SUMO road network coordinate conversion method and device based on adaptive learning algorithm, electronic equipment and storage medium

Abstract

The invention provides an SUMO road network coordinate conversion method and device based on an adaptive learning algorithm, electronic equipment and a storage medium, and belongs to the technical field of data processing, and the method comprises the steps: extracting node information, edge information and lane information from road network data; generating a point geometric object based on the node information; generating a line geometric object of the road edge based on the edge information; generating a surface geometric object based on the lane information; converting the two-dimensional coordinates of the geometric object into WGS84 space coordinates by using an integrated learning model fused with various heterogeneous machine learning algorithms; if the WGS84 space coordinates obtained through conversion are abnormal, interpolation restoration is carried out, restoration is carried out based on preset geometric constraints, or restoration is carried out based on network topology connectivity; and updating road network data based on the WGS84 space coordinates. The non-linear deformation of the complex road network can be captured, and the conversion precision is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a SUMO road network coordinate transformation method, apparatus, electronic device, and storage medium based on an adaptive learning algorithm. Background Technology

[0002] SUMO, as a microscopic traffic simulation software, uses a local Cartesian coordinate system to describe traffic elements (lanes, nodes, traffic lights, etc.) in its road network data. However, practical applications require WGS84 geographic coordinates to support scenarios such as GIS analysis and high-precision map matching.

[0003] Traditional coordinate transformation methods require pre-defined, fixed transformation parameters. For example, the seven-parameter transformation method employs a multi-step transformation process (WGS84 → Cartesian → 7 parameters → Clarke80), but requires pre-determined fixed transformation parameters. Another example is the projection parameter transformation method, which uses the Proj4js library for coordinate system transformation and relies on pre-defined projection parameters. This approach depends on fixed parameters, and the transformation accuracy is limited by the accuracy of the pre-defined parameters. It lacks an adaptive optimization mechanism, resulting in limited coordinate transformation accuracy.

[0004] Therefore, improving the accuracy of SUMO road network coordinate transformation based on adaptive learning algorithms has become an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides a SUMO road network coordinate transformation method, device, electronic device, and storage medium based on an adaptive learning algorithm, which solves the defect in the prior art where the accuracy of SUMO road network coordinate transformation based on adaptive learning algorithm is limited by the accuracy of preset parameters, resulting in insufficient coordinate transformation accuracy.

[0006] This invention provides a SUMO road network coordinate transformation method based on an adaptive learning algorithm, comprising the following steps: Node information, edge information, and lane information are extracted from road network data; the node information is the attribute information of intersections, the edge information is the attribute information of road edges, and the lane information is the attribute information of lanes. Based on the node information, a point geometry object is generated; Based on the edge information, a line geometry object of the road edge is generated; Based on the line geometry object, generate the surface geometry object of the lane surface; Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometry, the line geometry of the road edge, and the surface geometry are converted into WGS84 spatial coordinates; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation is performed based on the WGS84 spatial coordinates of neighboring points, or based on preset geometric constraints, or based on network topology connectivity. The road network data is updated based on the WGS84 spatial coordinates.

[0007] According to the present invention, a SUMO road network coordinate transformation method based on an adaptive learning algorithm is provided, wherein the ensemble learning model includes a base layer, a core layer, and a fusion layer; The base layer expands the two-dimensional coordinates to a cubic polynomial feature space, resulting in a nine-dimensional feature vector. The core layer includes a parallel support vector machine regression model, a random forest regression model, and a gradient boosting regression model. The support vector machine regression model, the random forest regression model, and the gradient boosting regression model generate prediction results based on the nine-dimensional feature vector, respectively. The fusion layer includes a ridge regressor learner that generates WGS84 spatial coordinates based on the three prediction results.

[0008] According to the SUMO road network coordinate transformation method based on an adaptive learning algorithm provided by the present invention, before performing interpolation repair based on the WGS84 spatial coordinates of neighboring points, or repair based on preset geometric constraints, or repair based on network topology connectivity if the transformed WGS84 spatial coordinates are abnormal, the method further includes: The continuity of the WGS84 spatial coordinates is detected in real time. If the WGS84 spatial coordinates change abruptly, it is determined that the WGS84 spatial coordinates are abnormal. The geographic rationality of the WGS84 spatial coordinates is detected in real time. If the WGS84 spatial coordinates are not within the preset geographic range, the WGS84 spatial coordinates are determined to be abnormal. The transformation accuracy of the WGS84 spatial coordinates is detected in real time. If the transformation accuracy is less than a preset accuracy threshold, the WGS84 spatial coordinates are determined to be abnormal.

[0009] According to the present invention, a SUMO road network coordinate transformation method based on an adaptive learning algorithm is provided, wherein the SUMO road network coordinate transformation method based on the adaptive learning algorithm further includes: The updated road network data will be output as Shapefile format data.

[0010] According to the present invention, a SUMO road network coordinate transformation method based on an adaptive learning algorithm is provided, wherein the SUMO road network coordinate transformation method based on the adaptive learning algorithm further includes: The vehicle trajectory data is parsed to identify the vehicle ID and trajectory point data; Based on the vehicle ID and the trajectory point data, construct the driving trajectory of each vehicle; Generate lane surface data based on Shapefile format data of the lane centerline; Vehicle trajectory data is generated based on the driving trajectory and the lane surface data.

[0011] According to the SUMO road network coordinate transformation method based on an adaptive learning algorithm provided by the present invention, the generation of lane surface data from Shapefile format data based on lane centerlines includes: Generate lane regions based on Shapefile format data of lane centerlines; Project the WGS84 spatial coordinates of the lane centerline onto the UTM 50N coordinate system to obtain the UTM 50N spatial coordinates; Based on the lane width information, a buffering operation is performed on the UTM 50N spatial coordinates to generate a lane surface; After removing overlaps and repairing gaps in the lane surface, the repaired data is projected into the WGS84 space to obtain lane surface data.

[0012] The present invention also provides a SUMO road network coordinate transformation device based on an adaptive learning algorithm, comprising the following modules: The information extraction module is used to extract node information, edge information, and lane information from road network data; the node information is the attribute information of intersections, the edge information is the attribute information of road edges, and the lane information is the attribute information of lanes. The point generation module is used to: generate point geometric objects based on the node information; The edge generation module is used to: generate line geometry objects of the road edge based on the edge information; The central generation module is used to: generate a surface geometry object of the lane surface based on the line geometry object; The coordinate transformation module is used to: convert the two-dimensional coordinates of the point geometric object, the line geometric object of the road edge, and the surface geometric object into WGS84 spatial coordinates using a pre-trained ensemble learning model; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates; The anomaly repair module is used to: if the converted WGS84 spatial coordinates are abnormal, perform interpolation repair based on the WGS84 spatial coordinates of nearby points, or repair based on preset geometric constraints, or repair based on network topology connectivity. The road network update module is used to update the road network data based on the WGS84 spatial coordinates.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the SUMO road network coordinate transformation method based on the adaptive learning algorithm as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the SUMO road network coordinate transformation method based on the adaptive learning algorithm as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the SUMO road network coordinate transformation method based on the adaptive learning algorithm as described above.

[0016] This invention provides a SUMO road network coordinate transformation method, device, electronic device, and storage medium based on an adaptive learning algorithm. It extracts node information, edge information, and lane information from road network data. The node information refers to the attribute information of intersections, the edge information refers to the attribute information of road edges, and the lane information refers to the attribute information of lanes. Based on the node information, point geometric objects are generated; based on the edge information, line geometric objects of road edges are generated; based on the line geometric objects, surface geometric objects of lane surfaces are generated. Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometric objects, the line geometric objects of the road edges, and the surface geometric objects are converted into WGS84 spatial coordinates. The ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation repair is performed based on the WGS84 spatial coordinates of neighboring points, or repair is performed based on preset geometric constraints, or repair is performed based on network topology connectivity. Based on the WGS84 spatial coordinates, the road network data is updated. Traditional methods rely on preset fixed transformation parameters or standard projection models, making it difficult to handle issues such as local distortion and nonlinear deformation in SUMO road networks, resulting in large transformation errors. This invention, by introducing an ensemble learning model, can automatically learn local spatial transformation rules based on actual road network data, significantly improving the accuracy of coordinate transformation. This is particularly evident in scenarios with complex urban structures and elevated highway ramps, where it achieves significantly higher transformation accuracy than traditional methods. Traditional methods require manual adjustment of transformation parameters in different cities or terrain regions, lacking adaptability. In contrast, the ensemble learning model used in this invention possesses excellent generalization capabilities, supports joint training with samples from multiple cities and terrains, and can dynamically update model parameters through an online learning mechanism, significantly improving the system's adaptability and transferability. Furthermore, this invention reconstructs point geometric objects, line geometric objects at road edges, and surface geometric objects in a hierarchical manner. Different geometric objects exhibit different behavioral characteristics during spatial deformation. For example, intersection nodes are more suitable for precise point-to-point mapping, while lane centerlines are more suitable for linear fitting and smoothing. Through hierarchical modeling, a more suitable transformation strategy can be designed for each type of geometric object, avoiding the "one-size-fits-all" error caused by a globally uniform model and significantly improving the overall transformation accuracy. Moreover, the SUMO road network has a complex topological structure (such as connection relationships, lane affiliation, traffic light control, etc.). This topological information still needs to be maintained after geographic coordinate transformation. By modeling point, line, and surface objects separately and performing consistency verification, the topological relationships such as road connectivity, lane continuity, and intersection structure can be effectively maintained, avoiding problems such as "road breaks" and "misalignments". Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts of the SUMO road network coordinate transformation method based on adaptive learning algorithm provided by the present invention; Figure 2 This is the second flowchart of the SUMO road network coordinate transformation method based on adaptive learning algorithm provided by the present invention; Figure 3 This is the third flowchart of the SUMO road network coordinate transformation method based on adaptive learning algorithm provided by the present invention; Figure 4 This is a schematic diagram of the SUMO road network coordinate transformation device based on adaptive learning algorithm provided by the present invention; Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention; Figure 6 This is a schematic diagram of the line geometry of the lane centerline provided by the present invention; Figure 7 This is a schematic diagram of the surface geometry of the lane surface provided by the present invention; Figure 8 This is a schematic diagram of the linear geometry of a lane intersection provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] It should be noted that in the description of the embodiments of the present invention, 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. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0022] The following is combined with Figures 1-8 This invention describes the SUMO road network coordinate transformation method, apparatus, electronic device, and storage medium based on an adaptive learning algorithm provided in embodiments of the present invention.

[0023] Figure 1 This is one of the flowcharts illustrating the SUMO road network coordinate transformation method based on adaptive learning algorithm provided by this invention, such as... Figure 1 As shown, the method includes the following: S110, extract node information, edge information and lane information from road network data; the node information is the attribute information of the intersection, the edge information is the attribute information of the road edge, and the lane information is the attribute information of the lane. S120, Based on the node information, generate a point geometry object; S130, Based on the edge information, generate a line geometry object of the road edge; S140, Based on the line geometry object, generate the surface geometry object of the lane surface; S150, using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometry object, the line geometry object of the road edge, and the surface geometry object are converted into WGS84 spatial coordinates; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates; S160. If the converted WGS84 spatial coordinates are abnormal, then interpolation is performed based on the WGS84 spatial coordinates of the neighboring points, or based on preset geometric constraints, or based on network topology connectivity. S170, Update the road network data based on the WGS84 spatial coordinates.

[0024] It should be noted that the execution subject of the SUMO road network coordinate transformation method based on adaptive learning algorithm provided in the embodiments of the present invention can be a server, computer device, such as a tablet computer, laptop computer, handheld computer, vehicle electronic device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc.

[0025] In S110, the road network data, i.e., the SUMO network file to be converted, is stored in XML format and contains a complex nested structure. Here, different types of road network elements are identified from the road network data. The parsing process first reads the location tag in the file header and extracts key information such as network offset, coordinate boundaries, and projection parameters. This information provides important references for subsequent coordinate transformation. Specifically, all junction tags are traversed, and internal type nodes used within the SUMO are automatically filtered out, retaining only the node data representing actual intersections. For each valid node, its ID, coordinates, node type, and other attribute information are extracted to obtain node information. Node types include priority, traffic light, and right-before-left, which are of great significance in GIS analysis. Edge labels are processed to identify road connection relationships. Each edge contains attributes such as the starting node ID, ending node ID, and road type to obtain edge information. Lane labels under the edge are parsed to obtain detailed lane-level information, including lane ID, speed limit, length, and precise geometric shape point sequence. The geometric shape is stored in string format and needs to be parsed into an array of coordinate points to obtain lane information.

[0026] In S120, the node layer creates point geometry objects. Each intersection corresponds to a Point geometry, retaining the original attribute information, such as node ID, coordinate position, and type.

[0027] In S130, the edge layer creates line geometry objects, generating LineString line geometry based on the coordinates of the start and end nodes of the edge.

[0028] Figure 6 This is a schematic diagram of the linear geometry of the lane centerline provided by the present invention, such as... Figure 6 As shown, each lane centerline corresponds to a line geometry object.

[0029] Figure 8 This is a schematic diagram of the linear geometry of a lane intersection provided by the present invention, such as... Figure 8 As shown, a lane intersection includes intersecting transverse and longitudinal line geometric objects.

[0030] In S140, such as Figure 7As shown, the lane layer creates refined line geometry objects. After generating high-precision lane centerlines based on the detailed geometric points of each lane, it is tiled and expanded into surface geometry objects for the lane surface. In essence, a hierarchical spatial data structure is constructed based on the parsed raw data. The execution order of S120, S130, and S140 is not strictly limited and can be determined according to actual usage requirements.

[0031] It should be noted that the original topological relationships are strictly maintained during the creation of geometric objects. Connections between edges and nodes are established through node IDs to ensure the connectivity of the road network. Simultaneously, the hierarchical relationship between edges and lanes is maintained to support multi-level spatial queries and analysis.

[0032] In S150, a trained ensemble learning model is applied to transform the coordinates of each geometric object. Point geometric objects directly transform their coordinates, while line geometric objects transform their shape point sequence point by point.

[0033] In S160, quality checks are performed in real time during the transformation process. For abnormal WGS84 spatial coordinates detected, interpolation repair is performed based on the transformation results of neighboring points; or, geometric constraints are used to maintain the shape characteristics of linear features (such as road curvature) for repair; or, topology checks are used to ensure the correctness of network connections.

[0034] In S170, the SUMO road network coordinates in the road network data are replaced with the corresponding WGS84 spatial coordinates to obtain the updated road network data.

[0035] The SUMO road network coordinate transformation method based on an adaptive learning algorithm provided in this invention extracts node information, edge information, and lane information from road network data. The node information refers to the attribute information of intersections, the edge information refers to the attribute information of road edges, and the lane information refers to the attribute information of lanes. Based on the node information, point geometric objects are generated; based on the edge information, line geometric objects of road edges are generated; based on the line geometric objects, surface geometric objects of lane surfaces are generated. Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometric objects, the line geometric objects of road edges, and the surface geometric objects are converted into WGS84 spatial coordinates. The ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation repair is performed based on the WGS84 spatial coordinates of neighboring points, or repair is performed based on preset geometric constraints, or repair is performed based on network topology connectivity. Based on the WGS84 spatial coordinates, the road network data is updated. This invention predicts and generates WGS84 spatial coordinates by integrating multiple heterogeneous machine learning algorithms into an ensemble learning model. This model can capture the nonlinear deformation of complex road networks, such as spiral ramps of interchanges, and achieve centimeter-level conversion accuracy of road networks. In addition, it performs hierarchical reconstruction of point geometric objects, line geometric objects of road edges, and surface geometric objects to ensure the integrity of the conversion data.

[0036] In an optional embodiment, the ensemble learning model includes a base layer, a core layer, and a fusion layer; The base layer expands the two-dimensional coordinates to a cubic polynomial feature space, resulting in a nine-dimensional feature vector. The core layer includes a parallel support vector machine regression model, a random forest regression model, and a gradient boosting regression model. The support vector machine regression model, the random forest regression model, and the gradient boosting regression model generate prediction results based on the nine-dimensional feature vector, respectively. The fusion layer includes a ridge regressor learner that generates WGS84 spatial coordinates based on the three prediction results.

[0037] In this embodiment of the invention, a three-layer integrated learning architecture is constructed. The base layer employs a polynomial feature expansion technique to extend the original two-dimensional coordinate features (x, y) to a cubic polynomial feature space, generating a nine-dimensional feature vector including x, y, x², y², xy, x³, y³, x²y, and xy². This nonlinear feature expansion can capture complex coordinate mapping relationships and is particularly suitable for handling nonlinear transformations in geographic coordinate transformations.

[0038] In this embodiment of the invention, the core layer integrates three heterogeneous machine learning algorithms. Support Vector Regression (SVR) uses a radial basis function kernel, which can handle high-dimensional nonlinear mappings; Random Forest Regression utilizes an ensemble voting mechanism of multiple decision trees, exhibiting good anti-overfitting ability; and Gradient Boosting Regression (XGBoost) achieves high-precision prediction through iterative optimization of residuals. Each algorithm has its advantages: SVR performs excellently with small sample sizes, Random Forest is insensitive to outliers, and XGBoost has extremely high prediction accuracy.

[0039] In this embodiment of the invention, the fusion layer employs a stacked ensemble strategy, using the prediction results of three base models as input features to train a ridge regression meta-learner. The meta-learner can learn the weight allocation of different base models and automatically select the model combination that performs best in different spatial regions.

[0040] Optionally, when training the ensemble learning model, training sample data is obtained in the following manner: The SUMO simulation coordinate system needs to be converted to the WGS84 standard coordinate system. First, a basic reference point dataset is established. Road center point information is located using GPS, and the SUMO local coordinates and corresponding WGS84 spatial coordinates of key geographical locations are obtained and packaged into CSV files. These reference points are distributed across different locations within the simulation area, including major intersections, road start and end points, and area boundary points. Specifically, the data includes: reference point 1 with local coordinates of (562.57, 1244.07) and corresponding WGS84 coordinates of (120.368655°E, 36.293704°N); reference point 2 with local coordinates of (19.14, 1690.72) and corresponding WGS84 coordinates of (120.362474°E, 36.297593°N), and so on. To improve model training performance, data augmentation techniques are employed to expand the original n (e.g., 10) reference points to m (e.g., 50) training samples. Data augmentation methods include: generating intermediate points between known reference points using spatial interpolation algorithms; calculating the coordinates of interpolation points using distance-weighted inverse interpolation; and ensuring the geographical validity of the interpolation points through cross-validation. The augmented dataset better covers the entire transformation area, improving the model's generalization ability.

[0041] Optionally, Bayesian optimization techniques are employed for intelligent hyperparameter tuning of the ensemble learning model. Traditional grid search or random search methods are inefficient, while Bayesian optimization, based on Gaussian process modeling, can find the globally optimal parameter combination within a finite number of trials. In specific implementation, a hyperparameter search space is defined, with the SVR regularization parameter C being 10. -3Logarithmically uniform distribution between 10³ and 10³, kernel function parameter gamma is within 10 -4 The number of trees in a random forest varies between 100 and 500 (integer distribution), and the maximum depth varies between 5 and 20. The learning rate of XGBoost varies between 0.01 and 0.3 (logarithmic distribution), the number of trees varies between 100 and 300, and the maximum depth varies between 3 and 15. The Bayesian optimizer constructs a probabilistic model of the objective function based on historical trial results, uses a collection function to balance exploration and utilization, and automatically selects the next set of most promising parameters for experimentation. After 100 optimization trials, the optimal parameter combination that minimizes the mean squared error of the 5-fold cross-validation is found.

[0042] Furthermore, to ensure the geographical reasonableness of the transformation results, a geographical constraint regularization mechanism is introduced: First, a reasonable geographical boundary is set to ensure that the transformed latitude and longitude coordinates fall within a reasonable range within my country (longitude 70°-140°, latitude 15°-55°). For transformation results exceeding this reasonable range, soft constraints are used for correction to avoid accuracy loss caused by hard truncation. Uncertainty quantification is achieved through the prediction variance of the integrated model: the standard deviation of the prediction results from the three base models is calculated as a measure of the reliability of the transformation results. Transformation results with high uncertainty are marked and manually reviewed to ensure data quality.

[0043] The SUMO road network coordinate transformation method based on an adaptive learning algorithm provided in this invention integrates a learning model with a base layer to combat nonlinearity, a core layer to ensure robustness, and a fusion layer to optimize accuracy. These three layers collaborate in a hierarchical manner. A meta-learner adaptively adjusts the weight distribution of each sub-model based on local geographic features of the spatial region (such as terrain undulation, road density, and topological complexity), thereby dynamically optimizing the predictive performance of the integrated model. Compared to existing fixed-parameter models (such as the seven-parameter transformation method) or standard projection transformation methods (such as transformations based on the Proj4 library), this invention exhibits stronger adaptability and higher transformation accuracy when handling complex nonlinear deformations and local spatial distortions. It is particularly suitable for applications with stringent requirements for spatial mapping accuracy and consistency, such as geographic coordinate transformation, high-precision map matching, autonomous driving simulation, and urban digital twins.

[0044] In an optional embodiment, before performing interpolation repair based on the WGS84 spatial coordinates of neighboring points, or repair based on preset geometric constraints, or repair based on network topology connectivity if the converted WGS84 spatial coordinates are abnormal, the method further includes: The continuity of the WGS84 spatial coordinates is detected in real time. If the WGS84 spatial coordinates change abruptly, it is determined that the WGS84 spatial coordinates are abnormal. The geographic rationality of the WGS84 spatial coordinates is detected in real time. If the WGS84 spatial coordinates are not within the preset geographic range, the WGS84 spatial coordinates are determined to be abnormal. The transformation accuracy of the WGS84 spatial coordinates is detected in real time. If the transformation accuracy is less than a preset accuracy threshold, the WGS84 spatial coordinates are determined to be abnormal.

[0045] In this embodiment of the invention, quality checks are performed in real time during the conversion process: The continuity of coordinate transformation is checked to avoid abrupt changes, meaning the distance between adjacent coordinate points should conform to geographical reality. Abrupt changes are corrected through interpolation, for example, using cubic spline interpolation to reconstruct coordinates based on five adjacent valid points; or, through geometric constraints, the distance between points is forced to not exceed a preset threshold, such as the road segment's design speed multiplied by the sampling time interval (e.g., 0.5s interval). The geographical rationality of the coordinates is checked to ensure that the coordinates fall within the expected geographical range. For example, the buffer zone is automatically expanded by a certain proportion based on the road network MBR (Minimum Bounding Rectangle) to obtain the expected geographical range. Calculate the conversion accuracy index and evaluate the conversion reliability of each point.

[0046] The SUMO road network coordinate transformation method based on adaptive learning algorithm provided in this embodiment of the invention detects the transformation results in real time during the transformation process, automatically identifies abnormal coordinates, and performs data repair to ensure the reliability of the transformation results.

[0047] In an optional embodiment, the SUMO road network coordinate transformation method based on the adaptive learning algorithm further includes: The updated road network data will be output as Shapefile format data.

[0048] In this embodiment of the invention, to address the issue of missing details in existing road network conversion methods when processing lane connection details at intersections, the road network data (XML data file) is converted to OpenDRIVE format (.xodr) with high fidelity. This method can completely preserve the static structural information of the road network and some dynamic traffic elements. The OpenDRIVE file uses... <planview>Elements and their interior <geometry>The sub-elements precisely define the microscopic geometry and curvature characteristics of the road. Simultaneously, this format supports standardized modeling of road ancillary facilities such as traffic lights, stop lines, and lane boundaries. Furthermore, by importing the converted .xodr file into TESSNG software and leveraging its built-in geometric reconstruction and structure recognition algorithms, the road network information is further output as richer Shapefile (.shp) format data. This file not only contains information such as lane width, lane turning, and stop lines, but also accurately describes the complex lane connections within intersections, effectively compensating for the information gaps in traditional methods when modeling complex intersection structures.

[0049] The SUMO road network coordinate transformation method based on adaptive learning algorithm provided in this invention converts the XML format road network data output by the SUMO simulation system into the OpenDRIVE standard format with high fidelity. It fully preserves the static structural information in the original road network (such as road topology, lane division, intersection layout, etc.) and effectively retains some dynamic traffic elements (such as traffic light control logic, lane speed limit information, etc.), realizing a semantic-level mapping from simulation modeling data to a standardized autonomous driving map format. On this basis, this invention further converts the OpenDRIVE format road network data into Shapefile Geographic Information System (GIS) format. Through geometric modeling and logical reasoning of lane connection relationships, it realizes automatic completion and topological correction of missing or ambiguous lane connection line data, significantly improving the spatial expression accuracy and logical consistency of road network data.

[0050] In an optional embodiment, the SUMO road network coordinate transformation method based on the adaptive learning algorithm further includes: The vehicle trajectory data is parsed to identify the vehicle ID and trajectory point data; Based on the vehicle ID and the trajectory point data, construct the driving trajectory of each vehicle; Generate lane surface data based on Shapefile format data of the lane centerline; Vehicle trajectory data is generated based on the driving trajectory and the lane surface data.

[0051] In practice, vehicle trajectory data is typically stored in CSV format. However, different versions of SUMO may have different output formats. Therefore, the format is inferred by reading the first few lines of the file, and then an appropriate parsing strategy is selected. Specifically, the delimiter type (comma, semicolon, tab, etc.) and column structure of the CSV file are identified. For files containing column headers, fields such as time, vehicle ID, coordinates, and lane are directly identified based on the header name. For files without headers, the meaning of each column is automatically inferred based on the data type and numerical range. Data type conversion and outlier detection are performed during the parsing process to ensure the data quality of subsequent processing.

[0052] Furthermore, based on the parsed trajectory point data, complete driving trajectories are constructed by grouping by vehicle ID. The trajectory points for each vehicle are sorted by timestamp, forming an ordered spatiotemporal sequence. Quality checks are performed on the trajectory data: consistency of time intervals is checked to identify abnormal time jumps; the continuity of spatial location is verified to avoid unreasonable instantaneous movements; driving speed and acceleration are calculated to identify abnormal motion states. A coordinate transformation model is applied to each trajectory point to convert SUMO local coordinates to latitude and longitude coordinates in WGS84 space. The transformed trajectory points are used to construct LineString line geometry objects, forming a complete vehicle driving trajectory line. Simultaneously, the temporal attributes of the trajectory are preserved, supporting spatiotemporal analysis and dynamic visualization.

[0053] In this embodiment of the invention, trajectory data in various formats is generated according to different analytical needs. Point format output contains detailed information for each trajectory point, suitable for refined behavioral analysis; line format output contains complete trajectory lines, suitable for path analysis and traffic statistics; aggregation format outputs statistical information aggregated by time window or spatial region, suitable for macro-level traffic analysis. All output data adopts a standard geographic information format and sets the correct coordinate reference system (EPSG:4326) to ensure compatibility with mainstream GIS software. The attribute table contains rich vehicle and trajectory information, supporting complex spatial queries and statistical analyses.

[0054] The SUMO road network coordinate transformation method based on adaptive learning algorithm provided in this invention automatically handles the output differences of multiple SUMO versions, eliminating the need for manual configuration; the spatial association between lane surfaces and trajectory lines supports refined behavior analysis; and point, line, and aggregate data constitute a complete spatiotemporal analysis system.

[0055] In an optional embodiment, generating lane surface data based on the Shapefile format data of the lane centerline includes: Generate lane regions based on Shapefile format data of lane centerlines; Project the WGS84 spatial coordinates of the lane centerline onto the UTM 50N coordinate system to obtain the UTM 50N spatial coordinates; Based on the lane width information, a buffering operation is performed on the UTM 50N spatial coordinates to generate a lane surface; After removing overlaps and repairing gaps in the lane surface, the repaired data is projected into the WGS84 space to obtain lane surface data.

[0056] In this embodiment of the invention, a planar lane region is generated based on lane centerline Shapefile data using a buffer construction method. The lane line data in the original WGS84 coordinate system is projected to the UTM 50N coordinate system, which is suitable for geometric operations. Based on the width information of each lane, the centerline is buffered to generate a lane surface with actual width. During the buffering process, flat-ended endpoints and acute-angle connections are used to more realistically reproduce the actual road shape. The generated buffered polygon is then projected back to the WGS84 coordinate system.

[0057] Furthermore, overlap removal and gap repair are performed on the lane surfaces to ensure the spatial continuity and topological correctness of the generated results. Through a traversal approach, a spatial difference operation is performed on each lane surface and the existing lane set to remove overlapping portions. To handle minute gaps, a "expansion-contraction" micro-buffering strategy is adopted. After projection onto the local UTM coordinate system, the lane surfaces are buffered at a small scale to effectively fill gaps caused by boundary errors. The repaired data is then projected back to WGS84, outputting continuous, complete, and geometrically valid lane surface data.

[0058] The SUMO road network coordinate transformation method based on an adaptive learning algorithm provided in this invention projects data onto the UTM 50N coordinate system for processing, making it suitable for geometric operations. It performs overlap removal and gap repair on lane surfaces to ensure the spatial continuity and topological correctness of the generated results. Compared to common direct transformation or simple geometric processing methods in the prior art, this invention effectively solves common problems such as spatial discontinuity, topological breaks, and geometric redundancy in SUMO road network data during cross-coordinate system transformation and geometric modeling by introducing a standardized projected coordinate system and overlap / gap optimization strategies, significantly improving the spatial consistency and geometric integrity of the road network data.

[0059] Figure 2 This is the second flowchart of the SUMO road network coordinate transformation method based on adaptive learning algorithm provided by this invention. Figure 3 This is the third flowchart of the SUMO road network coordinate transformation method based on adaptive learning algorithm provided by this invention, as shown below. Figure 2 and Figure 3 As shown, data verification and optimization are performed in real time during the conversion process. Specifically, a three-dimensional evaluation system is established, including statistical accuracy, spatial accuracy, and robustness assessment. Statistical accuracy is measured using traditional indicators such as the R² coefficient of determination, root mean square error (RMSE), and mean absolute error (MAE); spatial accuracy is based on calculating the spatial distribution characteristics of the conversion error based on geographical distance; robustness assessment is performed by testing the model's performance stability under different noise levels.

[0060] In this embodiment of the invention, a city traffic simulation project is used as the background. This project includes more than 1,000 traffic nodes, road edges, and nearly 10,000 lanes. It requires converting the SUMO simulation coordinate system to the WGS84 standard coordinate system for subsequent traffic platform visualization and analysis. The ensemble learning model achieves an R²=0.99996 goodness of fit for longitude transformation and an R²=0.99994 goodness of fit for latitude transformation, far exceeding the 0.9980 and 0.9975 of traditional linear regression models. The RMSE error is controlled within 0.12 meters, compared to the 0.8-meter error of traditional methods, representing an 85% improvement in accuracy. Under 10% Gaussian noise interference, the model accuracy retention rate reaches 96%, demonstrating excellent robustness.

[0061] A real-time quality monitoring system is established to continuously monitor conversion quality during the data transformation process. Monitoring indicators include: real-time statistics of conversion accuracy, frequency of outlier detection, violations of geographical constraints, processing speed, and resource consumption. When a quality anomaly is detected, the system automatically triggers a repair mechanism or issues an alert. Automatic repair strategies include: repairing isolated outliers based on nearest-neighbor interpolation; repairing outliers in the trajectory using time series smoothing techniques; repairing network connectivity through topological constraints; and repairing missing data based on historical statistical information. The entire repair process is recorded, generating detailed quality reports for user reference.

[0062] The following describes the SUMO road network coordinate transformation device based on the adaptive learning algorithm provided in the embodiments of the present invention. The SUMO road network coordinate transformation device based on the adaptive learning algorithm described below can be referred to in correspondence with the SUMO road network coordinate transformation method based on the adaptive learning algorithm described above.

[0063] Figure 4 This is a schematic diagram of the SUMO road network coordinate transformation device based on adaptive learning algorithm provided by the present invention, as shown below. Figure 4 As shown, the SUMO road network coordinate transformation device based on the adaptive learning algorithm may include, but is not limited to, the following: The information extraction module 410 is used to extract node information, edge information and lane information from road network data; the node information is the attribute information of the intersection, the edge information is the attribute information of the road edge, and the lane information is the attribute information of the lane. The point generation module 420 is used to: generate point geometric objects based on the node information; Edge generation module 430 is used to: generate line geometry objects of the road edge based on the edge information; The center generation module 440 is used to: generate a surface geometry object of the lane surface based on the line geometry object; The coordinate transformation module 450 is used to: convert the two-dimensional coordinates of the point geometric object, the line geometric object of the road edge, and the surface geometric object into WGS84 spatial coordinates using a pre-trained ensemble learning model; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates; The anomaly repair module 460 is used to: if the converted WGS84 spatial coordinates are abnormal, perform interpolation repair based on the WGS84 spatial coordinates of nearby points, or perform repair based on preset geometric constraints, or perform repair based on network topology connectivity. The road network update module 470 is used to update the road network data based on the WGS84 spatial coordinates.

[0064] It should be noted that the SUMO road network coordinate transformation device based on adaptive learning algorithm provided in this embodiment of the invention can execute the SUMO road network coordinate transformation method based on adaptive learning algorithm described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0065] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a SUMO road network coordinate transformation method based on an adaptive learning algorithm, which includes: Node information, edge information, and lane information are extracted from road network data; the node information is the attribute information of intersections, the edge information is the attribute information of road edges, and the lane information is the attribute information of lanes. Based on the node information, a point geometry object is generated; Based on the edge information, a line geometry object of the road edge is generated; Based on the line geometry object, generate the surface geometry object of the lane surface; Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometry, the line geometry of the road edge, and the surface geometry are converted into WGS84 spatial coordinates; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation is performed based on the WGS84 spatial coordinates of neighboring points, or based on preset geometric constraints, or based on network topology connectivity. The road network data is updated based on the WGS84 spatial coordinates.

[0066] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a 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 the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the SUMO road network coordinate transformation method based on the adaptive learning algorithm provided by the above methods, the method including: Node information, edge information, and lane information are extracted from road network data; the node information is the attribute information of intersections, the edge information is the attribute information of road edges, and the lane information is the attribute information of lanes. Based on the node information, a point geometry object is generated; Based on the edge information, a line geometry object of the road edge is generated; Based on the line geometry object, generate the surface geometry object of the lane surface; Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometry, the line geometry of the road edge, and the surface geometry are converted into WGS84 spatial coordinates; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation is performed based on the WGS84 spatial coordinates of neighboring points, or based on preset geometric constraints, or based on network topology connectivity. The road network data is updated based on the WGS84 spatial coordinates.

[0068] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the SUMO road network coordinate transformation method based on the adaptive learning algorithm provided by the above methods, the method comprising: Node information, edge information, and lane information are extracted from road network data; the node information is the attribute information of intersections, the edge information is the attribute information of road edges, and the lane information is the attribute information of lanes. Based on the node information, a point geometry object is generated; Based on the edge information, a line geometry object of the road edge is generated; Based on the line geometry object, generate the surface geometry object of the lane surface; Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometry, the line geometry of the road edge, and the surface geometry are converted into WGS84 spatial coordinates; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation is performed based on the WGS84 spatial coordinates of neighboring points, or based on preset geometric constraints, or based on network topology connectivity. The road network data is updated based on the WGS84 spatial coordinates.

[0069] The device embodiments described above are merely illustrative. 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 network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0070] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.< / geometry> < / planview>

Claims

1. A SUMO road network coordinate transformation method based on an adaptive learning algorithm, characterized in that, include: Extract node information, edge information, and lane information from road network data; The node information is the attribute information of the intersection, the edge information is the attribute information of the road edge, and the lane information is the attribute information of the lane. Based on the node information, a point geometry object is generated; Based on the edge information, a line geometry object of the road edge is generated; Based on the line geometry object, generate the surface geometry object of the lane surface; Using a pre-trained ensemble learning model, the two-dimensional coordinates of the point geometry, the line geometry of the road edge, and the surface geometry are converted into WGS84 spatial coordinates; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates. If the converted WGS84 spatial coordinates are abnormal, interpolation is performed based on the WGS84 spatial coordinates of neighboring points, or based on preset geometric constraints, or based on network topology connectivity. The road network data is updated based on the WGS84 spatial coordinates.

2. The SUMO road network coordinate transformation method based on adaptive learning algorithm according to claim 1, characterized in that, The ensemble learning model includes a base layer, a core layer, and a fusion layer; The base layer expands the two-dimensional coordinates to a cubic polynomial feature space, resulting in a nine-dimensional feature vector. The core layer includes a parallel support vector machine regression model, a random forest regression model, and a gradient boosting regression model. The support vector machine regression model, the random forest regression model, and the gradient boosting regression model generate prediction results based on the nine-dimensional feature vector, respectively. The fusion layer includes a ridge regressor learner that generates WGS84 spatial coordinates based on the three prediction results.

3. The SUMO road network coordinate transformation method based on adaptive learning algorithm according to claim 1, characterized in that, Before performing interpolation repair based on the WGS84 spatial coordinates of neighboring points, or repair based on preset geometric constraints, or repair based on network topology connectivity if the converted WGS84 spatial coordinates are abnormal, the method further includes: The continuity of the WGS84 spatial coordinates is detected in real time. If the WGS84 spatial coordinates change abruptly, it is determined that the WGS84 spatial coordinates are abnormal. The geographic rationality of the WGS84 spatial coordinates is detected in real time. If the WGS84 spatial coordinates are not within the preset geographic range, the WGS84 spatial coordinates are determined to be abnormal. The transformation accuracy of the WGS84 spatial coordinates is detected in real time. If the transformation accuracy is less than a preset accuracy threshold, the WGS84 spatial coordinates are determined to be abnormal.

4. The SUMO road network coordinate transformation method based on adaptive learning algorithm according to claim 1, characterized in that, The SUMO road network coordinate transformation method based on the adaptive learning algorithm also includes: The updated road network data will be output as Shapefile format data.

5. The SUMO road network coordinate transformation method based on adaptive learning algorithm according to claim 4, characterized in that, The SUMO road network coordinate transformation method based on the adaptive learning algorithm also includes: The vehicle trajectory data is parsed to identify the vehicle ID and trajectory point data; Based on the vehicle ID and the trajectory point data, construct the driving trajectory of each vehicle; Generate lane surface data based on Shapefile format data of the lane centerline; Vehicle trajectory data is generated based on the driving trajectory and the lane surface data.

6. The SUMO road network coordinate transformation method based on adaptive learning algorithm according to claim 5, characterized in that, The Shapefile format data based on the lane centerline is used to generate lane surface data, including: Generate lane regions based on Shapefile format data of lane centerlines; Project the WGS84 spatial coordinates of the lane centerline onto the UTM 50N coordinate system to obtain the UTM 50N spatial coordinates; Based on the lane width information, a buffering operation is performed on the UTM 50N spatial coordinates to generate a lane surface; After removing overlaps and repairing gaps in the lane surface, the repaired data is projected into the WGS84 space to obtain lane surface data.

7. A SUMO road network coordinate transformation device based on an adaptive learning algorithm, characterized in that, include: The information extraction module is used to extract node information, edge information, and lane information from road network data. The node information is the attribute information of the intersection, the edge information is the attribute information of the road edge, and the lane information is the attribute information of the lane. The point generation module is used to: generate point geometric objects based on the node information; The edge generation module is used to: generate line geometry objects of the road edge based on the edge information; The central generation module is used to: generate a surface geometry object of the lane surface based on the line geometry object; The coordinate transformation module is used to: convert the two-dimensional coordinates of the point geometric object, the line geometric object of the road edge, and the surface geometric object into WGS84 spatial coordinates using a pre-trained ensemble learning model; the ensemble learning model integrates the prediction results of multiple heterogeneous machine learning algorithms to generate WGS84 spatial coordinates; The anomaly repair module is used to: if the converted WGS84 spatial coordinates are abnormal, perform interpolation repair based on the WGS84 spatial coordinates of nearby points, or repair based on preset geometric constraints, or repair based on network topology connectivity. The road network update module is used to update the road network data based on the WGS84 spatial coordinates.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the SUMO road network coordinate transformation method based on the adaptive learning algorithm as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the SUMO road network coordinate transformation method based on the adaptive learning algorithm as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the SUMO road network coordinate transformation method based on the adaptive learning algorithm as described in any one of claims 1 to 6.

Images