A phase migration method across a road network based on a large model and intersection driving trajectories

Abstract

This invention discloses a phase migration method across road networks based on a large model and intersection driving trajectories, relating to the field of intelligent traffic signal control. It addresses the technical problems of inconvenient processing of heterogeneous phase data and the difficulty in achieving automated phase migration across road networks due to differences in road segments. The invention includes: constructing a basic driving direction system table for intersections based on a large model; matching intersection phases with the basic driving direction system table to construct a phase system table; obtaining basic data for the source and target road networks to be matched; generating valid driving paths for both the source and target road networks and saving them to a valid driving path database; matching basic driving directions for the target road network; and using these basic driving directions to match valid driving paths for both the source and target road networks, thus achieving a unique binding of source road network phases to target road network segments. This invention completely overcomes the dual barriers of heterogeneous phase data and differences in road segment naming, reducing the deployment and optimization costs of urban traffic management systems.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic signal control, specifically to a phase cross-road network migration method based on a large model and intersection driving trajectories. Background Technology

[0002] In urban intelligent traffic signal control systems, the timing scheme of intersection signal phases is a core element that determines the efficiency of traffic flow. The reuse of mature and optimized phase timing schemes has become a key path to reduce the deployment cost of new road network control and improve traffic control efficiency.

[0003] However, the current industry standard is that phase information is stored separately for each individual intersection. The phase data of each intersection is stored independently in the local control system, without a unified storage standard and sharing mechanism, which results in the phase data of different intersections naturally having heterogeneous attributes.

[0004] Furthermore, there is a lack of unified standards for road segment naming rules and segment division among road networks from different data sources: some road networks divide road segments by road name + directional identifier (such as "East lane of XX Road"), while others are named by lane function + number (such as "Left Turn Lane 1"). Moreover, there are significant differences in the granularity of road segment splitting (such as whether to divide U-turn lanes independently or to merge similar turning lanes).

[0005] The inconsistency in road segment naming and division, coupled with the heterogeneity of phase data stored separately at each intersection, means that the correspondence between phases and road segments lacks a unified reference point. A phase timing scheme validated in one road network cannot be directly migrated to road networks with different data sources. It requires manual comparison of intersection structures, phase logic, and road segment attributes to adapt the phase scheme. This manual matching process is not only time-consuming and labor-intensive but also prone to errors due to human judgment, severely limiting the efficiency of cross-road network reuse of phase timing schemes and increasing the optimization cost of traffic control systems.

[0006] To address the aforementioned issues, existing technologies lack efficient means for integrating heterogeneous phase data stored independently at individual intersections, and also fail to resolve the matching barriers caused by differences in road segment naming and division, making it difficult to achieve automated migration of phase schemes across road networks. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention proposes a phase cross-network migration method based on a large model and intersection driving trajectory, which solves the technical problem that the inconvenience of processing heterogeneous phase data and the difficulty in achieving automated cross-network migration of phase schemes due to differences in road segments.

[0008] A phase-based cross-road network migration method based on a large model and intersection driving trajectories includes: Step 1: Construct a basic driving direction system table for the intersection; Step 2: Use the large model to match the intersection phases with the basic driving direction system table to construct the phase system table, and then obtain the basic data of the source road network and the target road network to be matched; Step 3: Generate valid driving paths for the source road network and the target road network respectively and save them to the valid driving path library. The valid driving path includes several trajectory segments that correspond one-to-one with the basic driving direction. Step 4: Based on the phase system table, match the basic driving direction for the target road network, and use the basic driving direction to match the effective driving paths of the source road network and the target road network, so as to complete the unique binding of the source road network phase to the target road network segment, thereby migrating the intersection phase timing table of the source road network to the target road network and obtaining the phase timing scheme of the target road network.

[0009] Furthermore, the basic driving direction in step 1 includes the main road positive direction and the diagonal direction. The main road positive direction is the direction of travel at the intersection entrance, and the diagonal direction is the direction of travel between two adjacent main road positive directions. The diagonal direction is defined based on the azimuth angle interval relative to the main road, with the intersection main road positive direction as the reference.

[0010] Furthermore, step 2, which involves matching the intersection phases with the basic driving direction system table to construct the phase system table, includes: By using a large model to learn and summarize the tables of signal timing schemes and phase configuration documents at urban intersections, we can extract all the phases that have appeared in the actual phase timing schemes and build a phase system table covering all scenarios and all types of intersections. Meanwhile, in the process of constructing the phase system table, based on the constructed basic driving direction system, semantic parsing and logical decomposition are performed on each phase in the phase system table, and one or more corresponding basic driving directions are matched for it simultaneously.

[0011] Furthermore, the basic data of the source road network mentioned in step 2 includes: the intersection phase timing table and the intersection center point coordinates of the source road network. The intersection phase timing table includes: the name of each phase of the intersection, the timing scheme parameters, and the corresponding basic driving direction rules. The corresponding basic driving direction rules refer to the fact that each phase can be decomposed into a combination of several basic driving directions according to the constructed basic driving direction system. The basic data of the target road network includes: target road network vector data and the coordinates of the intersection center points of the target road network. The target road network vector data includes road segment coordinates and road segment topological connection relationships.

[0012] Further, step 3 includes: Step 3.1: Using the center point coordinates of the intersection as the center, randomly generate initial OD pairs covering all driving directions of the intersection within a preset spatial range, ensuring that the driving trajectory generated based on the OD crosses the central area of ​​the intersection and covers all possible driving directions of the intersection; Step 3.2: For each generated OD pair, based on the real road network topology data of the target intersection, accurately map the randomly generated OD pair to the real road network to generate the shortest driving path from the starting point to the end point, and verify the shortest driving path to obtain the valid driving path. Step 3.3: For the valid driving path, the traversed road segments are divided into several continuous trajectory segments. Each trajectory segment contains a unique identifier, a spatial coordinate sequence, start / end coordinates, azimuth angle, and standardized driving direction. The standardized driving direction is a basic driving direction determined based on the azimuth angle of the first segment of the valid driving path and the azimuth angle changes during the driving process, thereby achieving matching between the basic driving direction and the actual road segment.

[0013] Furthermore, the splitting process is a dynamic calculation based on the actual physical angle, including: calculating the absolute azimuth of the trajectory segment relative to the north baseline using trigonometric functions based on the spatial coordinates of the starting and ending points of the trajectory segment; calculating the annular angle difference between the absolute azimuth of each trajectory segment and the four absolute reference anchor points; extracting the actual azimuth with the smallest angle difference from each anchor point to invert the global rigid rotation deflection angle of the intersection; finally determining the actual reference azimuth of the four main roads under the true topology of the intersection; and finally calculating the relative angle difference between the absolute azimuth of the current trajectory segment and the rigid rotation angle. If the minimum relative angle difference is greater than the tolerance threshold, it indicates that the trajectory crosses and lies between the reference azimuths of two adjacent actual main roads. At this time, it is determined to be a true diagonal, and the dynamic sector it falls into is strictly mapped to the corresponding diagonal direction in the basic driving direction system table defined in step 1.

[0014] Furthermore, the verification of the shortest driving path includes: verifying the shortest driving path based on invalid verification rules. If any invalid verification rule is met, the shortest driving path is determined to be an invalid sequence and is removed. After verification, a valid driving path is obtained. The invalid verification rules include: 1) there is no effective coverage threshold of the center point of any trajectory segment entering the intersection in the driving path; 2) the path length exceeds a preset threshold; 3) the effective distance of the OD point is less than the effective distance threshold of the OD point.

[0015] Furthermore, the effective coverage threshold of the intersection center point, the effective distance threshold of the OD point, and the length threshold of the path are determined based on the actual physical spatial scale of urban standard intersections and in combination with engineering processing experience of vehicle navigation trajectory (TMC). The effective coverage threshold of the intersection center point is used to determine whether the extracted trajectory actually crosses the physical control core area of ​​the intersection. The path length threshold is used to eliminate invalid trajectory segments caused by abnormal vehicle detours; The effective distance threshold of the OD point is used to filter out cases where the starting point and ending point are too close to the core area.

[0016] Further, step 4 includes: matching possible basic driving directions for each phase at each intersection in the target road network based on the phase system table; determining the corresponding effective driving paths for the target road network and the source road network for the matched basic driving directions based on the effective driving path library, and performing segment matching between the effective driving paths of the target road network and the source road network; finally, applying the phase timing scheme of the source road network to the effective driving path of the target road network, completing the unique binding of the source road network phase to the target road network segment, and clarifying the actual physical segment range of the phase control.

[0017] The beneficial effects of this invention include: 1. It has achieved fully automated processing from basic data input to timing scheme implementation, completely replacing the traditional manual work mode. Single intersection matching can be completed in minutes, and it can support batch parallel processing of thousands of intersections at the city level, improving matching efficiency by more than 100 times. At the same time, it completely avoids the error of human annotation, and the matching results are highly consistent.

[0018] 2. By pre-setting a standardized basic driving direction system covering all scenarios, the heterogeneous interference caused by personalized phase naming and non-standard labeling is eliminated. The mapping of any custom phase to the standardized basic driving direction is completed through a large model, without relying on a unified phase storage specification and sharing mechanism. At the same time, the objective spatial direction characteristics of the intersection driving path are used as the matching benchmark, replacing the traditional subjective road segment naming and division rules. This completely avoids the matching obstacles caused by differences in road segment naming rules, division granularity, and identification systems between different road networks. It can achieve cross-road network compatibility of phase schemes without relying on a unified road segment coding and naming system. The matching generalization ability far exceeds that of traditional technical solutions.

[0019] 3. By using a large model to perform deep learning and summarizing on massive amounts of actual signal timing schemes and phase configuration documents at urban intersections, a complete set of phases covering all scenarios and all types of intersections was constructed. During the construction of the complete set, a pre-set standardized basic driving direction was used as a unified benchmark to simultaneously complete the semantic parsing and logical decomposition of all phases. Each phase was matched with one or more corresponding basic driving directions, forming a directly reusable phase-basic driving direction mapping library. It is compatible with any custom phase naming and any phase combination logic, achieving full coverage of all types of phase scenarios in urban traffic, and its generalization ability is significantly better than existing technologies.

[0020] 4. Based on a hierarchical and precise matching logic from phase, standardized basic driving direction, effective driving path to actual controlled road segment, it can accurately distinguish the phase corresponding to the driving direction of all types of intersections such as straight, left turn, right turn, U-turn, and diagonal, completely solving the core pain points of phase confusion and matching errors caused by traditional single-segment direction matching; at the same time, through mechanisms such as all-directional OD generation, invalid path filtering, and same-direction path deduplication, it ensures full coverage of all preset phases at the intersection, with no matching gaps or omissions.

[0021] 5. Through a multi-round validity verification mechanism, it can effectively adapt to various urban road intersection scenarios such as crossroads, T-shaped intersections, Y-shaped intersections, and irregularly shaped complex channelized intersections. The matching results have strict spatial logic and mathematical interpretability, avoiding irregular matching deviations. Attached Figure Description

[0022] Figure 1 This is a flowchart of a phase cross-road network migration method based on a large model and intersection driving trajectory, which is an embodiment of this application.

[0023] Figure 2 This is a schematic diagram illustrating the dynamic calculation based on the actual physical angle involved in the embodiments of this application.

[0024] Figure 3 This corresponds to the basic driving direction involved in the embodiments of this application and the actual effective driving path at a standard intersection of urban main roads.

[0025] Figure 4 This corresponds to the basic driving direction involved in the embodiments of this application and the actual effective driving path at asymmetrical channelized intersections in the urban core area.

[0026] Figure 5 This refers to the correspondence between the basic driving direction involved in the embodiments of this application and the actual effective driving path at a multi-branched complex intersection. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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 a part of the embodiments of this application, and not all of the embodiments. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. Example 1

[0028] The following is in conjunction with the appendix Figure 1 Specific embodiments of the present invention will be described in detail; A phase-based cross-road network migration method based on a large model and intersection driving trajectories includes the following steps: Step 1: Construct a basic driving direction system table for the intersection; A basic driving direction system table for vehicles at intersections is constructed, as shown in Table 1, such as straight travel from south to north, straight travel from north to south, etc. The main road's positive direction is used as the reference, and the diagonal directions are defined based on the azimuth angle interval relative to the main road. The main road's positive direction is the direction of travel at the intersection's approach lanes, and the diagonal directions are the directions of travel between two adjacent main road positive directions.

[0029] Table 1 Basic Driving Direction System

[0030] Step 2: Use the large model to match the intersection phases with the basic driving direction system table to construct the phase system table, and then obtain the basic data of the source road network and the target road network for preprocessing; Step 2.1: Use the large model to learn and summarize the massive number of urban intersection signal timing scheme tables and phase configuration documents, extract all phases that have appeared in the actual phase timing schemes, and construct a phase system table covering all scenarios and all types of intersections.

[0031] The large model uses the pre-trained DeepSeek-V3 basic version of the basic large language model. Through prompt word templates, it extracts all custom phase names from the input timing document and decomposes and maps them to one or more standard directions in the basic driving direction system table based on their context logic.

[0032] Simultaneously, during the construction of the phase system table, based on the constructed basic driving direction system, semantic parsing and logical decomposition are performed on each phase in the phase system table, and one or more corresponding basic driving directions are matched for it. For example, the phase "north-south straight" is matched with the basic driving directions "south to north straight" and "north to south straight".

[0033] Step 2.2: Obtain basic data and perform standardization processing; Two types of data are obtained for the intersections to be matched. The first type is the intersection phase timing table and the intersection center point coordinates of the source road network. The intersection phase timing table includes: the name of each phase of the intersection, the timing scheme parameters, and the corresponding basic driving direction rules. The corresponding basic driving direction rules refer to the combination of several basic driving directions that each phase can be decomposed according to the constructed basic driving direction system. The second type is the target road network vector data and the intersection center point coordinates of the target road network. The target road network vector data includes road segment coordinates and road segment topological connection relationships, etc.

[0034] The standardization process includes: data format conversion, that is, converting the data of the phase timing table into a structured data dictionary that the system can read; and converting the phase names of the intersection into basic driving direction combinations based on the basic driving direction system.

[0035] In this embodiment, the source road network is a map source provided by the city traffic management bureau, which is a road network with existing timing schemes, and the target road network is obtained from the Gaode Map platform.

[0036] Step 3: Construct effective driving paths at intersections: Generate effective driving paths for the source road network and the target road network respectively and save them to the effective driving path library. The effective driving path includes several trajectory segments that correspond one-to-one with the basic driving direction. Step 3.1: Using the center point coordinates of the intersection as the center, randomly generate initial OD pairs covering all driving directions of the intersection within a preset spatial range, ensuring that the driving trajectory generated based on the OD crosses the central area of ​​the intersection and covers all possible driving directions of the intersection; The preset spatial range requires that the start and end points of the generated OD pairs fall within the edge area of ​​the preset spatial range, and must be significantly larger than the effective coverage threshold of the intersection center point. This ensures that each generated driving trajectory not only approaches the intersection but also completely traverses the core control area of ​​the intersection, realistically simulating the entry and exit actions of vehicles.

[0037] The specific preset and generation method is as follows: Using the center point of the target intersection as the geometric coordinate origin, and based on the conventional physical scale of urban intersections, a spatial radius is preset to ensure complete coverage of the intersection without including interfering road segments. When generating OD pairs, a radial projection is performed to the surrounding eight basic spatial directions, using the radius combined with a random perturbation coefficient, to generate the starting point (O) and ending point (D). This avoids OD points being generated inside the intersection, causing the trajectory to not cross the core area, while ensuring that the generated OD pairs uniformly cover all potential entrance and exit combinations of the target intersection. Specifically, in this embodiment, the preset spatial range is a radius of 100 meters to 500 meters.

[0038] Step 3.2: For each generated OD pair, based on the real road network topology data of the target intersection, accurately map the randomly generated OD pair onto the real road network to generate the shortest driving path from the starting point to the destination.

[0039] The shortest driving path is checked based on invalidity verification rules. If any invalidity verification rule is met, the shortest driving path is determined to be an invalid sequence and is removed. After the verification is completed, a valid driving path is obtained. The invalidity verification rules include: 1) There is no valid coverage threshold of the center point of any trajectory segment entering the intersection in the driving path; 2) The path length exceeds the preset threshold; 3) The effective distance of the OD point is less than the effective distance threshold of the OD point.

[0040] The above three thresholds are mainly determined based on the actual physical spatial scale of standard urban intersections and combined with engineering experience in vehicle navigation trajectory (TMC).

[0041] The effective coverage threshold of the intersection center point is used to define the core area of ​​the intersection, ensuring that the trajectory of the correct driving direction passes through the core area of ​​the intersection; the radius of the core conflict area (i.e., the intersection of lanes in all directions) of a typical urban arterial road intersection is usually within this range. This threshold is set to accurately determine whether the extracted trajectory has actually crossed the physical control core area of ​​the intersection, avoiding misclassifying a trajectory that merely passes nearby as a trajectory that crosses the intersection. In this embodiment, it is configured as 20 meters to 40 meters. Effective coverage threshold (e.g., radius 20-40 meters): The path length threshold is defined as follows: the length of a single continuous core trajectory segment corresponding to a vehicle completing a standard intersection turn or straight-line movement is finite. This upper limit is set to automatically eliminate excessively long invalid trajectory segments caused by abnormal vehicle detours. In this embodiment, it is configured to be 80 meters to 200 meters.

[0042] The effective distance threshold for OD points is used to constrain the generated start point (O) and end point (D) to be located a certain distance outside the intersection's core area. This is used to filter out cases where the start and end points are too close to the core area, preventing minor movements within the intersection from being mistaken for complete and valid entry and exit trajectories. In this embodiment, it is configured to be greater than 15 meters.

[0043] Step 3.3: For the valid driving path, the traversed road segments are divided into several continuous trajectory segments. Each trajectory segment contains a unique identifier, a spatial coordinate sequence, start / end coordinates, azimuth angle, and standardized driving direction. The standardized driving direction is a basic driving direction determined based on the azimuth angle of the first segment of the valid driving path and the azimuth angle changes during the driving process, thereby achieving matching between the basic driving direction and the actual road segment. The trajectory segment (i.e., TMC trajectory segment) is the smallest physical road segment, equivalent to a road segment unit. The road segments matched in the subsequent step 4 are composed of TMCs, equivalent to a TMC sequence.

[0044] The splitting process itself is based on the dynamic calculation of the actual physical angle, but the calculated result must be strictly compared and mapped (normalized) to the basic driving direction system table defined in step 1.

[0045] The dynamic calculation based on the actual physical angle includes: (1) Calculation of absolute azimuth of trajectory: based on the spatial coordinates of the starting point of trajectory segmentation Spatial coordinates of the endpoint The absolute azimuth of the trajectory segment relative to the north baseline is calculated using trigonometric functions.

[0046] In the formula, This represents the horizontal distance between the starting and ending spatial coordinates of a trajectory segment. This represents the perpendicular distance between the starting and ending spatial coordinates of a trajectory segment. Indicates the azimuth angle of the trajectory segments. These represent the longitude and latitude of the starting point spatial coordinates of the trajectory segment, respectively. These represent the longitude and latitude of the endpoint spatial coordinates of the trajectory segment, respectively.

[0047] (2) Dynamic learning of intersection reference azimuth: For the numerous non-orthogonal "sloping / abnormal" intersections in heterogeneous road networks, directly using absolute geographical angles to rigidly define trajectory directions will misjudge normal straight trajectories on sloping main roads as "diagonal" trajectories. To solve this problem, the actual intersections (such as...) Figure 2 The main driving directions (as indicated by the red line segment in the diagram) are modeled as standard orthogonal intersections with absolute reference anchor points of 0° North, 90° East, 180° South, and 270° West (e.g., Figure 2 (The black line segment in the diagram is shown) The topology structure formed after rigid rotation and fixed deflection angle; To this end, the absolute azimuth angles of all effective trajectory segments in the target intersection are extracted first, the annular angle difference between each trajectory azimuth angle and the four absolute reference anchor points is calculated, the actual azimuth angle with the smallest angle difference with each anchor point is extracted to invert the global rigid rotation deflection angle of the intersection, and finally the actual reference azimuth angles of the four main roads under the real topology of the intersection are determined. In essence, it is to match the real traffic flow direction that is closest to due north, due east, due south and due west of the intersection, and complete the overall topology calibration of the intersection, which serves as a unified reference for subsequent trajectory driving direction determination.

[0048] (3) Calculation of relative angles and interval mapping: Calculate the azimuth angle of the current trajectory segment. With rigid rotation angle The relative angle difference between them is given by the formula:

[0049] Positive direction mapping: If the extracted minimum If the value is less than the preset positive direction tolerance threshold, it means that the trajectory is traveling normally along the "sloping main road", and it is directly mapped to the corresponding positive direction of the main road; True oblique judgment: if the minimum If the deviation exceeds the tolerance threshold, it indicates that the trajectory crosses and lies between two adjacent actual main road reference azimuth angles. Only then does the system determine it as a "true diagonal" trajectory and strictly map the dynamic sector it falls into to the corresponding diagonal direction in the basic driving direction system table defined in step 1.

[0050] The specific correspondence between the basic driving direction and the actual effective driving path at the intersection is obtained through... Figures 3 to 5 The following three sets of typical intersection examples are used for illustration. These typical intersections include standard crossroads on urban arterial roads, asymmetrical channelized intersections in the urban core area, and multi-way complex intersections. Figure 3 Corresponding to standard crossroads of urban main roads, Figure 4 Corresponding to asymmetric channelized intersections in the core urban area, Figure 5 Corresponding to multi-branch complex intersections.

[0051] Taking the standardized basic driving directions of turning right from north to south and going straight from north to south, which are included in all three sets of examples, as an example, the colored highlighted lines in the figure are the core trajectory segments of the effective path for the corresponding driving direction, which fully present the entire driving process of the vehicle entering the intersection from the entrance lane, completing the corresponding turning / straight movement, and exiting from the exit lane. The gray lines are the remaining road network segments of the target intersection. All paths fully cover the core control area of ​​the intersection.

[0052] The three sets of examples correspond to target road networks with different channelization structures, different road segment division rules, and different intersection forms. Although the spatial form, the naming / division granularity of the associated actual road segments, and the intersection topology of the paths in the same direction are significantly different, they can all be uniformly determined as the basic driving direction defined in step 1 based on the azimuth angles at the beginning and end of the path and the clockwise angle changes during the driving process.

[0053] This verifies that the present invention uses the objective spatial direction characteristics of the driving trajectory at the intersection as the matching benchmark, which can completely avoid the matching barriers caused by differences in the naming and division of different road network segments and different intersection shapes, and provides a standardized reference for the unified cross-road network matching of phase-segment in the future.

[0054] All generated valid driving paths are deduplicated, and only one driving path is retained for each basic driving direction. All the retained valid driving paths constitute a valid driving path library.

[0055] Step 4: Match the phase timing scheme with the target road network segments: Based on the phase system table, possible basic driving directions are matched for each phase at each intersection in the target road network. Based on the effective driving path database, the corresponding effective driving paths of the target road network and the source road network are determined for the matched basic driving directions, and road segment matching is performed between the effective driving paths of the target road network and the source road network. Finally, the phase timing scheme of the source road network is applied to the effective driving path of the target road network, completing the unique binding of the source road network phase to the target road network segment, clarifying the actual physical road segment range of the phase control, and obtaining the phase timing scheme of the target road network. Specific examples are shown in Tables 2-4, corresponding to the timing schemes for standard cross intersections on urban arterial roads, asymmetric channelized intersections in the urban core area, and multi-way complex intersections, respectively.

[0056] Table 2 Timing Scheme for Standard Crossroads on Urban Arterial Roads

[0057] Table 3 Timing schemes for asymmetric channelized intersections in the urban core area

[0058] Table 4 Timing schemes for multi-way intersections

[0059] The embodiments described above merely illustrate specific implementation methods of this application, and while the descriptions are detailed and specific, they should not be construed as limiting the scope of protection of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the technical solution of this application, and these modifications and improvements all fall within the scope of protection of this application.

Claims

1. A phase-crossing road network migration method based on a large model and intersection driving trajectories, characterized in that, Includes the following steps: Step 1: Construct a basic driving direction system table for the intersection; Step 2: Use the large model to match the intersection phases with the basic driving direction system table to construct the phase system table, and then obtain the basic data of the source road network and the target road network to be matched; Step 3: Generate valid driving paths for the source road network and the target road network respectively and save them to the valid driving path library. The valid driving path includes several trajectory segments that correspond one-to-one with the basic driving direction. Step 4: Based on the phase system table, match the basic driving direction for the target road network, and use the basic driving direction to match the effective driving paths of the source road network and the target road network, so as to complete the unique binding of the phase of the source road network to the road segment of the target road network, thereby migrating the intersection phase timing table of the source road network to the target road network and obtaining the phase timing scheme of the target road network. Step 3 includes: Step 3.1: Using the center point coordinates of the intersection as the center, randomly generate initial OD pairs covering all driving directions of the intersection within a preset spatial range, ensuring that the driving trajectory generated based on the OD crosses the central area of ​​the intersection and covers all possible driving directions of the intersection; Step 3.2: For each generated OD pair, based on the real road network topology data of the target intersection, accurately map the randomly generated OD pair to the real road network to generate the shortest driving path from the starting point to the end point, and verify the shortest driving path to obtain the valid driving path. Step 3.3: For the valid driving path, the road segments are divided into several continuous trajectory segments. Each trajectory segment contains a unique identifier, spatial coordinate sequence, start / end coordinates, azimuth angle, and standardized driving direction. The standardized driving direction is a basic driving direction determined based on the azimuth angle of the first and second road segments of the valid driving path and the changes in azimuth angle during the driving process, thereby achieving the matching of the basic driving direction with the actual road segments. The segmentation process is a dynamic calculation based on actual physical angles, including: calculating the absolute azimuth of the trajectory segment relative to the north baseline using trigonometric functions based on the spatial coordinates of the starting and ending points of the trajectory segment; calculating the annular angle difference between the absolute azimuth of each trajectory segment and the four absolute reference anchor points; extracting the actual azimuth with the smallest angle difference from each anchor point to invert the global rigid rotation deflection angle of the intersection; finally determining the actual reference azimuth of the four main roads under the true topology of the intersection; and finally calculating the relative angle difference between the absolute azimuth of the current trajectory segment and the rigid rotation angle. If the minimum relative angle difference is less than the preset positive direction tolerance threshold, it is directly mapped to the corresponding positive direction of the main road. If the minimum relative angle difference is greater than the positive direction tolerance threshold, it indicates that the trajectory crosses and lies between two adjacent actual main road reference azimuths. At this time, it is determined to be a true diagonal, and the dynamic sector it falls into is strictly mapped to the corresponding diagonal direction in the basic driving direction system table defined in step 1.

2. The phase cross-road network migration method based on a large model and intersection driving trajectory according to claim 1, characterized in that, The basic driving direction in step 1 includes the main road positive direction and the diagonal direction. The main road positive direction is the direction of travel at the intersection entrance, and the diagonal direction is the direction of travel between two adjacent main road positive directions. The diagonal direction is defined based on the azimuth angle interval relative to the main road, with the intersection main road positive direction as the reference.

3. The phase cross-road network migration method based on a large model and intersection driving trajectory according to claim 1, characterized in that, Step 2, which involves matching the intersection phases with the basic driving direction system table to construct the phase system table, includes: By using a large model to learn and summarize the tables of signal timing schemes and phase configuration documents at urban intersections, we can extract all the phases that have appeared in the actual phase timing schemes and build a phase system table covering all scenarios and all types of intersections. Meanwhile, in the process of constructing the phase system table, based on the constructed basic driving direction system, semantic parsing and logical decomposition are performed on each phase in the phase system table, and one or more corresponding basic driving directions are matched for it simultaneously.

4. The phase cross-road network migration method based on a large model and intersection driving trajectory according to claim 1, characterized in that, The basic data of the source road network mentioned in step 2 includes: the intersection phase timing table and the intersection center point coordinates of the source road network. The intersection phase timing table includes: the name of each phase of the intersection, the timing scheme parameters, and the corresponding basic driving direction rules. The corresponding basic driving direction rules refer to decomposing each phase into a combination of several basic driving directions according to the constructed basic driving direction system. The basic data of the target road network includes: target road network vector data and the coordinates of the intersection center points of the target road network. The target road network vector data includes road segment coordinates and road segment topological connection relationships.

5. The phase cross-road network migration method based on a large model and intersection driving trajectory according to claim 1, characterized in that, The verification of the shortest driving path includes: checking the shortest driving path based on invalid verification rules. If any invalid verification rule is met, the shortest driving path is determined to be an invalid sequence and is removed. After the verification is completed, a valid driving path is obtained. The invalid verification rules include: 1) there is no effective coverage threshold of the center point of any trajectory segment entering the intersection in the driving path; 2) the path length exceeds a preset threshold; 3) the effective distance of the OD point is less than the effective distance threshold of the OD point.

6. The phase cross-road network migration method based on a large model and intersection driving trajectory according to claim 5, characterized in that, The effective coverage threshold of the intersection center point, the effective distance threshold of the OD point, and the path length threshold are determined based on the actual physical spatial scale of urban standard intersections and in combination with engineering processing experience of vehicle navigation trajectories. The effective coverage threshold of the intersection center point is used to determine whether the extracted trajectory actually crosses the physical control core area of ​​the intersection. The path length threshold is used to eliminate invalid trajectory segments caused by abnormal vehicle detours; The effective distance threshold of the OD point is used to filter out cases where the starting point and ending point are too close to the core area.

7. The phase cross-road network migration method based on a large model and intersection driving trajectory according to claim 1, characterized in that, Step 4 includes: based on the phase system table, matching possible basic driving directions for each phase at each intersection in the target road network; based on the effective driving path library, determining the corresponding effective driving path of the target road network and the effective driving path of the source road network for the matched basic driving directions, and performing segment matching between the effective driving paths of the target road network and the effective driving paths of the source road network; finally, applying the phase timing scheme of the source road network to the effective driving path of the target road network, completing the unique binding of the source road network phase to the target road network segment, and clarifying the actual physical segment range of the phase control.

Images