A transportation-based fuzzy route intention resolution and path planning method and system
Patent Information
- Application Number
- CN202611233568.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-14
- Publication Date
- 2026-11-03
AI Technical Summary
Traditional oversized cargo transportation permit approval processes suffer from problems such as inertia in government services, semantic ambiguity of place names, disordered location of linear road segments, and disordered input order of waypoints, resulting in large errors in route planning and making it difficult to achieve automation and precision.
A fuzzy route intent parsing method based on AI Large Language Model (LLM) is adopted. By semantic parsing and directional reasoning of unstructured route text, a baseline route with origin, destination and highway toll station is generated. Anchor points are evenly set along the route and the POI coordinates of road segments are snapped up. Finally, the route is rearranged in an orderly manner to generate a complete route.
It achieves absolute accuracy in the transportation routes of large items, reduces the error rate of route planning, avoids deviations and detours, improves the automation of the approval process, and reduces the cost of manual verification.
Smart Images
Figure CN123028620A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of natural language parsing, intelligent route planning, and smart transportation technology, and in particular to a transportation-based fuzzy route intent parsing and route planning method and system. Background Technology
[0002] With economic and engineering development, the demand for transporting heavy equipment, wind turbine blades, large transformers, and other bulky items is increasing. Because bulky transport vehicles typically have excessive length, width, height, or weight, they place extremely high demands on road conditions. To ensure transportation safety and the integrity of along-route infrastructure, traffic management and approval departments need to rigorously verify the applied transport routes, especially confirming whether the vehicles can safely pass through key points such as highways, specific toll stations, and bridges.
[0003] Traditional approval processes and pathway planning suffer from the following pain points: (1) Constraints of inertia in government services: In the current approval process for oversized cargo transportation permits, applicants still rely heavily on manually filling in route descriptions in text form. Since the transportation routes are often highly repetitive, applicants tend to directly copy the route text from existing application forms to improve efficiency. This operational inertia makes it impossible for the government approval process to temporarily eliminate the text route input function, and it is also difficult to directly introduce a map route planning module to provide applicants with standardized structured route guidance, thus forming a bottleneck in the digitalization of the approval process.
[0004] (2) Semantic ambiguity of place names: users’ colloquial place names and abbreviations do not match the standard names of POIs on the map. Roads with the same name are distributed in many places across the country. Single point keyword search is very likely to match wrong points in different places. (3) Incorrect location of linear road segments: The linear road segments (such as G105, port access road, etc.) filled in by the applicant are only returned by the map POI search (which may be the coordinates of a single point such as the start, end, midpoint, service area, ramp toll station, etc. of the entire road segment). The points are often located on the highway auxiliary road, interchange ramp, and opposite lane. Directly using them as passing points will force vehicles to enter and exit ramps, which is very likely to result in invalid large-scale detours. (4) Disorder of waypoint input: When applicants for large-item transportation fill in the route, it is very easy to have disordered waypoint input. If it is used directly for route planning, there will be a large number of detours. (5) The entrance and exit of the toll station cannot be automatically distinguished: There are two independent POI points for the entrance and exit of the same toll station. Manual screening is inefficient and the wrong entrance or exit will cause large vehicles to exit the road illegally. Summary of the Invention
[0005] In view of the above problems, the present invention is proposed to provide a transportation-based fuzzy route intent parsing and path planning method and system to overcome or at least partially solve the above problems.
[0006] According to one aspect of the present invention, a transportation-based fuzzy route intent parsing and path planning method is provided, the path planning method comprising: Step S1: Semantic parsing and directional reasoning of unstructured route text; Step S2: Generate coordinates for the origin and destination, highway toll station, and baseline route L1; Step S3: Evenly set anchor points along the route to search for and snap to the coordinates of POIs of road points and road sections within a 10km radius; Step S4: Ordered coordinate rearrangement and generation of L2 complete path.
[0007] Optionally, step S1, semantic parsing and directional reasoning of the unstructured route text, includes: Receive unstructured natural language route descriptions input from the oversized cargo transportation declaration system, and construct system prompts that include entity recognition and physical intent constraints; Named entity recognition is performed by calling the AI Large Language Model (LLM). An ordered sequence of path entities is extracted; The AI large language model LLM infers the entrance / exit attributes of each toll station entity in the current route and marks them as "entrance" or "exit".
[0008] Optionally, the path entity sequence includes a start point, an end point, a highway number, a toll station, and a regular point of interest (POI).
[0009] Optionally, step S1, semantic parsing and directional reasoning of the unstructured route text, specifically includes: Step S1.1: Input the user's natural language route description, LLM splits the structured text, extracts the start point, end point, and waypoints, distinguishes the toll station entrance / exit types, and outputs JSON structured data {"start":"start point name","end":"end point name","via":["point 1","point 2",...]}; Step S1.2 Keyword preprocessing: The LLM adds entry / exit suffixes to text containing the "toll station" field to optimize the accuracy of subsequent POI retrieval; Step S1.3: Obtain the starting text S, the ending text E, and the original waypoint WayTxt=[K1,K2...Km].
[0010] Optionally, step S2, generating the origin and destination, the coordinate conversion of the highway toll station, and the baseline route L1 specifically includes: Step S2.1: Call the map's location retrieval API, and let LLM determine the optimal selection of candidate points. Convert the starting point S and ending point E into latitude and longitude coordinates respectively. O , Des ; Step S2.2: Filter all toll station names in WayTxt, retrieve toll station coordinates in advance, and use LLM to determine the optimal selection of candidate points returned by the map API; after selection, use them as temporary anchor points to plan the baseline route, ensuring that the baseline route L1 passes through the designated toll station; Step S2.3: Generate a baseline route L1 without common waypoints, parse the polyline trajectory string returned by the API, and split it to obtain a discrete trajectory point set L1={p1,p2,...,pn}. L1 serves as the geometric calculation baseline for the entire process.
[0011] Optionally, step S3, uniformly setting anchor points along the route to search for and absorb the coordinates of waypoints and POIs within a 10km range, specifically includes: Step S3.1: Using a 20-equal-division uniform anchor point strategy, uniformly sample N=20 anchor points Anchor={A1,A2...A20} from the baseline route L1 trajectory list. For a single route keyword, call the map POI retrieval interface cyclically with each anchor point as the center and R=10km as the radius, and preset 10 candidate POIs. Step S3.2: Determine the rationality of 10 candidate POIs through three methods: (1) string similarity, (2) LLM semantic recognition, and (3) manual interactive verification; finally, summarize all anchor points and return POIs to generate a candidate point set for the keyword. ; Step S3.3: Filter and disambiguate POIs with the same name based on spherical distance, and preset the along-path distance threshold. T =5000m, for the candidate set Filter each one individually, according to the Haversine great circle distance calculation formula. Where R = 6371 km, the average radius of the Earth; λ Longitude ϕ Calculate candidate points using latitude. Distances to each trajectory point of L1 are retained, satisfying the following conditions. Points were selected, and invalid POIs with the same name in different locations exceeding 5000m were removed. The selected points were then placed within the set. The smallest POI is used as the original matching coordinate for this keyword. P raw,j ; Step S3.4: Use LLM / regular expressions to identify whether the named entity is a regular road or a highway, and retrieve the original coordinates of the representative points of the road segment. P raw,j , Substitute into the formula Solving for adsorption sites P snap,j ,use P snap,j Replace the original POI coordinates, and the snapped point will fall on the L1 main line trajectory.
[0012] Optionally, step S4, ordered coordinate rearrangement and L2 complete path generation specifically includes: Step S4.1: Sort the origin and destination, the corrected toll stations, and the successfully absorbed waypoints according to the L1 path travel direction to form a complete POI sequence; Step S4.2: Input the sorted route coordinate array, origin and destination coordinates, and heavy truck parameters into the map route planning API to generate the final commercial standard navigation route L2; the generated L2 route forms the actual freight trajectory because all waypoints are on the correct main line; Step S4.3: Output a structured waypoint table, an HTML interactive map, and an Excel waypoint detail table.
[0013] Optionally, the route point details table specifically includes: number, keywords, matching place name, latitude and longitude, distance from the route, and adsorption remarks.
[0014] This invention also provides a transportation-based fuzzy route intent parsing and path planning system, which applies the above-described transportation-based fuzzy route intent parsing and path planning method. The path planning system includes: The semantic parsing module is used for semantic parsing and directional reasoning of unstructured route text; The baseline generation module is used for the coordinate conversion of the start and end points, highway toll stations, and the generation of baseline route L1; The coordinate snapping module is used to search for and snap the coordinates of POIs along a 10km range by evenly setting anchor points along the route. The path generation module is used for ordered coordinate rearrangement and L2 complete path generation. This invention provides a fuzzy route intent parsing and path planning method and system based on transportation. The path planning method includes: step S1, semantic parsing and direction reasoning of unstructured route text; step S2, coordinate conversion of origin and destination, highway toll stations, and generation of baseline route L1; step S3, searching for waypoints and POI coordinates within a 10km range by uniformly setting anchor points along the route; and step S4, ordered coordinate rearrangement and generation of the complete L2 path. The system generates routes using natural language with a single click, along with a visualized map and Excel output, significantly reducing the manual verification costs in the approval process for oversized cargo transportation routes.
[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 A flowchart illustrating a transportation-based fuzzy route intent parsing and path planning method provided in an embodiment of the present invention; Figure 2 The process of generating the baseline route L1 provided in this embodiment of the invention; Figure 3 POI filtering provided in this embodiment of the invention; Figure 4 Adsorption of principal spatial coordinates provided in the embodiments of the present invention; Figure 5 This is a schematic diagram of the path point processing results provided in an embodiment of the present invention. Detailed Implementation
[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0019] The terms "comprising" and "having," and any variations thereof, in the specification, embodiments, claims, and drawings of this invention are intended to cover non-exclusive inclusion, such as including a series of steps or units.
[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.
[0021] Example 1 This invention provides a fuzzy route intent parsing and path planning method for oversized cargo transportation permit approval based on a four-layer decoupled architecture. Building upon semantic alignment achieved using a large AI model (LLM), it innovatively proposes a linear road segment adsorption algorithm based on two-level anchoring at core toll stations and the Haversine great circle distance, significantly eliminating deviation and detour problems and ensuring the absolute accuracy of oversized cargo transportation routes.
[0022] The technical solution adopted in this invention includes the following core decoupling layers and algorithm flow: S1. Semantic parsing and directional reasoning of unstructured route text.
[0023] S2, origin and destination, highway toll station coordinates conversion and baseline route L1 generation.
[0024] S3. Anchor points are evenly set along the route to search for and absorb the coordinates of POIs of road points and road sections within a 10km range.
[0025] S4, Ordered coordinate rearrangement and L2 complete path generation.
[0026] The system receives unstructured natural language route descriptions input from the oversized cargo transportation declaration system, constructs system prompts that include entity recognition and physical intent constraints, and performs named entity recognition by calling an AI Large Language Model (LLM) to extract an ordered sequence of path entities. The sequence of path entities includes the starting point, ending point, highway number, toll station, and ordinary POI. At the same time, the LLM infers the import / exit attributes of each toll station entity in the current route and marks them as "entrance" or "exit".
[0027] S1 includes: S1.1 Input the user's natural language route description, LLM splits the structured text, extracts the start point, end point, and waypoints, distinguishes the toll station entrance / exit types, and outputs JSON structured data {"start":"start point name","end":"end point name","via":["point 1","point 2",...]}; S1.2 Keyword Preprocessing: LLM adds entry / exit suffixes to text containing the "toll station" field to optimize the accuracy of subsequent POI retrieval; S1.3, obtain the starting text S, the ending text E, and the original path point WayTxt=[K1,K2...Km].
[0028] S2 includes: S2.1, Call the map's location retrieval API, and let the LLM determine the optimal selection of candidate points. Convert the starting point S and ending point E into latitude and longitude coordinates respectively. O , Des ; S2.2 Filter all toll station names within WayTxt, retrieve toll station coordinates in advance, and use LLM to determine the optimal selection of candidate points returned by the map API. After selection, use these as temporary anchor points to plan the baseline route, ensuring that the baseline route L1 passes through the designated toll station; S2.3 Generate a baseline route L1 that does not contain common waypoints. Parse the polyline trajectory string returned by the API and split it to obtain a discrete trajectory point set L1={p1,p2,...,pn}. L1 serves as the baseline for the entire process geometric calculation.
[0029] S3 includes: S3.1. Using a 20-equal-division uniform anchor point strategy, N=20 anchor points Anchor={A1,A2...A20} are uniformly sampled from the baseline route L1 trajectory list. For a single route keyword, the map POI retrieval interface is called cyclically with each anchor point as the center and R=10km as the radius, and 10 candidate POIs are preset.
[0030] S3.2. The rationality of 10 candidate POIs is determined through three methods: (1) string similarity, (2) LLM semantic recognition, and (3) manual interactive verification. This avoids having routes with obvious problems assigned to the map API for path planning. Finally, all anchor points are summarized and returned as POIs, generating a candidate point set for the keyword. .
[0031] S3.3, POI filtering and disambiguation based on spherical distance, with a preset along-path distance threshold. T =5000m, for the candidate set Filter them one by one. According to Haversine's formula for calculating great circle distance Where R = 6371 km, the average radius of the Earth; λ Longitude ϕ Calculate candidate points (for latitude) Distances to each trajectory point of L1 are retained, satisfying the following conditions. Points were selected, and invalid POIs with the same name in different locations exceeding 5000m were removed. The selected points were then placed within the set. The smallest POI is used as the original matching coordinate for this keyword. P raw,j ; S3.4. Use LLM / regular expressions to identify whether named entities are ordinary roads or expressways (the identification rules are consistent with the national expressway numbering method. For example, expressway numbers that begin with the letters G or S and contain 1, 2, or 4 digits), and retrieve the original coordinates of the representative points of the road segment. P raw,j , Substitute into the formula Solving for adsorption sites P snap,j ,use P snap,j Replace the original POI coordinates, and after snapping, the point will fall on the L1 main line trajectory, eliminating detours via ramps and toll stations.
[0032] S4 includes: S4.1 Sort the origin, destination, corrected toll stations, and successfully absorbed waypoints according to the L1 path travel direction to form a complete POI sequence.
[0033] S4.2. Input the sorted route coordinate array, origin and destination coordinates, and heavy truck parameters into the map route planning API to generate the final commercial standard navigation route L2. At this point, the generated L2 route forms a real freight trajectory without detours or deviations, and conforms to business logic, since all waypoints are on the correct main line.
[0034] S4.3 Output a structured waypoint table, an HTML interactive map, and an Excel waypoint detail table (number, keywords, matching place names, latitude and longitude, distance from the route, and snap notes).
[0035] Example 2 S1. Semantic parsing and directional reasoning of unstructured route text.
[0036] The system receives natural language route descriptions, form fields, or interface parameters from user input. If the input is in natural language, the system uses a large language model or preset prompts to extract the start point, end point, and via list from the text, and identifies the highway number, toll station name, and toll station entrance / exit intention.
[0037] The structured output is represented as follows: , Where A is the starting text, B is the ending text, and Q is the sequence of keywords along the route. Keyword q can be a highway number, toll station, road name, POI name, AOI name, or user-input coordinates.
[0038] S2, the coordinate conversion of the origin and destination place names and the generation of the baseline route L1.
[0039] The system calls the location text retrieval interface to parse A and B into start coordinates A and end coordinates B; when the location text retrieval fails, the geocoding interface is called as a backup. For scenarios involving oversized cargo transport or heavy vehicles, the system configures vehicle type, width, height, gross weight, load capacity, number of axles, and traffic restriction avoidance parameters, so that the route planning interface returns a route that complies with vehicle traffic constraints when possible. Keywords containing "toll station" or with toll station type tags are identified from Q to obtain a set of toll stations. S= {s1,...,sm}. Pre-search for keywords at each toll station and obtain the toll station coordinates. Then, prioritize using A, s1 to sm, B as path planning nodes to generate the L1 reference path. If there are no toll station anchor points, directly generate L1 from A to B, as shown in the appendix. Figure 2 As shown.
[0040] S3. Uniform sampling of discrete anchor points along the route and adsorption of spatial coordinates of the main line.
[0041] For the L1 baseline trajectory, extract 20 interval anchor points, call the map POI retrieval, and return no less than 10 candidate points. Taking candidate point P5 as an example, calculate the spherical distance between this POI and all points on the L1 trajectory point by point. If the threshold is satisfied, then the remaining locations of P are considered. >5000m directly eliminated, as shown in the attached document. Figure 3 As shown. For road segments, it is necessary to snap the candidate points with the minimum distance to the main trajectory points. Replace the original POI coordinates of the road segment representative points and snap them together. As attached Figure 4 As shown.
[0042] S4, Ordered coordinate rearrangement and L2 complete path generation.
[0043] In S3, all waypoint coordinates are pulled back to specific trajectory points on the L1 mainline. The system uses the index number of each waypoint in L1 to perform absolute ascending sorting (i.e., the Progress Index sorting algorithm), automatically rearranging the waypoint sequence according to the index value from smallest to largest. Even if the user's original input order is disordered, the final waypoint arrangement strictly matches the vehicle's driving trajectory from the starting point to the destination. Finally, the system inputs the sorted waypoint coordinate array, the starting and ending point coordinates, and the parameters of the oversized truck into the map path API to generate the final commercial standard navigation route L2. The waypoint processing result is as follows: Figure 5 As shown.
[0044] Beneficial effects: 1. The 4-layer architecture offers strong fault tolerance: It uses LLM to parse fuzzy path named entities, generates the L1 baseline route using the origin, destination, and highway toll stations as core anchor points, and uniformly samples the waypoints of the fuzzy path using discrete anchor points along the way to calibrate the waypoints. Finally, it rearranges the coordinates of the waypoints to generate the final complete path L2. The 4-layer architecture can significantly reduce the path planning error rate. 2. Segmented anchor point retrieval overcomes long-distance missed searches: The method of 20 equally divided anchor points + 10km overlap radius can effectively determine the rationality of waypoints, and the retrieval accuracy is not affected by the length of the route. 3. Adsorption of representative points for road segments to avoid detours: The main line trajectory of the baseline route is used to correct the offset POI coordinates, solving the industry pain point that the representative points of road segments fall on the auxiliary road / ramp / opposite lane; 4. Automatically sort waypoints using projection method: It does not depend on the user input order, but adjusts the order of passing points autonomously based on spatial location, eliminating detours caused by disordered input. 5. Full-process automation: One-click generation of routes using natural language, along with visual maps and Excel output, significantly reduces the manual verification costs in the approval process for oversized cargo transportation routes.
[0045] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A fuzzy route intent parsing and path planning method based on transportation, characterized in that, The path planning method includes: Step S1: Semantic parsing and directional reasoning of unstructured route text; Step S2: Generate coordinates for the origin and destination, highway toll station, and baseline route L1; Step S3: Evenly set anchor points along the route to search for and snap to the coordinates of POIs of road points and road sections within a 10km radius; Step S4: Ordered coordinate rearrangement and generation of L2 complete path.
2. The fuzzy route intent parsing and path planning method based on transportation according to claim 1, characterized in that, Step S1, semantic parsing and directional reasoning of the unstructured route text, includes: Receive unstructured natural language route descriptions input from the oversized cargo transportation declaration system, and construct system prompts that include entity recognition and physical intent constraints; Named entity recognition is performed by calling the AI large language model LLM; An ordered sequence of path entities is extracted; The AI large language model LLM infers the entrance / exit attributes of each toll station entity in the current route and marks them as "entrance" or "exit".
3. The fuzzy route intent parsing and path planning method based on transportation according to claim 2, characterized in that, The path entity sequence includes the starting point, the ending point, the highway number, the toll station, and the ordinary POI.
4. The fuzzy route intent parsing and path planning method based on transportation according to claim 1, characterized in that, Step S1, semantic parsing and directional reasoning of the unstructured route text, specifically includes: Step S1.1: Input the user's natural language route description, LLM splits the structured text, extracts the start point, end point, and waypoints, distinguishes the toll station entrance / exit types, and outputs JSON structured data {"start":"start point name","end":"end point name","via":["point 1","point 2",...]}; Step S1.2 Keyword preprocessing: The LLM adds entry / exit suffixes to text containing the "toll station" field to optimize the accuracy of subsequent POI retrieval; Step S1.3: Obtain the starting point text S, the ending point text E, and the original waypoint WayTxt=[K1,K2...Km].
5. The fuzzy route intent parsing and path planning method based on transportation according to claim 1, characterized in that, The specific steps of generating the origin and destination points, highway toll station coordinates, and baseline route L1 include: Step S2.1: Call the map's location retrieval API, and let LLM determine the optimal selection of candidate points. Convert the starting point S and ending point E into latitude and longitude coordinates respectively. O , Des ; Step S2.2: Filter all toll station names in WayTxt, retrieve toll station coordinates in advance, and use LLM to determine the optimal selection of candidate points returned by the map API; after selection, use them as temporary anchor points to plan the baseline route, ensuring that the baseline route L1 passes through the designated toll station; Step S2.3: Generate a baseline route L1 without common waypoints, parse the polyline trajectory string returned by the API, and split it to obtain a discrete trajectory point set L1={p1,p2,...,pn}. L1 serves as the geometric calculation baseline for the entire process.
6. The fuzzy route intent parsing and path planning method based on transportation according to claim 1, characterized in that, Step S3, which involves uniformly setting anchor points along the route to search for and absorb the coordinates of POIs (Points of Interest) along a 10km radius, specifically includes: Step S3.1: Using a 20-equal-division uniform anchor point strategy, uniformly sample N=20 anchor points Anchor={A1,A2...A20} from the baseline route L1 trajectory list. For a single route keyword, call the map POI retrieval interface cyclically with each anchor point as the center and R=10km as the radius, and preset 10 candidate POIs. Step S3.2: Determine the rationality of 10 candidate POIs through three methods: (1) string similarity, (2) LLM semantic recognition, and (3) manual interactive verification; finally, summarize all anchor points and return POIs to generate a candidate point set for the keyword. ; Step S3.3: Filter and disambiguate POIs with the same name based on spherical distance, and preset the along-path distance threshold. T =5000m, for the candidate set Filter each one individually, according to the Haversine great circle distance calculation formula. Where R = 6371 km, the average radius of the Earth; λ Longitude ϕ Calculate candidate points based on latitude. Distances to each trajectory point of L1 are retained, satisfying the following conditions. Points were selected, and invalid POIs with the same name in different locations exceeding 5000m were removed. The selected points were then placed within the set. The smallest POI is used as the original matching coordinate for this keyword. P raw,j ; Step S3.4: Use LLM / regular expressions to identify whether the named entity is a regular road or a highway, and retrieve the original coordinates of the representative points of the road segment. P raw,j , Substitute into the formula Solving for adsorption sites P snap,j ,use P snap,j Replace the original POI coordinates, and the snapped point will fall on the L1 main line trajectory.
7. The fuzzy route intent parsing and path planning method based on transportation according to claim 1, characterized in that, Step S4, ordered coordinate rearrangement and L2 complete path generation, specifically includes: Step S4.1: Sort the origin and destination, the corrected toll stations, and the successfully absorbed waypoints according to the L1 path travel direction to form a complete POI sequence; Step S4.2: Input the sorted route coordinate array, origin and destination coordinates, and heavy truck parameters into the map route planning API to generate the final commercial standard navigation route L2; the generated L2 route forms the actual freight trajectory because all waypoints are on the correct main line; Step S4.3: Output a structured waypoint table, an HTML interactive map, and an Excel waypoint detail table.
8. The fuzzy route intent parsing and path planning method based on transportation according to claim 7, characterized in that, The detailed list of waypoints includes: number, keywords, matching place names, latitude and longitude, distance from the route, and adsorption remarks.
9. A transportation-based fuzzy route intent parsing and path planning system, employing the aforementioned transportation-based fuzzy route intent parsing and path planning method, characterized in that, The path planning system includes: The semantic parsing module is used for semantic parsing and directional reasoning of unstructured route text; The baseline generation module is used for the coordinate conversion of the start and end points, highway toll stations, and the generation of baseline route L1; The coordinate snapping module is used to search for and snap the coordinates of POIs along a 10km range by evenly setting anchor points along the route. The path generation module is used for ordered coordinate rearrangement and L2 complete path generation.