Riding-sharing boarding point recommendation method and system for inter-city online car hailing

Abstract

The invention provides a car-sharing boarding point recommendation method and system for inter-city online car hailing. The method comprises the following steps: acquiring inter-city traffic network data; projecting the three-dimensional road network to a two-dimensional topological plane; calculating the dynamic weight of each road based on the real-time traffic data; folding the secondary road according to the dynamic weight and the folding threshold to generate a backbone network; generating candidate recommendation results on the backbone network, wherein the candidate recommendation results comprise necessary points, selectable points and dynamic adjustment points; and carrying out optimization screening on the candidate recommendation results to obtain recommended boarding points. According to the invention, through dynamic weight calculation and backbone network generation, changes of traffic conditions can be reflected in real time, more accurate recommended boarding points can be generated, the ride-sharing efficiency can be improved, and through generation of diversified candidate recommendation points and optimization of candidate results, ride-sharing demands of different passengers can be satisfied, and the satisfaction degree of the passengers can be improved. The generation of the backbone network and the optimization of the candidate points effectively improve the calculation efficiency, so that the recommendation system can respond to the passenger request more quickly.

Description

Technical Field

[0001] The present invention relates to the technical field of online carpooling for intercity trips, and particularly to a method and system for recommending pick-up points for intercity online carpooling. Background Art

[0002] With the acceleration of the urbanization process, the demand for intercity trips is increasing day by day. As a convenient and flexible travel mode, online carpooling for intercity trips is favored by a large number of passengers. However, in the scenario of intercity online carpooling, the selection of pick-up points is crucial for carpooling efficiency and passenger satisfaction.

[0003] Traditional methods for recommending pick-up points often rely on static geographical information or simple rules, and it is difficult to reflect the changes in traffic conditions in real time, resulting in inaccurate recommended pick-up points and affecting carpooling efficiency and passenger experience. Summary of the Invention

[0004] Based on this, the object of the present invention is to propose a method and system for recommending pick-up points for intercity online carpooling to solve the above-mentioned problems.

[0005] According to a method for recommending pick-up points for intercity online carpooling proposed by the present invention, the method includes:

[0006] Obtain intercity traffic network data, including road nodes, edges, and weight information;

[0007] Based on the intercity traffic network data, project the three-dimensional road network onto a two-dimensional topological plane to establish a graph theory representation;

[0008] Based on real-time traffic data, calculate the dynamic weight of each road;

[0009] Fold secondary roads according to the dynamic weight and folding threshold to generate a backbone network;

[0010] Generate candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjustable points;

[0011] Optimize and screen the candidate recommendation results to obtain the final recommended pick-up points.

[0012] Furthermore, the step of obtaining intercity traffic network data includes:

[0013] Obtain GIS road data of the target area, including node coordinates, road lengths, and road grades;

[0014] Establish an initial traffic network graph G=(V, E), where V is the set of intersection nodes and E is the set of road edges;

[0015] Assign an initial weight w(e) to each edge e∈E 初始= Road class × Basic vehicle speed coefficient.

[0016] Furthermore, the steps of calculating the dynamic weight of each road include:

[0017] Access the real-time traffic data interface to obtain the current average vehicle speed v_real(e) of each road segment;

[0018] Calculate the dynamic weight, and the formula is:

[0019] w(e) = α·v_real(e)·δ(t) + β·Road class·θ(e) + γ·Accident probability coefficient·

[0020] where δ(t) is the time penalty coefficient, θ(e) is the road function coefficient, is the weather influence coefficient, and α, β, and γ are weight coefficients pre-calibrated according to regional characteristics.

[0021] Furthermore, the steps of folding secondary roads according to the dynamic weight and the folding threshold to generate the backbone network include:

[0022] Set the initial folding threshold;

[0023] Traverse all edges e ∈ E;

[0024] Dynamically calculate the folding threshold θ = Average dynamic weight × First adjustment coefficient;

[0025] If the dynamic weight w(e) < folding threshold θ, then perform the folding operation to merge the two end nodes v_i and v_j of edge e into a super node V_new;

[0026] Update the associated edge set to redirect the edges originally connected to nodes v_i and v_j to the super node V_new and recalculate the dynamic weights of the affected edges;

[0027] Repeat the iteration until the network scale converges.

[0028] Furthermore, the steps of merging the two end nodes v_i and v_j of edge e into a super node V_new include:

[0029] Calculate the latitude coordinate of the super node V_new, and the formula is:

[0030]

[0031] where φ_new is the latitude of the super node, φ_i is the latitude of node v_i, φ_j is the latitude of node v_j, f_i is the traffic flow weight of the road associated with node v_i, and f_j are the traffic flow weights of the roads associated with node v_j respectively;

[0032] Calculate the longitude coordinate of V_new of the super node, and the formula is:

[0033]

[0034] Among them, λ_new is the longitude of the super node, λ_i is the longitude of node v_i, and λ_j is the longitude of node v_j;

[0035] Inherit all the attributes of nodes v_i and v_j to V_new, including the connected road edges, traffic signal information, and historical data;

[0036] Remove the original edge e connecting nodes v_i and v_j from the network graph.

[0037] Furthermore, the steps of generating candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjustable points, include:

[0038] Select all nodes with degree ≥ all degree thresholds as core boarding points and define them as mandatory points;

[0039] For the shortest path between each pair of mandatory points, take the midpoint position as a candidate boarding point and define it as an optional point;

[0040] Generate temporary recommendation points within the super node neighborhood according to the real-time folding state as dynamically adjustable points.

[0041] Furthermore, the steps of generating temporary recommendation points within the super node neighborhood according to the real-time folding state as dynamically adjustable points include:

[0042] Taking the super node V_new as the center, establish an initial circular neighborhood, and combine with the Voronoi diagram constraint to exclude the service range of the adjacent super node V_new;

[0043] According to the current number of unmatched carpool requests and the number of available vehicles, dynamically adjust the neighborhood radius, and the formula is;

[0044]

[0045] Among them, k is the second adjustment coefficient, N_requests is the current number of unmatched carpool requests, and N_vehicles is the number of available vehicles;

[0046] Based on the neighborhood, generate temporary recommendation points with multiple objectives;

[0047] Comprehensively consider the traffic matching degree, demand density, and facility importance, and calculate the priority for each temporary recommendation point;

[0048] Dynamically adjust the candidate points.

[0049] Further, the step of generating temporary recommendation points based on the neighborhood and for multiple objectives includes:

[0050] Sample the flow velocity gradient along the boundary road of the neighborhood, and select the positions where the absolute value of the flow velocity gradient exceeds the gradient threshold as temporary recommendation points;

[0051] Apply a clustering algorithm to cluster the passenger request coordinates, and select the clustering core points as temporary recommendation points;

[0052] Identify POIs, and select the POIs closer to the super node V_new as temporary recommendation points.

[0053] Further, the step of optimizing the candidate recommendation results includes:

[0054] Establish a candidate boarding point set S = {mandatory points + optional points + dynamically adjusted points};

[0055] For each carpooling request, calculate the walking distance cost C_walk, vehicle detour cost C_detour, and passenger matching degree P_match to each point in the candidate boarding point set S;

[0056] Select the point with the minimum comprehensive cost C as the recommended boarding point, and the formula is:

[0057] C = ω1C_walk + ω2C_detour - ω3P_match, where ω1, ω2, and ω3 are the importance scores of the walking distance cost, vehicle detour cost, and passenger matching probability respectively.

[0058] The present invention also proposes a carpooling boarding point recommendation system for intercity online car-hailing, which is used to implement the above-mentioned carpooling boarding point recommendation method for intercity online car-hailing. The system includes:

[0059] Data acquisition module: used to acquire intercity traffic network data, including road nodes, edges, and weight information;

[0060] Two-dimensional conversion module: used to project the three-dimensional road network onto a two-dimensional topological plane based on the intercity traffic network data to establish a graph theory representation;

[0061] Dynamic weight module: used to calculate the dynamic weight of each road based on real-time traffic data;

[0062] Folding module: used to fold secondary roads according to the dynamic weight and folding threshold to generate a backbone network;

[0063] Candidate generation module: used to generate candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjusted points;

[0064] Recommendation point generation module: used to optimize and screen the candidate recommendation results to obtain the final recommended pick-up point.

[0065] In summary, for the carpool pick-up point recommendation method for intercity online car-hailing of the present invention, intercity traffic network data is obtained, and the complex intercity traffic network data (including road nodes, edges and their weight information in three-dimensional geographical coordinates) is integrated and projected onto a two-dimensional topological plane to establish a graph theory representation, so as to clearly represent the structure and connection relationship of the road network, providing a solid foundation for subsequent dynamic weight calculation, network folding and recommendation point generation; the dynamic weight of each road is calculated based on real-time traffic data, and the calculation of the dynamic weight can reflect the changes in traffic conditions in real time, improving the timeliness and accuracy of the recommended pick-up point; secondary roads are folded according to the dynamic weight and folding threshold to generate a backbone network, simplifying the network structure. The generation of the backbone network not only improves the calculation efficiency, but also makes the selection of the recommended pick-up point more focused on key positions, improving the accuracy of the recommendation; candidate recommendation results are generated on the backbone network, including mandatory points, optional points and dynamically adjustable points, so as to comprehensively consider the network structure, traffic conditions and passenger needs to generate diverse recommended points. The diverse recommended points can meet the carpooling needs of different passengers, improving the flexibility and convenience of carpooling. At the same time, the introduction of dynamically adjustable points enables the recommendation results to adapt to changes in traffic conditions in real time; the candidate recommendation results are optimized and screened to obtain the final recommended pick-up point. This step can comprehensively evaluate and screen the candidate points by considering factors such as walking distance cost, vehicle detour cost and passenger matching probability. The finally obtained recommended pick-up point can not only consider the convenience of passengers, but also take into account the driving efficiency of the vehicle and the passenger matching probability, thereby improving the overall efficiency of carpooling and passenger satisfaction.

[0066] Through the calculation of dynamic weights and the generation of backbone networks, the present invention can reflect the changes in traffic conditions in real time, generate more accurate recommended pick-up points, improve the efficiency of carpooling, and meet the carpooling needs of different passengers by generating diverse candidate recommendation points and optimizing the candidate recommendation results, improving passenger satisfaction. In addition, the generation of the backbone network and the optimization of candidate points effectively improve the calculation efficiency, enabling the recommendation system to respond to passenger requests more quickly.

[0067] The additional aspects and advantages of the present invention will be partly given in the following description, partly will become obvious from the following description, or will be understood through the embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, wherein:

[0069] Figure 1Flow chart of a method for recommending carpool pick-up points for intercity online car-hailing according to Embodiment 1 of the present invention;

[0070] Figure 2 System block diagram of a system for recommending carpool pick-up points for intercity online car-hailing according to Embodiment 2 of the present invention. Detailed implementation manners

[0071] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0072] It should be noted that when an element is referred to as being "fixedly provided on" another element, it can be directly on the other element or there can also be an intermediate element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intermediate element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are only for the purpose of illustration.

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0074] Embodiment 1

[0075] Please refer to Figure 1 , the present invention proposes a method for recommending carpool pick-up points for intercity online car-hailing, and the method includes steps S101 to S106:

[0076] S101, obtaining intercity traffic network data, including road nodes, edges and weight information.

[0077] It should be noted that obtaining the core data of the intercity traffic network, including road nodes, edges and weight information, etc., these data are the key data basis for graph theory modeling, dynamic weight calculation, backbone network generation and candidate recommendation result optimization in the subsequent steps.

[0078] The node data includes basic attributes (such as longitude and latitude coordinates), topological features (including node connectivity, the number of adjacent edges, etc., the node connectivity of intersections > the node connectivity of ordinary nodes) and extended attributes (whether it includes transportation hubs, such as high-speed railway stations, airports, whether it has multimodal transfer functions, etc.).

[0079] The edge data includes physical attributes (such as road length, number of lanes, type of physical isolation, etc.), a hierarchical system (which can be classified according to the Highway Engineering Technical Standards, expressway > first-class highway > second-class highway > third-class highway > fourth-class highway), and traffic parameters (such as design speed, speed limit value, historical average daily traffic volume, etc.).

[0080] The weight information includes an initial weight, which is equal to the road grade × the basic speed coefficient. A dynamic weight interface is also set in the weight information layer, including access ports for dynamic parameters such as real-time speed, accident probability, and weather impact. The weight is refreshed every preset time, such as once every 5 minutes.

[0081] Further optionally, the steps of obtaining the intercity traffic network data include:

[0082] Obtain the GIS road data of the target area, including node coordinates, road length, and road grade;

[0083] Establish an initial traffic network graph G=(V, E), where V is the set of intersection nodes and E is the set of road edges;

[0084] Assign an initial weight w(e) to each edge e∈E 初始 = road grade × basic speed coefficient.

[0085] It can be understood that by establishing a graph theory model G=(V, E) with GIS road data, the physical road system is transformed into a computable mathematical structure. The node coordinate accuracy needs to reach the sub-meter level (typical value: ±0.5m) to ensure the accuracy of the network topology. The road grade classification needs to conform to the Highway Engineering Technical Standards (such as expressway > first-class highway > second-class highway).

[0086] Establish a static attribute benchmark, and assign an initial weight to each edge e∈E. The formula is: w(e) 初始 = road grade × basic speed coefficient, realizing the coupling characterization of two factors. The basic speed coefficient needs to be obtained through traffic surveys. For example, 85% of the design speed is taken as the benchmark for expressways.

[0087] Convert the road network into a directed or undirected graph G=(V, E) to form a static network map, providing a benchmark reference for subsequent dynamic adjustments.

[0088] S102. Based on the intercity traffic network data, project the three-dimensional road network onto a two-dimensional topological plane to establish a graph theory representation.

[0089] It should be noted that integrating and projecting complex intercity traffic network data (including road nodes, edges, and their weight information in three-dimensional geographical coordinates) onto a two-dimensional topological plane to establish a graph theory representation clearly shows the structure and connection relationships of the road network, providing a solid foundation for subsequent dynamic weight calculation, network folding, and recommended point generation.

[0090] Specifically, to convert three-dimensional geographical coordinates (longitude, latitude, altitude) into two-dimensional plane coordinates, an appropriate projection method can be selected, such as Mercator projection or UTM projection, to minimize the projection error.

[0091] Remove redundant nodes and edges in the two-dimensional network, and only retain key road nodes and connection relationships.

[0092] Merge adjacent nodes to simplify the network structure and improve calculation efficiency.

[0093] Represent the simplified two-dimensional network as nodes and edges in graph theory.

[0094] Assign initial weights to each node and edge, considering factors such as road grade and length.

[0095] Determine the connection relationships between nodes and establish an adjacency matrix or adjacency list.

[0096] Calculate the distances between nodes and the weights of edges, providing a basis for subsequent dynamic weight calculation and network folding.

[0097] S103. Based on real-time traffic data, calculate the dynamic weight of each road.

[0098] It should be noted that by considering real-time traffic data (such as vehicle speed, accident probability, etc.) and road static attributes (such as road grade), the dynamic weight of each road is calculated. The dynamic weight calculation can reflect the changes in traffic conditions in real time, improving the timeliness and accuracy of recommended boarding points.

[0099] Further optionally, the steps for calculating the dynamic weight of each road include:

[0100] Access the real-time traffic data interface to obtain the current average vehicle speed v_real(e) of each road segment.

[0101] Calculate the dynamic weight, and the formula is:

[0102] w(e) = α·v_real(e)·δ(t) + β·road grade·θ(e) + γ·accident probability coefficient·

[0103] where δ(t) is the time penalty coefficient, θ(e) is the road function coefficient, is the weather influence coefficient.

[0104] It is understandable that when calculating the dynamic weight formula, the second-level data update is implemented through the v_real(e) interface. For example, the update frequency is 10 - 30 seconds. The time penalty coefficient δ(t) needs to be combined with traffic flow theory. For example, the BPR function is adopted: δ(t) = 1 + 0.15 * (V / C). 4 , the road function coefficient θ(e) needs to reflect the road planning priority. For example, for the main road, θ = 1.5, and for the branch road, θ = 0.6. The weather influence coefficient needs to be docked with meteorological data, such as rain and snow weather

[0105] The α parameter can associate the real-time vehicle speed v_real(e) with the road grade, realizing the non-linear mapping of vehicle speed - weight, that is, the fusion of dynamic attributes and static attributes. When α = 1, the real-time vehicle speed directly affects the weight in proportion; when α > 1, the sensitivity of the vehicle speed is enhanced.

[0106] The road grade is non-linearly amplified through the θ(e) function (such as θ = 1.5 for the main road and θ = 0.6 for the branch road). The β parameter plays a role in strengthening the road grade, maintaining the basic role of the road grade, and at the same time enhancing the advantages of high-grade roads in dynamic calculations.

[0107] The γ parameter, as a dynamic adjustment factor, reflects the inhibitory effect of accident risk on route selection and can couple the accident probability with the weather influence and automatically amplify the influence of the accident probability in rain and snow weather . It is adjusted according to road safety ratings, real-time traffic events, and insurance actuarial data. For example, the γ value needs to be increased for accident black spot sections, the γ value is dynamically increased during sudden accidents, and the γ value needs to be increased for high-risk sections.

[0108] The calculated dynamic weight integrates three dimensions: physical attributes (i.e., road grade), real-time state (i.e., vehicle speed), and environmental variables (i.e., weather). By calculating the dynamic weight, a two-layer network model of static benchmark and dynamic perturbation is formed, which can improve the fidelity of the representation, enabling the system to not only maintain the inherent attributes of the road network but also respond to the complex changes of the traffic system in real time, providing an accurate dynamic network model for subsequent network folding and boarding point recommendation.

[0109] S104, fold the secondary roads according to the dynamic weight and the folding threshold to generate a backbone network.

[0110] It should be noted that folding the secondary roads according to the dynamic weight and the folding threshold to generate a backbone network is to remove redundant information in the network, retain key road nodes and connection relationships, and simplify the network structure. The generation of the backbone network not only improves the calculation efficiency but also makes the selection of recommended boarding points more focused on key positions, improving the accuracy of the recommendation.

[0111] Further optionally, the step of folding secondary roads according to dynamic weights and folding thresholds to generate a backbone network includes:

[0112] Set an initial folding threshold;

[0113] Traverse all edges e ∈ E;

[0114] Dynamically calculate the folding threshold θ = average dynamic weight × first adjustment coefficient;

[0115] If the dynamic weight w(e) < folding threshold θ, perform a folding operation to merge the two nodes v_i and v_j at both ends of edge e into a supernode V_new;

[0116] Update the associated edge set to redirect the edges originally connected to nodes v_i and v_j to the supernode V_new and recalculate the dynamic weights of the affected edges;

[0117] Repeat the iteration until the network scale converges.

[0118] It can be understood that first, an initial folding threshold is set as a benchmark for judging the importance of roads; all edges e ∈ E in the network are traversed, and the dynamic weight w(e) of each edge is calculated, which is based on three-dimensional dynamic factors of physical attributes (i.e., road grade), real-time status (i.e., vehicle speed), and environmental variables (i.e., weather); and the folding threshold θ is dynamically calculated based on the average dynamic weight and the first adjustment coefficient, and this threshold is dynamically adjusted according to the network state to ensure the adaptability of the folding process; if the dynamic weight w(e) of a certain edge e < folding threshold θ, then perform a folding operation: merge the two nodes v_i and v_j at both ends of edge e into a supernode V_new. The folding operation essentially deletes secondary roads from the network and simplifies the network structure by merging nodes; then update the associated edge set to redirect all the edges originally connected to v_i and v_j to the supernode V_new to ensure that the connectivity and topological relationship of the network still remain after folding, and recalculate the dynamic weights of the affected edges; repeat the above iterative process until the network scale converges, that is, the number of nodes is stable, indicating that there are no more secondary roads that can be folded and the backbone network generation is completed. The present invention performs folding based on dynamic weights and thresholds, can respond to changes in the traffic network in real time, and ensures the timeliness and accuracy of the backbone network. By merging secondary roads and retaining key nodes and edges, a concise backbone network is generated, which is convenient for the efficient implementation of applications such as traffic planning and path navigation. And the iterative folding process will gradually simplify the network, reduce the computational complexity, and improve the algorithm operation efficiency.

[0119] Further optionally, the step of merging the two nodes v_i and v_j at both ends of edge e into a supernode V_new includes:

[0120] Calculate the latitude coordinate of V_new of the supernode, and the formula is:

[0121]

[0122] Where φ_new is the latitude of the supernode, φ_i is the latitude of node v_i, φ_j is the latitude of node v_j, f_i is the traffic flow weight of the road associated with node v_i, and f_j is the traffic flow weight of the road associated with node v_j respectively;

[0123] Calculate the longitude coordinate of V_new of the supernode, and the formula is:

[0124]

[0125] Where λ_new is the longitude of the supernode, λ_i is the longitude of node v_i, and λ_j is the longitude of node v_j;

[0126] Inherit all the attributes of nodes v_i and v_j to V_new, including the connected road edges, traffic signal information, and historical data;

[0127] Remove the original edge e connecting nodes v_i and v_j from the network graph.

[0128] It can be understood that according to the latitudes φ_i and φ_j of nodes v_i and v_j, and the traffic flow weights f_i and f_j of the roads they are associated with, calculate the latitude of the supernode V_new to ensure that the position of the supernode is closer to the nodes with larger traffic flow, reflecting the importance of the nodes in the traffic network. Similarly, according to the longitudes λ_i and λ_j of nodes v_i and v_j, calculate the longitude λ_new of the supernode V_new.

[0129] Inherit all the attributes of nodes v_i and v_j to the supernode V_new, including the connected road edges, traffic signal information, historical data, etc. Specifically, retain the higher-level traffic signal configuration in the original nodes to ensure that the merged network still meets the traffic management requirements. Establish a two-way connection mapping table from the supernode V_new to all the original associated nodes to maintain the connectivity and topological relationship of the network. Merge the historical congestion records of nodes v_i and v_j to establish a time series feature vector. And remove the original edge e connecting nodes v_i and v_j from the network graph.

[0130] S105, generate candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjustable points.

[0131] It should be noted that candidate recommended results are generated on the backbone network, including mandatory points, optional points, and dynamically adjustable points, so as to comprehensively consider the network structure, traffic conditions, and passenger demands, generate diverse recommended points, and the diverse recommended points can meet the carpooling demands of different passengers, improving the flexibility and convenience of carpooling. At the same time, the introduction of dynamically adjustable points enables the recommended results to adapt to changes in traffic conditions in real time.

[0132] Further optionally, the step of generating candidate recommended results on the backbone network, including mandatory points, optional points, and dynamically adjustable points, includes:

[0133] Select all nodes with degree ≥ all degree thresholds as core boarding points and define them as mandatory points;

[0134] For the shortest path between each pair of mandatory points, take the midpoint position as a candidate boarding point and define it as an optional point;

[0135] Generate temporary recommended points within the supernode neighborhood according to the real-time folding state as dynamically adjustable points.

[0136] It is understandable that all nodes with degree ≥ all degree thresholds (such as 3) are selected as core boarding points and defined as mandatory points. The node degree reflects the local attribute of the node in the road network, that is, the number of edges directly connected to the node. Nodes with higher degrees have higher traffic flow and stronger connectivity, which are key positions for passengers to get on and off or for vehicles to stop.

[0137] For the shortest path between each pair of mandatory points, take the midpoint position as a candidate boarding point and define it as an optional point. Taking the midpoint position of the shortest path as a candidate boarding point can make the passenger getting on and off positions more reasonable, reduce the walking distance or waiting time of passengers, and improve the convenience and comfort of carpooling. At the same time, the midpoint position is also convenient for vehicle parking and passenger gathering and dispersal.

[0138] Generate temporary recommended points within the supernode neighborhood according to the real-time folding state as dynamically adjustable points. The real-time folding state reflects the real-time changes in traffic conditions, and such changes may make the original recommended points no longer suitable for the current traffic demands. And the supernode is a key node in the traffic network with higher traffic flow and connectivity. Establishing a neighborhood centered on the supernode can cover a wider range of passenger demands. The temporary recommended points generated within the supernode neighborhood can be dynamically adjusted according to real-time traffic conditions (such as passenger request coordinates, traffic flow, etc.), making the recommended points more in line with the passenger getting on and off demands, thus improving the success rate of carpooling. And the temporary recommended points as dynamically adjustable points can be adjusted in real time according to the current carpooling request quantity and resource status (such as the number of available vehicles). This helps to optimize resource allocation and avoid situations of resource waste (such as empty vehicle running) or resource shortage (such as passengers waiting for too long).

[0139] Further optionally, the step of generating temporary recommendation points within the supernode neighborhood according to the real-time folding state as dynamic adjustment points includes:

[0140] Taking the supernode V_new as the center, an initial circular neighborhood is established, and in combination with the Voronoi diagram constraint, the service scope of the adjacent supernode V_new is excluded;

[0141] According to the current number of unmatched carpool requests and the number of available vehicles, the neighborhood radius is dynamically adjusted, and the formula is;

[0142]

[0143] where k is the second adjustment coefficient, N_requests is the current number of unmatched carpool requests, and N_vehicles is the number of available vehicles;

[0144] Based on the neighborhood, temporary recommendation points are generated with multiple objectives;

[0145] Combining the traffic flow matching degree, demand density, and facility importance, calculate the priority for each temporary recommendation point;

[0146] Dynamically adjust the candidate points.

[0147] It is understandable that taking the supernode V_new as the center, an initial circular neighborhood is established, and the size of the initial circular neighborhood can be set according to the actual situation to cover a certain range around the supernode.

[0148] In combination with the Voronoi diagram constraint, the service scope of the adjacent supernode V_new is excluded. The Voronoi diagram can be used to determine the influence range of traffic hubs, thereby optimizing the distribution of traffic flow. Through the Voronoi diagram constraint, it is possible to avoid the over-concentration or overlap of recommendation points, ensuring that each recommendation point can cover effective passenger demands.

[0149] According to the current number of unmatched carpool requests and the number of available vehicles, the neighborhood radius is dynamically adjusted. Specifically, by considering the ratio of the number of unmatched carpool requests to the number of available vehicles, the neighborhood radius can be adjusted in real time to adapt to different traffic demands. When the number of unmatched carpool requests is large and the number of available vehicles is small, appropriately expanding the neighborhood radius can cover more potential passengers and improve the carpool success rate. On the contrary, when the number of unmatched carpool requests is small and the number of available vehicles is large, the neighborhood radius can be appropriately reduced to avoid the over-dispersion or redundancy of recommendation points.

[0150] When generating temporary recommendation points, multiple objectives can be comprehensively considered, such as traffic flow, passenger demand, facility importance, etc., so as to more comprehensively reflect the actual situation of the transportation network and improve the accuracy and effectiveness of the recommendation points. Generating recommendation points within a dynamically adjusted neighborhood can ensure that the recommendation points are within the current reasonable range and can better adapt to the current traffic demand. Then, the priority can be calculated for each temporary recommendation point by comprehensively considering the flow matching degree, demand density, and facility importance.

[0151] Based on the adjustment mechanism, the candidate points are dynamically adjusted. Specifically, the candidate points are recalculated regularly, and the points with higher matching success rates among the historical candidate points are retained to adapt to the changes in traffic demand and ensure the timeliness and effectiveness of the recommendation results. When road closures, severe congestion, or special weather conditions are detected, local neighborhood reconstruction is immediately performed to regenerate candidate points that adapt to the current traffic conditions and improve the carpooling success rate. For candidate points that have not been selected for a long time or have a low matching success rate, they are demoted or removed. These candidate points may no longer be suitable as recommendation points due to changes in traffic demand, facility adjustments, etc. By demoting or removing them, the set of candidate points can be optimized to improve the accuracy and effectiveness of the recommendation results.

[0152] Further optionally, the step of generating temporary recommendation points based on the neighborhood and multiple objectives includes:

[0153] Sample the flow velocity gradient along the neighborhood boundary road, and select the positions where the absolute value of the flow velocity gradient exceeds the gradient threshold as temporary recommendation points;

[0154] Apply a clustering algorithm to cluster the passenger request coordinates, and select the clustering core points as temporary recommendation points;

[0155] Identify POIs, and select the POIs closer to the super node V_new as temporary recommendation points.

[0156] It is understandable that multiple objectives, such as traffic flow, passenger demand, facility importance, etc., can be comprehensively considered to generate temporary recommendation points, so as to more comprehensively reflect the actual situation of the transportation network and improve the accuracy and effectiveness of the recommendation points.

[0157] In specific implementation, the flow velocity gradient can be sampled along the neighborhood boundary road, and the positions where the absolute value of the flow velocity gradient exceeds the gradient threshold (such as set to 0.2 times the average flow velocity) are selected as temporary recommendation points. The flow velocity gradient reflects the change rate of traffic flow, and the positions where the absolute value exceeds the threshold represent areas with large traffic flow changes. These areas are usually hot spots for passengers to get on and off or for vehicles to park.

[0158] Apply a clustering algorithm to cluster passenger request coordinates and select the clustering core points as temporary recommended points. In the transportation network, the clustering algorithm can be used to identify dense areas of passenger request coordinates where the passenger demand is high. Using these areas as recommended points can increase the success rate of carpooling.

[0159] Identify POIs (Points of Interest) such as bus stops and parking lot entrances, and select the POIs closer to the super node V_new as temporary recommended points. These facilities are usually important locations for passengers to get on and off the vehicle. Using them as recommended points can facilitate passengers getting on and off the vehicle, improving the convenience and comfort of carpooling.

[0160] S106. Optimize and screen the candidate recommended results to obtain the final recommended boarding point.

[0161] It should be noted that the candidate recommended results are optimized and screened to obtain the final recommended boarding point. By considering factors such as walking distance cost, vehicle detour cost, and passenger matching probability, the candidate points can be comprehensively evaluated, screened, and optimized. The optimized recommended boarding point not only considers the convenience of passengers but also takes into account the driving efficiency of the vehicle and the passenger matching probability, thereby improving the overall efficiency of carpooling and passenger satisfaction.

[0162] Further optionally, the steps for optimizing the candidate recommended results include:

[0163] Establish a candidate boarding point set S = {mandatory points + optional points + dynamically adjusted points};

[0164] For each carpooling request, calculate the walking distance cost C_walk, vehicle detour cost C_detour, and passenger matching degree P_match to each point in the candidate boarding point set S;

[0165] Select the point with the minimum comprehensive cost C as the recommended boarding point. The formula is:

[0166] C = ω1C_walk + ω2C_detour - ω3P_match, where ω1, ω2, and ω3 are the importance scores of the walking distance cost, vehicle detour cost, and passenger matching probability respectively.

[0167] It is understandable that by integrating mandatory points, optional points, and dynamically adjusted points to establish a candidate boarding point set, a comprehensive and flexible candidate boarding point set is established by comprehensively considering mandatory points, optional points, and dynamically adjusted points to cover a wider range of passenger needs and increase the success rate of carpooling.

[0168] Then, comprehensively considering the walking distance cost \(C_{walk}\) (i.e., the walking distance of the passenger who issues the carpooling request from the current location to the candidate point), the vehicle detour cost \(C_{detour}\) (i.e., the detour distance of the vehicle from the current location to the candidate point), and the passenger matching degree \(P_{match}\) (i.e., the distance between the destination of the passenger who issues the carpooling request and the destination of the passengers waiting for carpooling in the vehicle. The closer the actual distance is, the higher the passenger matching degree), calculate the comprehensive cost \(C\) of each candidate point in the candidate pick-up point set \(S\), and select the point with the minimum comprehensive cost \(C\) as the final recommended pick-up point, so as to screen out the optimal pick-up point and improve the carpooling success rate.

[0169] In summary, the carpooling pick-up point recommendation method for intercity online car-hailing of the present invention obtains intercity traffic network data, integrates and projects complex intercity traffic network data (including road nodes, edges and their weight information in three-dimensional geographical coordinates) onto a two-dimensional topological plane, and establishes a graph theory representation to clearly represent the structure and connection relationship of the road network, providing a solid foundation for subsequent dynamic weight calculation, network folding and recommended point generation; calculates the dynamic weight of each road based on real-time traffic data, and the calculation of dynamic weight can reflect the changes in traffic conditions in real time, improving the timeliness and accuracy of the recommended pick-up point; folds secondary roads according to the dynamic weight and folding threshold to generate a backbone network and simplify the network structure. The generation of the backbone network not only improves the calculation efficiency, but also makes the selection of the recommended pick-up point more focused on key positions, improving the accuracy of the recommendation; generates candidate recommendation results on the backbone network, including mandatory points, optional points and dynamically adjustable points, so as to comprehensively consider the network structure, traffic conditions and passenger needs, and generate diverse recommended points. The diverse recommended points can meet the carpooling needs of different passengers, improving the flexibility and convenience of carpooling. At the same time, the introduction of dynamically adjustable points enables the recommendation result to adapt to the changes in traffic conditions in real time; optimizes and screens the candidate recommendation results to obtain the final recommended pick-up point. This step can comprehensively evaluate and screen the candidate points by considering factors such as walking distance cost, vehicle detour cost and passenger matching probability. The finally obtained recommended pick-up point can not only consider the convenience of passengers, but also take into account the driving efficiency of the vehicle and the passenger matching probability, thereby improving the overall efficiency of carpooling and passenger satisfaction.

[0170] Through the calculation of dynamic weights and the generation of the backbone network, the present invention can reflect the changes in traffic conditions in real time, generate more accurate recommended pick-up points, improve the efficiency of carpooling, and meet the carpooling needs of different passengers by generating diverse candidate recommendation points and optimizing the candidate recommendation results, improving passenger satisfaction. In addition, the generation of the backbone network and the optimization of candidate points effectively improve the calculation efficiency, enabling the recommendation system to respond to passenger requests more quickly.

[0171] Embodiment 2

[0172] Please refer to Figure 2 , a ridesharing pick-up point recommendation system for intercity online car-hailing proposed by the present invention, the system comprising:

[0173] A data acquisition module: used to acquire intercity traffic network data, including road nodes, edges, and weight information;

[0174] A two-dimensional conversion module: used to project a three-dimensional road network onto a two-dimensional topological plane based on intercity traffic network data to establish a graph theory representation;

[0175] A dynamic weight module: used to calculate the dynamic weight of each road based on real-time traffic data;

[0176] A folding module: used to fold secondary roads according to the dynamic weight and a folding threshold to generate a backbone network;

[0177] A candidate generation module: used to generate candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjustable points;

[0178] A recommended point generation module: used to optimize and screen the candidate recommendation results to obtain the final recommended pick-up point.

[0179] Further optionally, the data acquisition module is further used for:

[0180] Acquiring GIS road data of the target area, including node coordinates, road lengths, and road grades;

[0181] Establishing an initial traffic network graph G = (V, E), where V is a set of intersection nodes and E is a set of road edges;

[0182] Assigning an initial weight w(e) to each edge e ∈ E 初始 = road grade × basic vehicle speed coefficient.

[0183] Further optionally, the dynamic weight module is further used for:

[0184] Accessing a real-time traffic data interface to obtain the current average vehicle speed v_real(e) of each road section;

[0185] Calculating the dynamic weight, the formula being:

[0186] w(e) = α·v_real(e)·δ(t) + β·road grade·θ(e) + γ·accident probability coefficient·

[0187] where δ(t) is a time penalty coefficient, θ(e) is a road function coefficient, is a weather influence coefficient, and α, β, and γ are weight coefficients pre-calibrated according to regional characteristics.

[0188] Further optionally, the folding module is further configured to:

[0189] Set an initial folding threshold;

[0190] Traverse all edges e ∈ E;

[0191] Dynamically calculate the folding threshold θ = average dynamic weight × first adjustment coefficient;

[0192] If the dynamic weight w(e) < folding threshold θ, perform a folding operation to merge the two nodes v_i and v_j at both ends of edge e into a supernode V_new;

[0193] Update the associated edge set to redirect the edges originally connected to nodes v_i and v_j to the supernode V_new and recalculate the dynamic weights of the affected edges;

[0194] Repeat the iteration until the network scale converges.

[0195] Further optionally, the folding module is further configured to:

[0196] Calculate the latitude coordinates of the supernode V_new, and the formula is:

[0197]

[0198] where φ_new is the latitude of the supernode, φ_i is the latitude of node v_i, φ_j is the latitude of node v_j, f_i is the traffic flow weight of the road associated with node v_i, and f_j is the traffic flow weight of the road associated with node v_j respectively;

[0199] Calculate the longitude coordinates of the supernode V_new, and the formula is:

[0200]

[0201] where λ_new is the longitude of the supernode, λ_i is the longitude of node v_i, and λ_j is the longitude of node v_j;

[0202] Inherit all attributes of nodes v_i and v_j to V_new, including the connected road edges, traffic signal information, and historical data;

[0203] Remove the edge e that originally connected nodes v_i and v_j from the network graph.

[0204] Further optionally, the candidate generation module is further configured to:

[0205] Select all nodes with degrees ≥ all degree thresholds as core boarding points and designate them as mandatory points;

[0206] For the shortest path between each pair of mandatory points, take the midpoint as the candidate boarding point and define it as an optional point.

[0207] Generate temporary recommended points within the neighborhood of the supernode according to the real-time folding state as dynamically adjustable points.

[0208] Optionally further, the candidate generation module is further configured to:

[0209] Centered on the supernode V_new, establish an initial circular neighborhood, and combine with the Voronoi diagram constraint to exclude the service scope of the adjacent supernode V_new.

[0210] Dynamically adjust the neighborhood radius according to the current number of unmatched carpool requests and the number of available vehicles. The formula is;

[0211]

[0212] where k is the second adjustment coefficient, N_requests is the current number of unmatched carpool requests, and N_vehicles is the number of available vehicles.

[0213] Based on the neighborhood, generate temporary recommended points with multiple objectives.

[0214] Integrate the traffic matching degree, demand density, and facility importance to calculate the priority for each temporary recommended point.

[0215] Dynamically adjust the candidate points.

[0216] Optionally further, the candidate generation module is further configured to:

[0217] Sample the flow velocity gradient along the neighborhood boundary road, and select the positions where the absolute value of the flow velocity gradient exceeds the gradient threshold as temporary recommended points.

[0218] Apply a clustering algorithm to cluster the passenger request coordinates, and select the clustering core points as temporary recommended points.

[0219] Identify POIs, and select the POIs closer to the supernode V_new as temporary recommended points.

[0220] Optionally further, the recommended point generation module:

[0221] Establish a set S of candidate boarding points = {mandatory points + optional points + dynamically adjustable points};

[0222] For each carpool request, calculate the walking distance cost C_walk, vehicle detour cost C_detour, and passenger matching degree P_match to each point in the set S of candidate boarding points.

[0223] Select the point with the minimum comprehensive cost C as the recommended boarding point. The formula is:

[0224] C = ω1C_walk + ω2C_detour - ω3P_match, where ω1, ω2, and ω3 are the importance scores of the walking distance cost, the vehicle detour cost, and the passenger matching probability, respectively.

[0225] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. A method for recommending carpool pick-up points for intercity online car-hailing, characterized in that, The method includes: Obtain intercity traffic network data, including road nodes, edges, and weight information; Based on the intercity traffic network data, project the three-dimensional road network onto a two-dimensional topological plane to establish a graph theory representation; Based on real-time traffic data, calculate the dynamic weight of each road; Fold secondary roads according to the dynamic weight and folding threshold to generate a backbone network; Generate candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjustable points; Optimize and screen the candidate recommendation results to obtain the final recommended boarding points.

2. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 1, wherein The steps of obtaining intercity traffic network data include: Obtain GIS road data of the target area, including node coordinates, road lengths, and road grades; Establish an initial traffic network graph G=(V, E), where V is the set of intersection nodes and E is the set of road edges; Assign an initial weight w(e) to each edge e ∈ E 初始 = road grade × basic vehicle speed coefficient 3. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 1, wherein, The steps of calculating the dynamic weight of each road include: Access the real-time traffic data interface to obtain the current average vehicle speed v_real(e) of each road section; Calculate the dynamic weight, and the formula is: w(e) = α·v_real(e)·δ(t) + β·road grade·θ(e) + γ·accident probability coefficient Among them, δ(t) is the time penalty coefficient, θ(e) is the road function coefficient, is the weather influence coefficient, and α, β, γ are weight coefficients pre-calibrated according to regional characteristics.

4. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 3, wherein The steps of folding secondary roads according to the dynamic weight and folding threshold to generate a backbone network include: Set an initial folding threshold; Traverse all edges e∈E; Dynamically calculate the folding threshold θ = average dynamic weight × first adjustment coefficient; If the dynamic weight w(e) < folding threshold θ, perform a folding operation to merge the two nodes v_i and v_j at both ends of edge e into a supernode V_new; Update the associated edge set to redirect the edges originally connected to nodes v_i and v_j to the supernode V_new and recalculate the dynamic weights of the affected edges; Repeat the iteration until the network scale converges.

5. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 4, wherein The steps of merging the two nodes v_i and v_j at both ends of edge e into a supernode V_new include: Calculate the latitude coordinate of the supernode V_new, and the formula is: where φ_new is the latitude of the supernode, φ_i is the latitude of node v_i, φ_j is the latitude of node v_j, f_i is the traffic flow weight of the road associated with node v_i, and f_j are the traffic flow weights of the roads associated with node v_j respectively; Calculate the longitude coordinate of the supernode V_new, and the formula is: where λ_new is the longitude of the supernode, λ_i is the longitude of node v_i, and λ_j is the longitude of node v_j; Inherit all attributes of nodes v_i and v_j to V_new, including connected road edges, traffic signal information, and historical data; Remove the original edge e connecting nodes v_i and v_j from the network graph.

6. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 4, wherein The steps of generating candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjustable points, include: Select all nodes with degrees ≥ all degree thresholds as core boarding points and define them as mandatory points; For the shortest path between each pair of mandatory points, take the midpoint position as a candidate boarding point and define it as an optional point; Generate temporary recommendation points within the supernode neighborhood according to the real-time folding status as dynamically adjustable points.

7. The method for recommending a carpool pick-up point for intercity online car-hailing according to claim 6, wherein The steps of generating temporary recommendation points within the supernode neighborhood according to the real-time folding status as dynamically adjustable points include: Centered on the supernode V_new, establish an initial circular neighborhood, and combine with the Voronoi diagram constraint to exclude the service range of the adjacent supernode V_new; Dynamically adjust the neighborhood radius according to the current number of unmatched carpool requests and the number of available vehicles. The formula is: where k is the second adjustment coefficient, N_requests is the current number of unmatched carpool requests, and N_vehicles is the number of available vehicles; Based on the neighborhood, generate temporary recommendation points with multiple objectives; Calculate the priority for each temporary recommendation point by integrating traffic matching degree, demand density, and facility importance; Dynamically adjust the candidate points.

8. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 7, wherein, The step of generating temporary recommendation points with multiple objectives based on the neighborhood includes: Sample the flow velocity gradient along the neighborhood boundary road, and select the positions where the absolute value of the flow velocity gradient exceeds the gradient threshold as temporary recommendation points; Apply a clustering algorithm to cluster the passenger request coordinates, and select the cluster core points as temporary recommendation points; Identify POIs and select the POIs closer to the super node V_new as temporary recommendation points.

9. The ride-sharing pick-up point recommendation method for intercity online car-hailing according to claim 1, characterized in that, The step of optimizing the candidate recommendation results includes: Establish a candidate boarding point set S = {mandatory points + optional points + dynamically adjusted points}; For each carpool request, calculate the walking distance cost C_walk, vehicle detour cost C_detour, and passenger matching degree P_match to each point in the candidate boarding point set S; Select the point with the minimum comprehensive cost C as the recommended boarding point. The formula is: C = ω1C_walk + ω2C_detour - ω3P_match, where ω1, ω2, and ω3 are the importance scores of the walking distance cost, vehicle detour cost, and passenger matching probability respectively.

10. A ride-sharing pick-up point recommendation system for intercity online car-hailing, which is used to implement the ride-sharing pick-up point recommendation method for intercity online car-hailing described in any one of claims 1 to 9, characterized in that, The system includes: Data acquisition module: used to acquire intercity traffic network data, including road nodes, edges, and weight information; Two-dimensional conversion module: used to project the three-dimensional road network onto a two-dimensional topological plane based on the intercity traffic network data and establish a graph theory representation; Dynamic weight module: used to calculate the dynamic weight of each road based on real-time traffic data; Folding module: used to fold secondary roads according to the dynamic weight and folding threshold to generate a backbone network; Candidate generation module: used to generate candidate recommendation results on the backbone network, including mandatory points, optional points, and dynamically adjusted points; Recommendation point generation module: used to optimize and screen the candidate recommendation results to obtain the final recommended boarding point.

Images