Container truck transportation path mining and retrieval method and system based on GPS and industry characteristics

Abstract

The invention provides a container truck transportation path mining and retrieval method and system based on GPS and industry characteristics, and remarkable optimization and improvement are realized aiming at unique requirements of the container truck transportation industry. According to the method, the experience route is extracted by adopting a geo-fencing technology and constructing a Cartesian set method independent of the waybill, so that the coverage range of the experience route is enhanced. The scheme aims to meet the requirements of container transportation on specific road conditions, including but not limited to height limitation, weight limitation, special road conditions, special charging modes and the like. According to the method, a route planning strategy with high adaptability is specially designed, the safety and efficiency of the transportation process are effectively improved, and meanwhile the high-coverage route retrieval capability is achieved. Compared with the prior art, the scheme has the advantages that geographic fence information of a wharf and a storage yard and abundant historical GPS track data are deeply fused, and a public transportation track mode is accurately mined and constructed.

Description

Technical Field

[0001] The present invention relates to the field of traffic data mining based on GPS technology, and in particular to a method and system for mining and retrieving container truck transportation routes based on GPS and industry characteristics. Background Art

[0002] With the rapid development of the logistics industry, the transportation efficiency and cost control of container trailers have become the key to corporate competition. However, the navigation software currently on the market has obvious limitations in planning container transportation routes, which seriously restrict the improvement of transportation efficiency and the reduction of costs. Existing navigation software often ignores the unique needs of container transportation, such as vehicle size, load restrictions, road conditions, toll settlement methods and other factors. This leads to a significant gap between the navigation route and the actual transportation route of the driver, especially the navigation route is often longer than the actual route, which not only increases the transportation time and cost, but also may cause the vehicle to drive illegally or fall into sections that are not suitable for large vehicles. Although some studies have attempted to improve route planning problems through algorithm optimization, most of these studies lack industry specificity and actual application scenario verification. More importantly, these algorithms often fail to fully consider factors such as the actual road conditions and the driver's driving habits, so the effect in practice needs to be further optimized. Patent CN202011620267.2 uses a combination of algorithms such as DBSCAN, LCSS, Dijkstra and aprior to calculate the vehicle's frequently traveled routes, thereby improving the performance indicators in the field of commercial vehicle frequently traveled routes. However, it does not combine actual application scenarios to perform optimal route planning for specific areas. Existing path planning technology is difficult to obtain the optimal planning and navigation path in the field of container trailers. Summary of the invention

[0003] In order to meet the route planning needs of container trailers, solve the irrationality of current general navigation software in route planning, and avoid the problem of inaccurate actual mileage calculation during mileage metering, the present invention proposes a method and system for mining and retrieving container truck transportation routes based on GPS and industry characteristics, combining the characteristics of the container transportation industry. Taking the docks, yards and other operating scenes involved in the starting point of the task as the entry point, the GPS trajectory of the trailer is used for pattern mining, and more reasonable route planning is carried out by drawing on experience routes.

[0004] The specific plan is as follows:

[0005] A method for mining and retrieving container truck transport routes based on GPS and industry characteristics.

[0006] S1: Collect and access container trailer trajectory data, surrounding environment data, and truck transportation task data: including container trailer trajectory data formed by the GPS data of the container trailer, dock and yard geographic fence data, and truck transportation task data;

[0007] S2: Preprocess the data in S1 and construct a Cartesian set of candidate ODs that are separated from the waybill:

[0008] 1) Mining the stop points that exceed the set stay time in the historical data of container trailer trajectories as key nodes of the transportation task, preliminarily filtering the stop points through the map service, and encoding the remaining stop points as potential customer door-to-door stop points;

[0009] 2) any container trailer track entering or leaving the geographical fence of the terminal and the yard is truncated and split into container trailer sub-tracks; the geographical fence of the terminal and the yard marks the logical starting point or end point of the transportation task;

[0010] 3) constructing a Cartesian set of candidate ODs that are separated from the waybill based on each of the split container trailer sub-trajectories;

[0011] 4) Screen candidate ODs based on actual business data so that the control route subject is consistent with historical transportation experience as a candidate trajectory;

[0012] S3: Mining and extracting common sub-trajectories: In the candidate OD Cartesian set, for a group of transportation task trajectories with the same starting point and end point, trajectory similarity evaluation and cluster analysis are performed, trajectory distance is calculated using the Hausdoriff method, and then cluster analysis of trajectory distance is performed, the density of trajectory points in each cluster is calculated for optimization, and the trajectory with the shortest transportation time is selected as the benchmark trajectory, and then the common sub-trajectory is verified and restored based on the number of Geohash blocks to obtain an effective empirical route;

[0013] S4: Optimize empirical trajectories and enhance processing: Use intelligent denoising algorithms to automatically identify and delete redundant trajectory segments within a preset range around OD; use third-party route planning services to supplement the remaining trajectory information;

[0014] S5: Trajectory retrieval for nearby transportation tasks: Search for the OD experience trajectory centered on the target OD and use it as the main part of the planned trajectory; use the third-party route planning service to solve the problem of missing head and tail near the starting and ending points of the trajectory, thereby improving the experience trajectory and forming a complete route that conforms to historical business experience.

[0015] Preferably, the method for collecting and accessing container trailer data, surrounding environment data and container truck transport task data is:

[0016] S11: The location data sent by the truck connected to the GPS signal box. The collection time interval between every two GPS points is between 2-10 minutes, including the real-time latitude and longitude of the vehicle, sending timestamp, direction, and speed;

[0017] S12: The dock geo-fence data is manually maintained static basic data, and the shape of the geo-fence is the actual geographical shape of the dock;

[0018] S13: The truck transport task data includes: the starting and ending terminals of the transport, the customer's location, the required time of the transport task, and the trailer license plate number information.

[0019] Preferably, the preprocessing method in S2 is:

[0020] S21: Stop point mining and position standardization: Use the Traj-DBSCAN algorithm to mine significant stop points in the historical trajectory of the container truck as key nodes of the transportation task, and use Geohash technology to encode the position of each stop point;

[0021] S22: Task boundary definition and trajectory screening: Based on the conventional business model of container truck transportation, the actual geographical fence of the terminal is set as the logical starting point or end point of the transportation task; by setting a 48-hour time sliding window and searching for trajectory data across the fences of two terminals within this time period, the OD pairs formed by the Geohash of the stop points in these trajectories are marked and a candidate set of historical sub-trajectories is constructed;

[0022] S23: Track quality control and optimization: Double filtering is performed on candidate tracks to exclude tracks with abnormal transportation time; and tracks with insufficient GPS data points, i.e., less than the set value, are eliminated.

[0023] Preferably, the method for mining and extracting common sub-trajectories in S3 is: for a group of transport task trajectories with the same starting point and end point Geohash, if the number of the group of transport trajectories is greater than or equal to 3, mining common sub-trajectories is performed.

[0024] Preferably, the method for mining and extracting common sub-tracks in S3 is:

[0025] S31: Trajectory similarity evaluation and clustering: Hausdorff distance is used to calculate the similarity between each trajectory in the same group; based on the similarity of the trajectories, the DBSCAN algorithm is used to cluster the trajectories, and clusters are obtained according to the distance similarity. Isolated points are removed, and the top 5 clusters containing the most transport trajectories are selected for analysis;

[0026] S32: Track density screening and determination of reference track: Calculate the track density in each cluster, where the track density is the track transportation time divided by the number of GPS points, and give priority to retaining tracks with higher density; if there are less than 3 high-density tracks, gradually relax the density restriction; and select the track with the shortest transportation time as the reference track in the clusters that meet the conditions;

[0027] S33: Geohash-based trajectory comparison and extraction of common segments: Geohash encoding is performed on all trajectories, and each trajectory encoding is compared with the reference trajectory encoding one by one, and the Geohash blocks that are the same or adjacent to the reference trajectory encoding are retained to construct the encoding framework of potential common sub-trajectories;

[0028] S34: Verify and restore the common sub-trajectory: Ensure that the first and last Geohash blocks of the common coding result are consistent with the reference trajectory, and automatically supplement them if they are inconsistent; at the same time, check the number of Geohash blocks in the common coding result for verification. If the verification passes, use the reference trajectory and the verified Geohash coding result to restore the common sub-trajectory; and determine whether it is a valid empirical route based on the number of GPS points in the restored common sub-trajectory.

[0029] Preferably, the verification requirement in S34 is: there are at least 3 Geohash blocks and they cover more than 30% of the reference track code; the method for judging the valid empirical route is: if the number of GPS points in the restored common sub-track is not less than 5, it is judged as a valid empirical route.

[0030] Preferably, the method for optimizing the experience trajectory and enhancing the processing in S4 is:

[0031] S41: After obtaining the empirical route described in S3, an intelligent denoising algorithm is used to eliminate unnecessary factors due to pathfinding and temporary stops near the start and end points, and redundant trajectory segments within a preset range around the OD are automatically identified and deleted;

[0032] S42: Use a third-party route planning service to supplement the track information, including supplementing track points, calculating track distance, counting the number of track GPS points, and transportation time;

[0033] Preferably, the method for retrieving the trajectory of the adjacent transport task in S5 is:

[0034] S51: When performing trajectory retrieval, taking the target OD as the center, setting the search radius and drawing the search area, looking for existing and settled OD experience trajectories; if an experience trajectory that meets the conditions is retrieved, it is used as the main part of the planning trajectory;

[0035] S52: Improve the experience track: Use a third-party route planning service to solve the problem of missing heads and tails near the start and end points of the track, and ultimately form a complete route that conforms to historical business experience.

[0036] A truck transport route mining and retrieval system based on GPS and industry characteristics, including

[0037] Data collection module: used to collect container trailer trajectory data, terminal and yard geo-fence data, and truck transport task data formed by GPS data of container trailers;

[0038] Preprocessing module: used to process the container trailer trajectory data, terminal and yard geo-fence data and truck transport task data in the data acquisition module: including: mining the stop points that exceed the set stay time in the historical data of container trailer trajectory as the key nodes of the transport task, preliminarily filtering the stop points through the map service, and encoding the remaining stop points as potential customer door-point stop points; a location encoding sub-module that truncates any container trailer trajectory that enters and exits the geographical fence of the terminal and yard, and splits it to obtain container trailer sub-trajectories; a candidate OD sub-module that constructs a Cartesian set of candidate ODs that are separated from the waybill based on each split container trailer sub-trajectory; a candidate trajectory sub-module that screens candidate ODs based on actual business data so that the control route body conforms to historical transportation experience as a candidate trajectory;

[0039] Mining and extracting common sub-track modules: including: evaluating and clustering the trajectory similarity of a group of transport task trajectories with the same starting and ending Geohash, using Hausdoriff to calculate the trajectory distance, calculating the density of trajectory points in each cluster for optimization, and selecting the trajectory with the shortest transportation time as the benchmark trajectory, and then verifying and restoring the common sub-trajectories based on the number of Geohash blocks to obtain the empirical route sub-module of the effective empirical route;

[0040] Optimize the empirical trajectory and enhance the processing module: including: a redundancy removal submodule that uses an intelligent denoising algorithm to automatically identify and delete redundant trajectory segments within a preset range around the OD; a trajectory enhancement submodule that uses a third-party route planning service to supplement trajectory information;

[0041] The trajectory retrieval module for the adjacent transportation task includes: a planning trajectory submodule that searches for the OD experience trajectory with the target OD as the center and uses it as the main part of the planning trajectory; an experience route submodule that uses a third-party route planning service to solve the problem of missing head and tail near the starting and ending points of the trajectory, thereby improving the experience trajectory and forming a transportation route that conforms to historical business experience; and a route display submodule that displays the transportation route that conforms to historical experience.

[0042] Beneficial effects:

[0043] The present invention proposes a method and system for mining and retrieving container truck transportation routes based on GPS and industry characteristics, which achieves significant optimization and improvement in view of the unique needs of the container truck transportation industry. The present invention adopts geo-fencing technology and constructs a Cartesian set method independent of the waybill to extract empirical routes, thereby enhancing the coverage of empirical routes. The solution is designed to meet the needs of container transportation for specific road conditions, including but not limited to height restrictions, weight restrictions, special road conditions, and special charging modes. The present invention specially designs a highly adaptable route planning strategy, which effectively improves the safety and efficiency of the transportation process, and has a high-coverage route retrieval capability. Compared with the existing technology, the solution accurately mines and constructs a public transportation trajectory model by deeply integrating the geo-fencing information of the terminal and the yard and the rich historical GPS trajectory data.

[0044] Firstly, based on the fact that docks and yards are limited, all tasks need to start or end at the dock / yard, and the present invention is based on finding all transportation tasks by cutting off the trajectory through geographic fences.

[0045] Secondly, the present invention uses AOI truncation technology to construct a Cartesian set separated from the waybill. Since the trajectory between two yards or terminals must involve at least one transportation task, the present invention directly identifies the transportation task from these trajectories and constructs a Cartesian set, rather than relying on the order information maintained in the system. In this way, all potential transportation task routes can be extracted for subsequent analysis. This solution solves the problem that waybill information is often inaccurate in the field of container transportation. For example, a driver may be assigned two transportation tasks at the same time, but the order in which they are actually executed may not follow the instructions on the waybill; the empty yard specified on the waybill may be changed to another yard due to waiting in line or lack of containers; the manual update of the waybill information may not have been reflected in the system. Usually, the trajectory segments of non-transportation routes will be excluded in the subsequent cluster analysis. The test results show that through this method, the present invention has achieved more than 90% coverage of empirical routes. In contrast, the coverage of empirical routes generated using waybill information is less than 70%, which fully proves its effectiveness and reliability in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flow chart of a method for mining and retrieving container truck transport routes based on GPS and industry characteristics.

[0047] Figure 2 Algorithm flow chart for searching OD pairs of the stay point hash within the time period.

[0048] Figure 3Flow chart of obtaining the experience route in the embodiment.

[0049] Figure 4 A truck transport route mining and retrieval system based on GPS and industry characteristics. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0051] The present invention uses the GPS data of container trailers, dock geo-fence data and container truck transport task data, performs calculation and analysis according to relevant calculation methods, and finally obtains one or more empirical routes of an OD (origin to destination) pair.

[0052] like Figure 1 and 3 As shown, the present invention is mainly divided into the following steps:

[0053] A method for mining and retrieving container truck transport routes based on GPS and industry characteristics.

[0054] S1: Collect and access container trailer trajectory data, surrounding environment data, and truck transportation task data: including container trailer trajectory data formed by the GPS data of the container trailer, dock and yard geographic fence data, and truck transportation task data;

[0055] S2: Preprocess the data in S1 and construct a Cartesian set of candidate ODs that are separated from the waybill:

[0056] 1) Mining the stop points that exceed the set stay time in the historical data of container trailer trajectories as key nodes of the transportation task, preliminarily filtering the stop points through the map service, and encoding the remaining stop points as potential customer door-to-door stop points;

[0057] 2) any container trailer track entering or leaving the geographical fence of the terminal and the yard is truncated and split into container trailer sub-tracks; the geographical fence of the terminal and the yard marks the logical starting point or end point of the transportation task;

[0058] 3) constructing a Cartesian set of candidate ODs that are separated from the waybill based on each of the split container trailer sub-trajectories;

[0059] 4) Screen candidate ODs based on actual business data so that the control route subject is consistent with historical transportation experience as a candidate trajectory;

[0060] S3: Mining and extracting common sub-trajectories: In the candidate OD Cartesian set, for a group of transportation task trajectories with the same starting point and end point, trajectory similarity evaluation and cluster analysis are performed, trajectory distance is calculated using the Hausdoriff method, and then cluster analysis of trajectory distance is performed, the density of trajectory points in each cluster is calculated for optimization, and the trajectory with the shortest transportation time is selected as the benchmark trajectory, and then the common sub-trajectory is verified and restored based on the number of Geohash blocks to obtain an effective empirical route;

[0061] S4: Optimize empirical trajectories and enhance processing: Use intelligent denoising algorithms to automatically identify and delete redundant trajectory segments within a preset range around OD; use third-party route planning services to supplement the remaining trajectory information;

[0062] S5: Trajectory retrieval for nearby transportation tasks: Search for the OD experience trajectory centered on the target OD and use it as the main part of the planned trajectory; use the third-party route planning service to solve the problem of missing head and tail near the starting and ending points of the trajectory, thereby improving the experience trajectory and forming a complete route that conforms to historical business experience.

[0063] The detailed details are as follows:

[0064] S1: Collect and access container trailer trajectory data, surrounding environment data, and truck transportation task data: The data required by this algorithm includes container trailer GPS data, terminal geo-fence data, and truck transportation task data:

[0065] a. The location data sent by the container truck connected to the GPS signal box, the collection time interval between every two GPS points is between 2-10 minutes, including the vehicle's real-time latitude and longitude, sending timestamp, direction, speed, etc.

[0066] b. The dock geo-fence data is manually maintained static basic data. The shape of the geo-fence is a polygon, which can be understood as the actual geographical shape of the dock.

[0067] c. Truck transport task data includes the starting and ending terminals of the transport, customer location, required time of the transport task, trailer license plate number information, etc.

[0068] S2: Preprocessing the container trailer trajectory data, surrounding environment data, and container truck transportation task data described in S1: Stop point mining and position standardization: First, the Traj-DBSCAN algorithm is used to deeply mine the significant stop points in the historical trajectory of the container truck, which often mark the key nodes of the transportation task. Then, the Geohash technology is used to encode the position of each stop point for subsequent comparison and analysis.

[0069] a. Task boundary definition and trajectory screening: Based on the conventional business model of container truck transportation, the actual geographical fence of the terminal is set as the logical starting point or end point of the transportation task. By setting a 48-hour time sliding window, we search for trajectory data spanning the fences of two terminals within this time period, paying special attention to the OD pairs formed by the Geohash of the stop points in these trajectories. Construct a series of potential candidate sets of historical sub-trajectories with practical business significance. (e.g. Figure 2 , preprocess the GPS data, use every 48 hours as a sliding window, and search for the OD pairs of the Hash of the stay point within the time period)

[0070] b. Trajectory quality control and optimization: In order to ensure the quality of the constructed empirical routes, the candidate trajectories are double filtered. On the one hand, those trajectories with abnormally long transportation time (more than 48 hours) or too short (less than 30 minutes) are excluded, because these trajectories may not be representative due to special reasons (such as failures, detours, etc.). On the other hand, trajectories with insufficient GPS data points (for example, less than 10) are eliminated, because these trajectories may have incomplete data due to signal problems, affecting the accuracy of the analysis results.

[0071] And remove the tracks with insufficient GPS data points, i.e., those with less than the set value.

[0072] The candidate ODs include:

[0073] Sub-trajectory set = {T1, T2}

[0074] The stop point set of sub-trajectory T1 = {pier 1, stop point 1, stop point 2, pier 2}

[0075] The stop point set of sub-trajectory T2 = {pier 3, stop point 3, pier 4}

[0076] Through the Cartesian product, all possible OD combinations of T1 are obtained:

[0077] T1 candidate OD combination = {(pier 1, stop 1), (stop 1, stop 2), (stop 2, pier 2), (pier 1, stop 2), (pier 1, pier 2), (stop 2, pier 2)}

[0078] T2 candidate OD combination = {(pier 3, stop point 3), (stop point 3, pier 4), (pier 3, stop point 4)}

[0079] Then all possible OD combinations are:

[0080] Candidate OD combination = T1 candidate OD combination × T2 candidate OD combination = {(terminal 1, stop point 1), (stop point 1, stop point 2), (stop point 2, terminal 2), (terminal 1, stop point 2), (terminal 1, terminal 2), (stop point 2, terminal 2), (terminal 3, stop point 3), (stop point 3, terminal 4), (terminal 3, stop point 4)}.

[0081] S3: Mining and extracting common sub-trajectories: For a group of transport task trajectories with the same starting and ending Geohash, if the number of transport trajectories in the group is greater than or equal to 3, the common sub-trajectory mining is performed according to the following process:

[0082] a. Trajectory similarity evaluation and clustering: Hausdorff distance is used to calculate the similarity between each trajectory in this set of trajectories. Based on the similarity of trajectories, DBSCAN algorithm is used to cluster the trajectories, and clusters are obtained according to distance similarity. Isolated points are removed, and the top 5 clusters containing the most transport trajectories are selected for in-depth analysis.

[0083] b. Trajectory density screening and base trajectory determination: Calculate the density of the trajectories in each cluster (the trajectory transportation time divided by the number of GPS points), and give priority to retaining trajectories with higher density to ensure the compactness and representativeness of the trajectory. If there are less than 3 high-density trajectories, gradually relax the density restrictions (for example, first to ≤150, then to ≤300, and finally to no restrictions) to expand the candidate range. Among the clusters that meet the conditions, select the trajectory with the shortest transportation time as the base trajectory, and use it as a reference for subsequent analysis.

[0084] a. Geohash-based trajectory comparison and common segment extraction: All trajectories are Geohash-encoded to facilitate efficient comparison of spatial positions. Each trajectory code is compared with the reference trajectory code one by one, and the Geohash blocks that are the same or adjacent to the reference trajectory code are retained to construct the coding framework of potential common sub-trajectories.

[0085] b. Verification and restoration of common sub-trajectory: Ensure that the first and last Geohash blocks of the common coding result are consistent with the reference trajectory. If they are inconsistent, they will be automatically supplemented to ensure the integrity of the trajectory. At the same time, check the number of Geohash blocks in the common coding result, which must be at least 3 and cover more than 30% of the reference trajectory coding to ensure the significance and practicality of the common sub-trajectory. If the verification passes, the reference trajectory and the verified Geohash coding result are used to restore the specific common sub-trajectory. If the number of GPS points in the restored trajectory is not less than 5, it is considered a valid empirical route.

[0086] It can be used for path planning and optimization of subsequent transportation tasks.

[0087] S4: Optimize experience trajectory and enhance processing:

[0088] a. After obtaining the preliminary empirical route, in order to eliminate detours caused by unnecessary factors such as path finding and temporary stops near the starting and ending points, we use an intelligent denoising algorithm to automatically identify and delete redundant trajectory segments within a preset range around the OD.

[0089] b. In order to further improve the completeness and accuracy of the empirical trajectory, a third-party route planning service (such as Baidu Eagle Eye) is used to supplement the trajectory information. The supplementary content includes supplementing trajectory points, calculating trajectory distance, counting the number of trajectory GPS points, transportation time and other information.

[0090] S5: Trajectory retrieval of nearby transport tasks:

[0091] When performing trajectory retrieval, a reasonable search radius is set with the target OD as the center, and a search area is drawn to find out whether there are existing and settled OD experience trajectories. Once an experience trajectory that meets the conditions is retrieved, it is used as the main part of the planning trajectory.

[0092] a. To further improve this track, a third-party route planning service is used to solve the problem of missing the beginning and end of the track near the starting and ending points, and finally form a complete route that conforms to historical business experience.

[0093] Taking the trajectory of an OD as an example, we will obtain the most frequently used route from multiple transport trajectories. The public sub-trajectory is mainly used to control the main route to conform to historical experience. The missing parts of the OD surrounding and public trajectory points are forced to be completed through the third-party route planning service to reduce detours.

[0094] When searching for trajectories, draw a circle with a certain radius to find out whether there are any accumulated OD experience trajectories in the surrounding POIs.

[0095] like Figure 4 As shown, the present invention also protects a container truck transportation path mining and retrieval system based on GPS and industry characteristics, including:

[0096] Data collection module: used to collect container trailer trajectory data, terminal and yard geo-fence data, and truck transport task data formed by GPS data of container trailers;

[0097] Preprocessing module: used to process the container trailer trajectory data, terminal and yard geo-fence data and truck transport task data in the data acquisition module: including: mining the stop points that exceed the set stay time in the historical data of container trailer trajectory as the key nodes of the transport task, preliminarily filtering the stop points through the map service, and encoding the remaining stop points as potential customer door-point stop points; a location encoding sub-module that truncates any container trailer trajectory that enters and exits the geographical fence of the terminal and yard, and splits it to obtain container trailer sub-trajectories; a candidate OD sub-module that constructs a Cartesian set of candidate ODs that are separated from the waybill based on each split container trailer sub-trajectory; a candidate trajectory sub-module that screens candidate ODs based on actual business data so that the control route body conforms to historical transportation experience as a candidate trajectory;

[0098] Mining and extracting common sub-track modules: including: evaluating and clustering the trajectory similarity of a group of transport task trajectories with the same starting and ending Geohash, using Hausdoriff to calculate the trajectory distance, calculating the density of trajectory points in each cluster for optimization, and selecting the trajectory with the shortest transportation time as the benchmark trajectory, and then verifying and restoring the common sub-trajectories based on the number of Geohash blocks to obtain the empirical route sub-module of the effective empirical route;

[0099] Optimize the empirical trajectory and enhance the processing module: including: a redundancy removal submodule that uses an intelligent denoising algorithm to automatically identify and delete redundant trajectory segments within a preset range around the OD; a trajectory enhancement submodule that uses a third-party route planning service to supplement trajectory information;

[0100] The trajectory retrieval module for the adjacent transportation task includes: a planning trajectory submodule that searches for the OD experience trajectory with the target OD as the center and uses it as the main part of the planning trajectory; an experience route submodule that uses a third-party route planning service to solve the problem of missing head and tail near the starting and ending points of the trajectory, thereby improving the experience trajectory and forming a transportation route that conforms to historical business experience; and a route display submodule that displays the transportation route that conforms to historical experience.

[0101] The above specific implementation modes are preferred embodiments of the present invention and cannot be used to limit the present invention. Any other changes or other equivalent replacement methods that do not deviate from the technical solution of the present invention are included in the protection scope of the present invention.

Claims

1. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics, characterized in that: S1: Collect and access container trailer trajectory data, surrounding environment data, and truck transportation task data: including container trailer trajectory data formed by the GPS data of the container trailer, terminal and yard geographic fence data, and truck transportation task data; S2: Preprocess the data in S1 and construct a Cartesian set of candidate ODs that are separated from the waybill: 1) Mining the stop points that exceed the set stay time in the historical data of container trailer trajectories as key nodes of the transportation task, preliminarily filtering the stop points through the map service, and encoding the remaining stop points as potential customer door-to-door stop points; 2) any container trailer track entering or leaving the geographical fence of the terminal and the yard is truncated and split into container trailer sub-tracks; the geographical fence of the terminal and the yard marks the logical starting point or end point of the transportation task; 3) constructing a Cartesian set of candidate ODs that are separated from the waybill based on each of the split container trailer sub-trajectories; 4) Screen candidate ODs based on actual business data so that the control route subject is consistent with historical transportation experience as a candidate trajectory; S3: Mining and extracting common sub-trajectories: In the candidate OD Cartesian set, for a group of transportation task trajectories with the same starting point and end point, trajectory similarity evaluation and cluster analysis are performed, trajectory distance is calculated using the Hausdoriff method, and then cluster analysis of trajectory distance is performed, the density of trajectory points in each cluster is calculated for optimization, and the trajectory with the shortest transportation time is selected as the benchmark trajectory, and then the common sub-trajectory is verified and restored based on the number of Geohash blocks to obtain an effective empirical route; S4: Optimize empirical trajectories and enhance processing: Use intelligent denoising algorithm to automatically identify and delete redundant trajectory segments within a preset range around OD; Use third-party route planning services to supplement the remaining trajectory information; S5: Trajectory retrieval for nearby transportation tasks: search for OD empirical trajectories centered on the target OD and use them as the main part of the planned trajectory; Use third-party route planning services to solve the problem of missing heads and tails near the starting and ending points of the trajectory, thereby improving the empirical trajectory and forming a complete route that conforms to historical business experience.

2. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The method for collecting and accessing container trailer data, surrounding environment data, and container truck transportation task data is as follows: S11: The location data sent by the truck connected to the GPS signal box. The collection time interval between every two GPS points is between 2-10 minutes, including the real-time latitude and longitude of the vehicle, sending timestamp, direction, and speed; S12: The dock geo-fence data is manually maintained static basic data, and the shape of the geo-fence is the actual geographical shape of the dock; S13: The truck transport task data includes: the starting and ending terminals of the transport, the customer's location, the required time of the transport task, and the trailer license plate number information.

3. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The preprocessing method in S2 is: S21: Stop point mining and position standardization: Use the Traj-DBSCAN algorithm to mine significant stop points in the historical trajectory of the container truck as key nodes of the transportation task, and use Geohash technology to encode the position of each stop point; S22: Task boundary definition and trajectory screening: Based on the conventional business model of container truck transportation, the actual geographical fence of the terminal is set as the logical starting point or end point of the transportation task; by setting a 48-hour time sliding window and searching for trajectory data across the fences of two terminals within this time period, the OD pairs formed by the Geohash of the stop points in these trajectories are marked and a candidate set of historical sub-trajectories is constructed; S23: Track quality control and optimization: Double filtering is performed on candidate tracks to exclude tracks with abnormal transportation time; and tracks with insufficient GPS data points, i.e., less than the set value, are eliminated.

4. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: In S2, the candidate ODs include: Sub-trajectory set = {T1, T2} The stop point set of sub-trajectory T1 = {pier 1, stop point 1, stop point 2, pier 2} The stop point set of sub-trajectory T2 = {pier 3, stop point 3, pier 4} Through the Cartesian product, all possible OD combinations of T1 are obtained: T1 candidate OD combination = {(pier 1, stop 1), (stop 1, stop 2), (stop 2, pier 2), (pier 1, stop 2), (pier 1, pier 2), (stop 2, pier 2)} T2 candidate OD combination = {(pier 3, stop point 3), (stop point 3, pier 4), (pier 3, stop point 4)} Then all possible OD combinations are: Candidate OD combination = T1 candidate OD combination × T2 candidate OD combination = {(terminal 1, stop point 1), (stop point 1, stop point 2), (stop point 2, terminal 2), (terminal 1, stop point 2), (terminal 1, terminal 2), (stop point 2, terminal 2), (terminal 3, stop point 3), (stop point 3, terminal 4), (terminal 3, stop point 4)}.

5. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The method of mining and extracting common sub-trajectories in S3 is as follows: for a group of transportation task trajectories with the same starting point and end point Geohash, if the number of the group of transportation trajectories is greater than or equal to 3, then common sub-trajectories are mined.

6. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The method for mining and extracting common sub-tracks in S3 is: S31: Trajectory similarity evaluation and clustering: Hausdorff distance is used to calculate the similarity between each trajectory in the same group; based on the similarity of the trajectories, the DBSCAN algorithm is used to cluster the trajectories, and clusters are obtained according to the distance similarity. Isolated points are removed, and the top 5 clusters containing the most transport trajectories are selected for analysis; S32: Track density screening and determination of reference track: Calculate the track density in each cluster, where the track density is the track transportation time divided by the number of GPS points, and give priority to retaining tracks with higher density; if there are less than 3 high-density tracks, gradually relax the density restriction; and select the track with the shortest transportation time as the reference track in the clusters that meet the conditions; S33: Geohash-based trajectory comparison and extraction of common segments: Geohash encoding is performed on all trajectories, and each trajectory encoding is compared with the reference trajectory encoding one by one, and the Geohash blocks that are the same or adjacent to the reference trajectory encoding are retained to construct the encoding framework of potential common sub-trajectories; S34: Verify and restore the common sub-trajectory: Ensure that the first and last Geohash blocks of the common coding result are consistent with the reference trajectory, and automatically supplement them if they are inconsistent; at the same time, check the number of Geohash blocks in the common coding result for verification. If the verification passes, use the reference trajectory and the verified Geohash coding result to restore the common sub-trajectory; and determine whether it is a valid empirical route based on the number of GPS points in the restored common sub-trajectory.

7. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The verification requirement in S34 is: there are at least 3 Geohash blocks and they cover more than 30% of the reference track code; the method for judging the valid empirical route is: if the number of GPS points in the restored common sub-track is not less than 5, it is judged as a valid empirical route.

8. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The method for optimizing the experience trajectory and enhancing the processing in S4 is: S41: After obtaining the empirical route described in S3, an intelligent denoising algorithm is used to eliminate unnecessary factors due to pathfinding and temporary stops near the start and end points, and redundant trajectory segments within a preset range around the OD are automatically identified and deleted; S42: Use a third-party route planning service to supplement the track information, including supplementing track points, calculating track distance, counting the number of track GPS points, and transportation time.

9. A method for mining and retrieving container truck transportation routes based on GPS and industry characteristics according to claim 1, characterized in that: The method for retrieving the trajectory of the adjacent transportation task in S5 is: S51: When performing trajectory retrieval, taking the target OD as the center, setting the search radius and drawing the search area, looking for existing and settled OD experience trajectories; if an experience trajectory that meets the conditions is retrieved, it is used as the main part of the planning trajectory; S52: Improve the experience track: Use a third-party route planning service to solve the problem of missing heads and tails near the start and end points of the track, and ultimately form a complete route that conforms to historical business experience.

10. A container truck transport route mining and retrieval system based on GPS and industry characteristics according to any one of claims 1 to 8, characterized in that: include Data collection module: used to collect container trailer trajectory data, terminal and yard geo-fence data, and truck transport task data formed by GPS data of container trailers; Preprocessing module: used to process the container trailer trajectory data, terminal and yard geo-fence data and truck transport task data in the data acquisition module: including: mining the stop points that exceed the set stay time in the historical data of container trailer trajectory as the key nodes of the transport task, preliminarily filtering the stop points through the map service, and encoding the remaining stop points as potential customer door-point stop points; a location encoding sub-module that truncates any container trailer trajectory that enters and exits the geographical fence of the terminal and yard, and splits it to obtain container trailer sub-trajectories; a candidate OD sub-module that constructs a Cartesian set of candidate ODs that are separated from the waybill based on each split container trailer sub-trajectory; a candidate trajectory sub-module that screens candidate ODs based on actual business data so that the control route body conforms to historical transportation experience as a candidate trajectory; Mining and extracting common sub-track modules: including: evaluating and clustering the trajectory similarity of a group of transport task trajectories with the same starting and ending Geohash, using Hausdoriff to calculate the trajectory distance, calculating the density of trajectory points in each cluster for optimization, and selecting the trajectory with the shortest transportation time as the benchmark trajectory, and then verifying and restoring the common sub-trajectories based on the number of Geohash blocks to obtain the empirical route sub-module of the effective empirical route; Optimize the empirical trajectory and enhance the processing module: including: a redundancy removal submodule that uses an intelligent denoising algorithm to automatically identify and delete redundant trajectory segments within a preset range around the OD; a trajectory enhancement submodule that uses a third-party route planning service to supplement trajectory information; The trajectory retrieval module for the adjacent transportation task includes: a planning trajectory submodule that searches for the OD experience trajectory with the target OD as the center and uses it as the main part of the planning trajectory; an experience route submodule that uses a third-party route planning service to solve the problem of missing head and tail near the starting and ending points of the trajectory, thereby improving the experience trajectory and forming a transportation route that conforms to historical business experience; and a route display submodule that displays the transportation route that conforms to historical experience.

Images