simulation parameter set corresponding to the target historical scenario.

2. The simulation parameter calibration method according to claim 1, characterized in that, The step of generating structured scene features and scene semantic vectors based on the multi-source trajectory observation data includes: Based on the multi-source trajectory observation data, structured scene features are generated; wherein, the structured scene features include average speed, speed standard deviation, flow rate, density, average headway, lane change rate, proportion of large vehicles, observation time window, and congestion index; Based on the multi-source trajectory observation data, the road network type, congestion status, vehicle type composition, and event type of the current scenario are determined. The scene semantic vector is determined based on the road network type, congestion status, vehicle type composition, and event type of the current scene.

3. The simulation parameter calibration method according to claim 1, characterized in that, The step of calculating the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library based on the structured scene features and scene semantic vectors includes: Based on the structured scene features, the numerical similarity between the current scene and multiple historical scenes in the preset simulation vector library is calculated to obtain the structured similarity. Based on the scene semantic vector, the semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library is calculated to obtain the semantic similarity.

4. The simulation parameter calibration method according to claim 1, characterized in that, The step of determining at least one similar historical scene and a multi-vehicle simulation parameter set corresponding to the similar historical scene based on the structural and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library includes: The structured similarity and semantic similarity are weighted and fused to obtain the matching score for each historical scene; Based on the ranking of the matching scores, at least one similar historical scenario is selected from the preset simulation vector library, and the multi-vehicle simulation parameter set corresponding to the similar historical scenario is obtained.

5. The simulation parameter calibration method according to claim 4, characterized in that, The verification of the multiple similar historical scenarios and the multi-vehicle simulation parameter sets corresponding to the similar historical scenarios includes: If the proportion of effective trajectory points, the missing rate of structured features, and the proportion of effective trajectory time windows meet the corresponding preset thresholds, determine whether the matching score of the top-ranked similar historical scene is not lower than the preset minimum similarity threshold. If the matching score of the first-ranked similar historical scene is not lower than the preset minimum similarity threshold, determine whether the difference in matching scores between the first and second-ranked similar historical scenes is not lower than the minimum interval threshold. If the difference in matching scores between the first and second most similar historical scenarios is not lower than the minimum interval threshold, the first most similar historical scenario is determined to have passed the verification.

6. The simulation parameter calibration method according to claim 5, characterized in that, The step of generating target simulation parameters based on the target historical scenario and the multi-vehicle simulation parameter set corresponding to the target historical scenario includes: If the verification passes, a first fast simulation verification is performed on the multi-vehicle simulation parameter set corresponding to the target historical scenario; If the first rapid simulation verification passes, the multi-vehicle simulation parameter set corresponding to the target historical scenario is determined as the target simulation parameter.

7. The simulation parameter calibration method according to claim 6, characterized in that, The method further includes: If the matching score of the top-ranked similar historical scene is lower than the preset minimum similarity threshold or the difference in matching scores between the top and second-ranked similar historical scenes is lower than the minimum interval threshold, a fast search is performed within a preset boundary using the multi-vehicle simulation parameter set corresponding to the top-ranked similar historical scene as the baseline to obtain the search results; wherein, the search results include at least one multi-vehicle simulation parameter set to be verified. Perform a second fast simulation verification on the at least one set of simulation parameters for multiple vehicle models to be verified; If any of the aforementioned multi-vehicle model simulation parameter sets to be verified passes the second fast simulation verification, the aforementioned multi-vehicle model simulation parameter set to be verified is determined as the target simulation parameter.

8. A simulation parameter calibration device, characterized in that, The device includes: The acquisition module is used to acquire multi-source trajectory observation data of the current scene; wherein, the multi-source trajectory observation data includes real-time vehicle trajectory data and historical vehicle trajectory data; The generation module is used to generate structured scene features and scene semantic vectors based on the multi-source trajectory observation data; The calculation module is used to calculate the structured similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library based on the structured scene features and scene semantic vectors; wherein each historical scene corresponds to a multi-vehicle simulation parameter set; The determination module is used to determine at least one similar historical scene and a multi-vehicle simulation parameter set corresponding to the similar historical scene based on the structural similarity and semantic similarity between the current scene and multiple historical scenes in the preset simulation vector library; The verification module is used to verify the multiple similar historical scenarios and the multi-vehicle simulation parameter sets corresponding to the similar historical scenarios; The scene determination module is used to select at least one similar historical scene that has passed the verification as the target historical scene. The parameter generation module is used to generate target simulation parameters based on the target historical scenario and the multi-vehicle simulation parameter set corresponding to the target historical scenario.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the simulation parameter calibration method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the simulation parameter calibration method as described in any one of claims 1 to 7.