Online car-hailing area minimum motorcade putting method and system based on travel sharing network

Abstract

The invention provides an online car-hailing area minimum motorcade putting method and system based on a travel sharing network, and the method comprises the steps: finding out key nodes with higher path influence through the analysis of a traffic network diagram, carrying out the clustering of a key node set, obtaining a plurality of clustering clusters, and carrying out the clustering of the clustering clusters; and constructing a shared network with each travel as a node by using a hot spot region formed by each cluster, analyzing the shared network by using a minimum path coverage method, and obtaining the minimum path number which is the minimum motorcade number which can be arranged. According to the method, an operable theoretical framework and practical reference are provided for transport capacity optimization of the urban high-demand area. Through forming a local motorcade cycle in a hot spot region, cross-region empty driving is reduced, and configuration optimization of regional-level online car-hailing transport capacity resources is realized.

Description

Technical Field

[0001] This invention belongs to the field of ride-hailing dispatch, and in particular relates to a method and system for deploying the smallest regional fleet of ride-hailing vehicles based on a ride-sharing network. Background Technology

[0002] In recent years, with the acceleration of urbanization and the development of the sharing economy, ride-hailing services have become an important mode of transportation for urban residents due to their flexible and efficient operation. However, the current allocation of transportation capacity resources in ride-hailing systems still faces significant challenges. On the one hand, urban transportation demand exhibits significant spatiotemporal heterogeneity, with supply and demand imbalances frequently occurring during morning and evening rush hours and in popular areas such as commercial centers and transportation hubs. On the other hand, existing dispatching mechanisms are mostly based on the assumption of global equilibrium, making it difficult to characterize the demand differences between regions, resulting in high empty-running rates and increased passenger waiting times, which is detrimental to resource conservation and traffic congestion relief. Therefore, constructing a method that can reflect regional demand differences and provide vehicles on demand to improve the efficiency of transportation capacity resource allocation has important theoretical significance and practical value for urban traffic management.

[0003] Currently, scholars both domestically and internationally have conducted extensive research on ride-hailing capacity optimization, primarily focusing on three aspects: shared network modeling, vehicle scheduling optimization, and hotspot area identification. In the shared network field, related research often utilizes graph theory methods to characterize the shareable relationships between trips. However, limited by static road network structures and fixed pricing mechanisms, these methods are still prone to getting stuck in local optima at the regional level, making it difficult to achieve global resource equilibrium.

[0004] In the identification of hotspot areas, density-based clustering algorithms are widely used. DBSCAN can identify high-density areas of arbitrary shapes based on spatial distribution characteristics and is widely used for hotspot mining of ride-hailing pick-up and drop-off points. Despite the abundance of related results, existing research is still mainly based on static historical data, which makes it difficult to fully reflect the demand correlation and dynamic changes between hotspot areas, and the response of hotspot boundaries to fluctuations in actual demand also shows a lag.

[0005] Current ride-hailing management primarily relies on order forecasting and real-time dispatch mechanisms. Platforms typically guide vehicles to areas with high order density, but they lack the ability to identify the actual demand and reusable trip chains within these areas. Due to the lack of spatial and temporal connectivity, platforms cannot determine the "minimum number of vehicles needed to satisfy a hotspot area within a specific time period," nor can they determine when an area is already saturated or overloaded. In this situation, ride-hailing drivers spontaneously congregate in hotspot areas, and platforms lack a basis for setting deployment limits. Furthermore, the government has not provided regional deployment control standards, ultimately leading to excessive vehicle concentration in hotspot areas, causing problems such as empty runs, delays, and localized congestion. Summary of the Invention

[0006] The main problem addressed by this invention is how to effectively schedule ride-hailing vehicles, increase vehicle reuse rates, and reduce empty mileage and regional congestion. It provides a method and system for deploying the smallest regional fleet of ride-hailing vehicles based on a ride-sharing network.

[0007] To solve the above technical problems, the technical solution adopted is: A method for deploying a minimum regional fleet of ride-hailing vehicles based on a ride-sharing network includes the following steps: Step 1: Obtain and analyze the GPS trajectory dataset of ride-hailing vehicles in the area to be studied to obtain the set of key travel nodes in the traffic network map of the area to be studied; Step 2: Perform cluster analysis on the key travel node set to obtain multiple hotspot clusters forming multiple travel hotspot areas; Step 3: Build a shared network for each travel hotspot area; Step 4: Analyze each shared network to obtain the minimum number of vehicles for each shared network.

[0008] Furthermore, it also includes constructing a global network for the area to be studied, analyzing the global network, and obtaining the minimum number of vehicles in the area to be studied.

[0009] Furthermore, the method for obtaining and analyzing the GPS trajectory dataset of ride-hailing vehicles in the area under study to obtain the set of key travel nodes in the traffic network map of the area under study is as follows: Step 1.1: Based on the GPS trajectory dataset of ride-hailing vehicles in the area to be studied, map and match the start and end points of each ride-hailing trip to spatial travel nodes in the transportation network, and count the number of times passengers get on and off at each spatial travel node. The trip refers to a service journey from start to finish. Step 1.2: Construct an undirected topology graph. The nodes in the undirected topology graph are the spatial travel nodes in the transportation network graph, and the edges are the road connections between the spatial travel nodes. Step 1.3: Calculate the comprehensive centrality score for each spatial travel node in the undirected topology graph. The comprehensive centrality score refers to the score after comprehensively considering the node's connectivity strength, the quality of the node's neighbors, and the node's structural perturbation rate. Step 1.4: Sort the overall centrality scores of each spatial travel node from high to low, and select the top-ranked spatial travel nodes as the key travel node set.

[0010] Furthermore, the method for calculating the comprehensive centrality score for each spatial travel node is as follows: The overall centrality score of spatial travel nodes is: ; Representing spatial travel nodes The strength of local connectivity within the k-th order neighborhood Indicates the quality of the node's neighbors. Indicates the node structure perturbation rate; They represent the weights, and ; For undirected topological graphs Any spatial travel node in Constructing spatial travel nodes k-order neighborhood , ;in, Representing spatial travel nodes , The number of hops in the shortest path between nodes. For any spatial travel node in the network; Space travel nodes The local connectivity strength is: ; As a node for space travel The local average shortest path in the k-th order neighborhood; ; As a node for space travel neighborhood Shortest path distance within the area; Representing spatial travel nodes The number of spatial travel nodes within the neighborhood; ; Representing spatial travel nodes The number of times passengers get on and off the bus within the statistical time window. This represents the maximum number of boarding and alighting times across all spatial travel nodes within the study area and time window. Space travel nodes Perturbation rate of its neighborhood results: ; in, Indicates the removal of a node. The average shortest path length within the k-th order neighborhood network.

[0011] Furthermore, the method for clustering analysis of key travel node sets is as follows: Step 2.1: Obtain the set of key travel nodes, as well as the geographic coordinates and overall centrality score of each key travel node; Step 2.2: Use the overall node centrality score of each key node as the node weight to replace the local density parameter in the density hierarchical clustering algorithm HDBSCAN for cluster analysis of the key node set; Step 2.3: After cluster analysis, multiple hotspot clusters are obtained, forming multiple travel hotspot areas, which are then output.

[0012] Furthermore, the method for building a shared network for each travel hotspot area is as follows: For each hotspot cluster, construct a directed acyclic graph (DAG). The nodes in the DAG represent each trip within the hotspot area. The directed edges in the DAG refer to the directed connections between the end of one trip and the start of the next trip for the same vehicle. The directed edges in the DAG satisfy the following conditions: the sum of the end time of the previous trip and the time between the end of the previous trip and the start time of the next trip must be earlier than the departure time of the next trip; and the time difference between the departure time of the next trip and the end time of the previous trip must meet the trip connection time threshold.

[0013] Furthermore, the method for constructing the global network of the region to be studied is as follows: A directed acyclic graph (DAG) is constructed in the study area. The nodes in the DAG represent each trip within the entire study area. The directed edges in the DAG refer to the directed connections between the end of one trip and the start of the next trip for the same vehicle. The directed edges in the DAG satisfy the following conditions: the sum of the end time of the previous trip and the time between the end of the previous trip and the start time of the next trip must be earlier than the departure time of the next trip; the time difference between the departure time of the next trip and the end time of the previous trip must meet the trip connection time threshold. Each directed edge in the DAG is assigned a weight, making the DAG a directed weighted graph.

[0014] Furthermore, the method to obtain the minimum number of vehicles for each shared network or global network is as follows: For a directed acyclic graph of a shared network or a global network, the minimum path coverage method is used to find the minimum number of consecutive paths that cover all travel nodes. The minimum number of consecutive paths represents the minimum number of vehicles required. The minimum path coverage method divides all nodes in the directed acyclic graph into several paths and satisfies that each node belongs to only one path, and all nodes are covered by the path.

[0015] Furthermore, the method for finding the minimum number of continuous paths using the minimum path cover method is to solve the directed acyclic graph using the Hopcroft-Karp algorithm with bisection matching.

[0016] This invention also provides a ride-hailing regional minimum fleet deployment system based on a ride-sharing network, which is implemented using the steps of the ride-hailing regional minimum fleet deployment method based on a ride-sharing network described above.

[0017] By adopting the above technical solution, the present invention has the following beneficial effects: This invention provides a method and system for minimum fleet deployment of ride-hailing services in a region based on a shared mobility network. By analyzing the traffic network map, key nodes with high path influence are identified. These key node sets are then clustered to obtain multiple clusters. A shared network is constructed using the hotspot areas formed by each cluster, with each trip as a node. The minimum path coverage method is used to analyze the shared network, and the number of minimum paths obtained is the minimum number of fleets that can be deployed. This invention provides an operable theoretical framework and practical reference for optimizing transportation capacity in high-demand urban areas. By forming local fleet circulation in hotspot areas, cross-regional empty runs are reduced, achieving optimized allocation of regional ride-hailing transportation capacity resources. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a system flowchart of the present invention; Figure 2 A map showing the transportation network and node distribution; Figure 3 Spatial distribution map of GPS map matching results; Figure 4 This is a spatial distribution map of key nodes based on a real road network; Figure 5 Cluster distribution map of hotspot areas; Figure 6 The diagrams are schematic diagrams of shared networks. (a) is a schematic diagram of the shared network of cluster 4, and (b) is a schematic diagram of the global shared network. Detailed Implementation

[0020] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Figures 1 to 6 This invention illustrates a specific embodiment of a method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network, comprising the following steps: Step 1: Obtain and analyze the GPS trajectory dataset of ride-hailing vehicles in the area to be studied to obtain the set of key travel nodes in the traffic network map of the area to be studied.

[0022] Since there are many nodes in the transportation network, it is necessary to identify key nodes for research. The importance of a node is usually measured by indicators such as degree centrality, betweenness centrality, and proximity centrality. Degree centrality represents the number of connections of a node, betweenness centrality represents the frequency of a node's occurrence in all shortest paths, and proximity centrality represents the average distance from a node to all other nodes. Traditional research methods calculate the importance of each node by combining these indicators. Reference 1, "Zhang Tingping, Wan Di. An algorithm for evaluating the importance of rail transit nodes based on rough set theory [J]. Journal of Jilin University (Engineering Science), 2022, 52(12): 2845-2851", proposes an algorithm for evaluating the importance of nodes from three dimensions: degree centrality, proximity centrality, and order centrality. However, it has certain limitations in reflecting the complex structural characteristics of the transportation network and is difficult to fully reflect the actual impact of nodes on path connectivity and vehicle reuse in their local structure. This embodiment proposes an improved LASPN local centrality model to further identify key nodes and combines density clustering algorithm to more accurately identify hotspots. In transportation networks, single common global centrality indicators often fail to effectively identify hotspot nodes with complex structures and heterogeneous demands. Therefore, based on LASPN (Local Average Shortest Path with Extended Neighborhood), a ternary centrality metric model integrating structural strength, neighbor quality, and path perturbation is constructed. Considering the actual characteristics and computational feasibility of shared mobility data, the above three indicators are quantitatively replaced by local connectivity, travel activity, and structural perturbation rate, respectively, thereby achieving the computability and empirical validation of the model.

[0023] To verify the effectiveness of the method proposed in this embodiment, a GPS trajectory dataset of ride-hailing vehicles from a portion of a city was selected for analysis and verification. The study period was from August 1st to August 4th, 2020, during the morning rush hour (7:00–9:00). The trajectory points covered longitudes of 116.30350°E - 116.35105°E and latitudes of 39.90526°N - 39.93146°N. From 80,620 trip data records, 1,662 trip data records were selected for the 7:00–9:00 time period. This area includes typical urban functional areas such as shopping malls, subway entrances, and residential areas, exhibiting traffic characteristics of a mix of commuting and flexible travel. The dataset contains many fields; we used the following: start time, arrival time, start longitude, start latitude, destination longitude, and destination latitude. Time measurement accuracy was in seconds.

[0024] During the collection of ride-hailing GPS data, abnormal data in the unprocessed raw data can directly affect the accuracy of subsequent hotspot area delineation and shared network construction. Therefore, the first step is to clean the data, removing outliers, including missing values ​​and duplicate records (such as identical trip start and end points, abnormal travel distances or durations), and performing coordinate system transformation to ensure that the ride-hailing GPS trajectory data and road network data are in the same coordinate system.

[0025] The map matching process used openstreetmap.org to create the city's street network. A filtering method was applied to the streets, omitting sidewalks, internal roads within residential areas, and paths unlikely to contain order origin and destination points. Street intersections were extracted to construct a network, where road intersections are treated as nodes, and roads connecting intersections are treated as directed edges. Figure 2 As shown in the diagram. Then, the GPS trip start and destination nodes are matched to the intersection nodes of the street network. These locations are within 100 meters of at least one intersection node in the street network, which aligns with most trip scenarios. Other redundant nodes are removed, forming the basic node input for the regional shared network and path reuse model. After cleaning and filtering the extracted road intersection network, some obvious inconsistencies and redundancies are removed, resulting in a final region containing 650 nodes and 1331 directed edges. After map matching, the trip's start and destination nodes possess clear road topology attributes, providing a spatial basis for subsequent shared network construction and accessibility determination.

[0026] To visually demonstrate the map matching effect and verify the rationality of the data processing, this application selected a portion of the trajectories during the morning rush hour (7:00–10:00) within the study area and plotted their spatial distribution map, revealing the coverage of the travel trajectories in the road network and their spatial characteristics, such as... Figure 3 As shown. From Figure 3As can be seen, the travel trajectories after map matching maintain a high degree of consistency with the actual road structure, indicating that map matching effectively improves the spatial accuracy of travel data. Meanwhile, the origin-destination (OD) points show a clear clustering phenomenon on major arterial roads, urban expressways, and key transportation corridors, consistent with the travel characteristics of the study area. This result provides a reliable data foundation for subsequent hotspot area extraction, shared network construction, and minimum path coverage model, and also ensures the credibility of subsequent experimental results.

[0027] In this embodiment, the method for obtaining the key travel node set in the traffic network map of the proposed study area by matching the ride-hailing GPS trajectory dataset with the map is as follows: Step 1.1: Based on the GPS trajectory dataset of ride-hailing vehicles in the area to be studied, map and match the start and end points of each ride-hailing trip to spatial travel nodes in the transportation network, and count the number of times the trip is picked up and dropped off at each spatial travel node. The trip refers to a service journey from start to finish.

[0028] In this embodiment, an undirected topological graph is constructed. ,in This is a set of nodes, representing spatial travel nodes (such as intersections or road junctions) in the urban road network. Let be the edge set, representing the road connections between nodes.

[0029] For any node , construct its Extended neighborhood: ; in, Representing spatial travel nodes The shortest path length between them. For any spatial travel node in the network; Step 1.2: Construct an undirected topology graph. The nodes in the undirected topology graph are the spatial travel nodes in the transportation network graph, and the edges are the road connections between the spatial travel nodes.

[0030] Step 1.3: Calculate the comprehensive centrality score for each spatial travel node in the undirected topology graph. The comprehensive centrality score refers to the score after comprehensively considering the node's connectivity strength, the quality of the node's neighbors, and the node's structural perturbation rate.

[0031] In this embodiment, the method for calculating the comprehensive centrality score for each spatial travel node is as follows: The overall centrality score of spatial travel nodes is: ; in, Representing spatial travel nodes Local connectivity in its k-order neighborhood subgraph Indicates the quality of the node's neighbors. Indicates the node structure perturbation rate; These represent the weight parameters, and The weighting parameters ensure that each indicator contributes equally.

[0032] For undirected topological graphs Any spatial travel node in Constructing spatial travel nodes k-order neighborhood , ;in, Representing spatial travel nodes , The number of hops in the shortest path between nodes. For any spatial travel node in the network; Space travel nodes The local connectivity strength is: ; As a node for space travel The local average shortest path in its k-th order neighborhood; ; As a node for space travel In its k-th order neighborhood The shortest path distance within; Representing spatial travel nodes The number of spatial travel nodes within the neighborhood; The smaller the value, the denser the neighborhood structure, and the more nodes... The stronger the local connectivity, the better. This indicates how dense the structure of a node is within its local neighborhood. For nodes The local average shortest path length reflects the average spatial connectivity distance between the node and its neighboring nodes. This is achieved by... Taking the reciprocal transforms the negative indicator of path length into a positive indicator of connectivity strength, meaning that the shorter the path around a node, the higher its local connectivity strength. This reflects the connectivity efficiency of a node in the local network.

[0033] ; Indicates the location of space travel nodes Number of times to get on and off the bus, The maximum number of boarding and alighting times among all spatial travel nodes within the k-th order neighborhood and the statistical time window; node neighbor quality. By from the node The number of times a location is used to board and alight is used to reflect the node's activity level, which reflects the intensity of demand for the node in actual travel. After normalization, it becomes... This reflects the relative travel intensity of the nodes.

[0034] Further measure the nodes The impact of disturbances on the neighborhood structure is calculated by comparing the changes in the average shortest path length of the neighborhood before and after a node is removed. If a node is removed, its average shortest path will change. The structural disturbance rate can be used to measure the degree of influence of a node on the connectivity of neighborhood paths. Therefore, spatial travel nodes... Perturbation rate of its neighborhood results: ; ; in, Indicates the removal of a node. Average shortest path length in the next neighborhood; For nodes The subgraph constructed from the k-th order neighborhood, For subgraph Remove node The perturbation subgraph is constructed using the node and its associated edges. A higher perturbation rate indicates that the node is more critical in the local structure, reflecting the degree of influence of the node on the stability of the network structure.

[0035] Step 1.4: Sort the overall centrality scores of each spatial travel node from high to low, and select the top-ranked spatial travel nodes as the key travel node set.

[0036] In this embodiment, the overall centrality score of all network nodes is calculated, and the nodes with significantly high centrality in the top 20% of the node set are selected in descending order of their overall centrality scores and denoted as the key node set. .

[0037] In shared mobility networks, key nodes often have high path influence. By constructing an undirected topology graph of the nodes in the transportation network and analyzing the comprehensive scores of each node, key nodes can be identified, providing a scientific basis for this identification. For example... Figure 4 As shown. Ultimately, 60 key nodes were identified. Figure 4 It contains 1,322 real roads, 226 traffic lights, and 713 bus stops. Nodes are categorized into high, medium, and low centrality based on their overall centrality score, represented by points of different colors and sizes.

[0038] Step 2: Perform cluster analysis on the key travel node set to obtain multiple hotspot clusters forming multiple travel hotspot areas.

[0039] In this embodiment, the method for clustering analysis of the key travel node set is as follows: Step 2.1: Obtain the set of key travel nodes, as well as the geographic coordinates and overall centrality score of each key travel node.

[0040] In this embodiment, the key node set is as follows: ; In the formula: Representing nodes, each node contains two attributes: geographic coordinates. ; Represents a two-dimensional coordinate point, referring to the position coordinates (longitude and latitude) of each node on a two-dimensional plane.

[0041] Step 2.2: Use the overall node centrality score of each key node as the node weight to replace the local density parameter in the density hierarchical clustering algorithm HDBSCAN for cluster analysis of the key node set; Step 2.3: After cluster analysis, multiple hotspot clusters are obtained, forming multiple travel hotspot areas, which are then output.

[0042] In this embodiment, the clustering method used is based on the density hierarchical clustering algorithm HDBSCAN, which is derived from "Sander J. Density-based clustering[M] / / Encyclopedia of machine learning and data mining. Springer, Boston, MA, 2017: 349-353.". Considering the differences in demand intensity among different key nodes, the node comprehensive centrality score is used as the node weight during the clustering process. That is, in the existing HDBSCAN algorithm, each node contributes the same to the local density, which can be equivalently understood as its contribution weight being 1. In this embodiment, the characteristic of each node having a unit weight of 1 in the local density estimation of the traditional clustering algorithm is changed. Node weights are used to replace the local density in the current clustering algorithm HDBSCAN, so that nodes with higher demand intensity have a greater influence in hotspot area identification. By introducing node weights into the clustering algorithm, more important points are more easily identified as hotspots during clustering.

[0043] This embodiment assigns weights based on the importance of key nodes, i.e., their comprehensive centrality scores. This departs from traditional density estimation processes that rely solely on spatial distance and the number of neighboring nodes, assuming each node contributes equally to density. This approach fails to reflect the differences in demand intensity and functionality among nodes in a real-world travel system. Higher node weights result in a greater effective contribution to density estimation, making it easier to form stable core nodes in subsequent hierarchical clustering. Node weights ensure that the clustering algorithm's influence on cluster formation is no longer solely determined by spatial proximity, but is simultaneously influenced by both travel demand intensity and network structural importance. In this way, demand information is naturally embedded into the clustering process while maintaining the stability and interpretability of the original HDBSCAN algorithm.

[0044] In this embodiment, to further achieve precise resource allocation in spatial regions, these key nodes need to be aggregated into hotspot regions with spatial continuity and structural connectivity. Traditional clustering methods (such as K-Means or DBSCAN) have certain limitations when dealing with spatially irregular and density-heterogeneous data, and neither can integrate non-spatial features such as node structural importance during the clustering process, which may lead to the final aggregation result deviating from the actual network topology. Therefore, the weighted HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) clustering model is adopted. Based on the HDBSCAN algorithm, node attribute weights are introduced, combining the advantages of DBSCAN and hierarchical clustering. It can automatically determine the number of clusters and identify noise points, without relying on a fixed neighborhood radius. It can effectively deal with the situation where hotspot regions are "of varying sizes and densities" in real data, and considers spatial adjacency and structural importance during clustering, identifying hotspot regions that are merged into multiple attributes. Compared to traditional density clustering methods that rely solely on Euclidean distance, this embodiment utilizes a weighted HDBSCAN clustering method based on weighted core distance and reach distance to spatially cluster node centrality, thereby extracting hotspot regions. The clustering process incorporates features such as node structural importance, making the final aggregation result closer to the actual road network topology, and providing a basis for subsequent minimum fleet size configuration and local area optimization. To more intuitively illustrate the spatial distribution characteristics of hotspot regions, a hotspot region clustering distribution map within the study area is shown below. Figure 5 As shown in Table 1, the clustering results are as follows: Figure 5It can be seen that travel demand within the study area exhibits a typical dual-core, multi-level clustering pattern. The western area is the most prominent high-density center, characterized by the darkest color and the densest isodensity lines, making it the highest-density hotspot center on the entire map. The surrounding area shows a concentric gradient decay, indicating that this region has the highest concentration of points / demands. The northeastern area is a secondary core area, while the areas between the two and the southeastern region are medium-to-low-density transition zones and edge areas, indicating a high degree of spatial imbalance and significant clustering characteristics between road / infrastructure resources and travel demand. Further incorporating travel demand intensity allows for a more comprehensive characterization of the actual travel hotspot structure. This ensures that the clustering results simultaneously reflect the importance of spatial continuity and demand intensity, resulting in hotspot areas with stable structures, concentrated demand, and clear outlines.

[0045] Table 1 Clustering Results

[0046] The clustering results reveal a clear hierarchical structure in the overall centrality of nodes within the study area. Clusters 1 and 0 exhibit significantly higher average centrality (the average of the overall centrality scores of all nodes within the same cluster, used to evaluate the cluster's hub status in the network), demonstrating prominent hub attributes and representing the core hotspots with the highest connectivity efficiency within the region. In contrast, clusters 2, 3, and 4 have larger node sizes and centrality values ​​ranging from 0.66 to 0.69, forming a secondary hotspot hierarchy supporting major traffic activities. Their spatial distribution typically aligns with arterial roads and mixed-use areas. Overall, the centrality differences among clusters reveal a significant differentiation in the structural importance of travel nodes within the study area, forming a continuous gradient from core hotspots to peripheral regions. This spatial structure not only reflects the hierarchical characteristics of the road network's functional layout but also provides a clear regional stratification basis for subsequent minimum path coverage solutions and minimum fleet configurations.

[0047] Step 3: Build a shared network for each travel hotspot area.

[0048] In this embodiment, the method for constructing a shared network for each travel hotspot area is as follows: A directed weighted graph is constructed for each hotspot cluster. The nodes in the directed weighted graph represent each trip within the hotspot area. A trip refers to the time from boarding to alighting. The directed edges in the directed weighted graph represent the connection between the end of one trip and the start of the next trip for the same vehicle. The directed edges in the directed weighted graph satisfy the following conditions: the sum of the end time of the previous trip and the time between the end of the previous trip and the start time of the next trip must be earlier than the departure time of the next trip; and the time difference between the departure time of the next trip and the end time of the previous trip must meet the trip connection time threshold.

[0049] Each travel node within the hotspot area It includes departure time. Arrival time ,starting point ,end The attribute of, let the first A total of 1 hotspot areas A valid trip is recorded as ,in Representing each trip. Constructed as a directed weighted graph. ,in This is the set of travel nodes in this region. Represents all directed edges in the network if and only if ( express , Travel time between (representing the formation of nodes) and (The upper limit of the connection time between two trips is the trip connection time threshold) If the network link (directed edge) between the two trips exists, it indicates that the two trips can be served continuously by a single vehicle, ensuring that the vehicles can connect between the two trips, while limiting the maximum waiting interval for the two trips to be completed sequentially. Time reachability constraints are used to characterize the connection conditions between trips.

[0050] In other words, nodes in a shared network are key nodes, and edges must satisfy the following two constraints to construct directed edges: ; ; Indicates itinerary Destination and Itinerary Travel time between origins This is the threshold for trip connection time.

[0051] Set a trip connection time threshold based on trip time. : ; In the formula: Refers to the current itinerary The duration of the trip. An exponential function is available, where parameter 'a' represents the base tolerance level, and 'b' reflects the strength of the impact of increased trip time on tolerance. Initial parameter settings for 'a' and 'b' can be based on experience (e.g., 'a' = 5 min, 'b' = 0.02). Setting the trip connection time threshold as a dynamic threshold related to trip duration is more consistent with actual passenger behavior than setting a fixed threshold. When the sharing tolerance threshold increases, the number of effective connections in the network increases, and the trip reuse rate improves, theoretically reducing the minimum fleet size required to achieve full coverage; conversely, it indicates limited reuse and decreased vehicle utilization.

[0052] Step 4: Analyze each shared network to obtain the minimum number of vehicles for each shared network.

[0053] In this embodiment, when it is necessary to configure the minimum number of vehicles for the entire study area, a global network for the study area can also be constructed using a shared network construction method. The method for constructing the global network for the study area is as follows: A directed acyclic graph (DAG) is constructed in the study area. The nodes in the DAG represent each trip within the entire study area. The directed edges in the DAG refer to the directed connections between the end of one trip and the start of the next trip for the same vehicle. The directed edges in the DAG satisfy the following conditions: the sum of the end time of the previous trip and the time between the end of the previous trip and the start time of the next trip must be earlier than the departure time of the next trip; the time difference between the departure time of the next trip and the end time of the previous trip must meet the trip connection time threshold. Each directed edge in the DAG is assigned a weight, making the DAG a directed weighted graph.

[0054] In this embodiment, the method for analyzing each shared network or global network and obtaining the minimum number of vehicles for each shared network or global network is the same, namely: For a directed acyclic graph (DAG) in a shared or global network, the minimum path cover method is used to find the minimum number of consecutive paths covering all trip nodes. The minimum number of consecutive paths represents the minimum number of vehicles required. Minimum path cover involves dividing all nodes in the DAG into several paths, ensuring that each node belongs to only one path, and all nodes are covered by a path. A path in a DAG corresponds to a trip sequence that can be continuously served by a single vehicle.

[0055] This embodiment, based on the constructed shared network, transforms the problem into finding the minimum path cover in a directed acyclic graph. That is, without omitting any trips, it aims to cover all trip nodes with the minimum number of consecutive paths (vehicles) to obtain the minimum fleet size required for the region. Unlike traditional static calculation methods based on order volume, this method fully considers the spatiotemporal connectivity between trips and incorporates the continuous service potential of vehicles into capacity allocation analysis, thereby avoiding redundant vehicle configurations and possessing greater practical application value.

[0056] In this embodiment, the method for finding the minimum number of continuous paths using the minimum path cover method is to solve the directed acyclic graph using the Hopcroft-Karp algorithm with bisection matching.

[0057] In a Directed Acyclic Graph (DAG), a time-connection constraint is introduced to filter out paths that can be continuously connected. Under the DAG structure, the minimum path cover problem can be transformed into a maximum matching problem. The core idea of ​​minimum path cover is to divide all nodes in the graph into several paths, satisfying that each node belongs to only one path; each path is a directed path in the graph; and all nodes are covered by paths. Each path represents a continuous task for a vehicle. Finding the minimum number of paths that can cover all journeys is equivalent to finding the minimum number of vehicles needed. According to graph theory, this can be expressed by the following formula: ; In the formula: This represents the minimum path coverage. This represents the total number of nodes in the DAG. This represents the maximum matching number of the bipartite graph corresponding to this graph.

[0058] In this embodiment, the Hopcroft-Karp algorithm for bipartite matching (Hopcroft JE, Karp R M. An n5=2 algorithm for maximum matching in bipartite graphs[J]. SIAM J.Comput, 1973, 2: 225-231) is used to copy each node of the directed acyclic graph into two copies and place them into the left and right node sets respectively; if there exists a directed edge Then, in the bipartite graph, connect the left side. With the right side The classic maximum matching algorithm, Hopcroft-Karp, is used to solve this bipartite graph, where each matching edge represents a successful journey connection. According to the minimum path cover theory of directed acyclic graphs, the minimum number of paths required to cover all journeys is the total number of nodes minus the maximum number of matching edges, thus yielding the minimum vehicle deployment size.

[0059] The Hopcroft–Karp algorithm continuously searches for and expands augmenting paths until no augmenting paths remain in the graph, at which point the matching size is maximized, thus ensuring the optimality of the minimum path cover result. It is highly efficient in handling large-scale sparse bipartite graph matching problems, with a time complexity of O(log n). After obtaining the maximum number of matches using this algorithm, the formula is used. This allows you to calculate the minimum number of vehicles in the corresponding area.

[0060] This embodiment transforms the problem of finding the minimum fleet size into a directed acyclic graph (DAG) constructed with each trip as a node, thus converting the problem into a pathfinding problem within the DAG. This allows for the rapid determination of the minimum required fleet size. Compared to traditional methods, this embodiment further reduces the number of nodes to study by analyzing the traffic network graph to identify key nodes. Then, it uses clustering to delineate hotspot areas, reflecting the variability in travel demand. This significantly improves the rationality and accuracy of capacity allocation while ensuring coverage.

[0061] This embodiment constructs a shared network for each hotspot area and a global shared network for the area to be studied. For example... Figure 6 As shown, Figure 6 In (a), different colors represent different sub-regions, and the black border represents the connection relationship of cluster 4. Figure 6 In (b), blue nodes represent travel nodes in cluster 3, brown nodes represent travel nodes in cluster 4, and dark blue nodes represent key hub nodes in the network, which may be important travel nodes connecting multiple clusters or nodes with high centrality in the network. Gray edges represent intra-cluster connections, and red edges represent inter-cluster connections. Nodes are arranged along the circumference for easy observation of the overall network structure; the more edges, the more connection opportunities there are in that area. Figure 6The results show that the internal shared network has stronger local connectivity and more concentrated high-centrality nodes, reflecting the characteristics of strong trip reuse and high demand density within the hotspot area. In contrast, the global shared network has more dispersed node relationships, exhibiting a multi-centrality characteristic with multiple high-centrality nodes coexisting, reflecting the diversity of travel demand and trip reuse relationships between different functional areas at the urban scale. The comparative results demonstrate that different scale network construction strategies—local shared networks built based on hotspot areas and global shared networks built based on the entire study area—can reveal different levels of trip reuse structures, providing multi-scale structural basis for subsequent minimum path coverage solutions and minimum fleet configuration.

[0062] After constructing the shared network at different scales, the maximum number of matching edges (each matching edge represents the number of trip connections and vehicle reuses that can be achieved in one trip), minimum vehicle demand, and vehicle saving effect were calculated for each hotspot area and the global network based on the minimum path coverage model. Simultaneously, the structural indicators such as the number of vehicles, path length, and single-trip path ratio were analyzed using the obtained optimal path set to compare the differences in the reuse capacity of the shared network in different areas. The corresponding results are shown in Tables 2 and 3.

[0063] Table 2. Results of Minimum Path Cover Solution

[0064] Table 3 Statistical Characteristics of Vehicle Routes

[0065] The results show significant differences in minimum vehicle requirements and reuse efficiency among different hotspot regions. Clusters 3 and 4, despite their larger node sizes, still achieved vehicle savings rates exceeding 53%–56%, indicating that mesoscale hotspot regions possess high structural reuse potential. The global network's savings rate further increased to 57.6%, demonstrating that cross-regional network integration can achieve even stronger overall reuse efficiency.

[0066] Path structure metrics further validate this trend: the global network not only has a longer average path length, but the longest path also extends to 7, indicating that under global network conditions, a single ride-hailing vehicle can theoretically complete up to 7 consecutive single trips while meeting time and space constraints. This also demonstrates a more complete cross-regional reuse chain. In contrast, small-scale hotspots (such as Cluster 2) only form a short path structure of 1 vehicle and 2 trips, with limited reuse space. Overall, as the network scales up, the path coverage chain becomes longer and the reuse degree is stronger, ultimately leading to a more significant vehicle saving effect.

[0067] This embodiment proposes a hotspot area identification method based on weighted HDBSCAN clustering and improved LASPN centrality analysis within the framework of shared network modeling and minimum path coverage solution. Furthermore, it introduces regional demand differences for edge weight design, providing urban traffic management departments with a regional-level ride-hailing capacity allocation method. This method forms local fleet circulation in hotspot areas, reducing cross-regional empty runs, and identifies long-term demand centers through the shared network structure, thereby optimizing the allocation of regional ride-hailing capacity resources. The results show that this method can effectively characterize the travel characteristics of urban hotspot areas and provides a new approach to calculating the minimum fleet size. Future research can be further extended to fields such as autonomous ride-hailing, providing a basis for long-term urban planning and offering more practical support for urban traffic management and travel services.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for deploying a minimum regional fleet of ride-hailing vehicles based on a ride-sharing network, characterized in that, Includes the following steps: Step 1: Obtain and analyze the GPS trajectory dataset of ride-hailing vehicles in the area to be studied to obtain the set of key travel nodes in the traffic network map of the area to be studied; Step 2: Perform cluster analysis on the key travel node set to obtain multiple hotspot clusters forming multiple travel hotspot areas; Step 3: Build a shared network for each travel hotspot area; Step 4: Analyze each shared network to obtain the minimum number of vehicles for each shared network.

2. The method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network according to claim 1, characterized in that, It also includes constructing a global network for the area to be studied, analyzing the global network, and obtaining the minimum number of vehicles in the area to be studied.

3. The method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network according to claim 2, characterized in that, The method for obtaining and analyzing the GPS trajectory dataset of ride-hailing vehicles in the area under study to obtain the set of key travel nodes in the traffic network map of the area under study is as follows: Step 1.1: Based on the GPS trajectory dataset of ride-hailing vehicles in the area to be studied, map and match the start and end points of each ride-hailing trip to spatial travel nodes in the transportation network, and count the number of times passengers get on and off at each spatial travel node. The trip refers to a service journey from start to finish. Step 1.2: Construct an undirected topology graph. The nodes in the undirected topology graph are the spatial travel nodes in the transportation network graph, and the edges are the road connections between the spatial travel nodes. Step 1.3: Calculate the comprehensive centrality score for each spatial travel node in the undirected topology graph. The comprehensive centrality score refers to the score after comprehensively considering the node's connectivity strength, the quality of the node's neighbors, and the node's structural perturbation rate. Step 1.4: Sort the overall centrality scores of each spatial travel node from high to low, and select the top-ranked spatial travel nodes as the key travel node set.

4. The method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network according to claim 3, characterized in that, The method for calculating the overall centrality score for each spatial travel node is as follows: The overall centrality score of spatial travel nodes is: ; Representing spatial travel nodes The strength of local connectivity within the k-th order neighborhood Indicates the quality of a node's neighbors. Indicates the node structure perturbation rate; They represent the weights, and ; For undirected topological graphs Any spatial travel node in Constructing spatial travel nodes k-order neighborhood , ;in, Representing spatial travel nodes , The shortest path hop count between nodes. For any spatial travel node in the network; Space travel node The local connectivity strength is: ; As a node for space travel The local average shortest path in the k-th order neighborhood; ； As a node for space travel neighborhood Shortest path distance within the interior; Representing spatial travel nodes The number of spatial travel nodes within the neighborhood; ； Representing spatial travel nodes The number of times passengers get on and off the bus within the statistical time window. This represents the maximum number of boarding and alighting times across all spatial travel nodes within the study area and time window. Space travel node Perturbation rate of its neighborhood results: ； in, Indicates the removal of a node. The average shortest path length within the k-th order neighborhood network.

5. The method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network according to claim 4, characterized in that, The method for clustering analysis of key travel node sets is as follows: Step 2.1: Obtain the set of key travel nodes, as well as the geographic coordinates and overall centrality score of each key travel node; Step 2.2: Use the overall node centrality score of each key node as the node weight to replace the local density parameter in the density hierarchical clustering algorithm HDBSCAN for cluster analysis of the key node set; Step 2.3: After cluster analysis, multiple hotspot clusters are obtained, forming multiple travel hotspot areas, which are then output.

6. The method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network according to claim 5, characterized in that, The method for building a shared network for each travel hotspot area is: For each hotspot cluster, construct a directed acyclic graph (DAG). The nodes in the DAG represent each trip within the hotspot area. The directed edges in the DAG refer to the directed connections between the end of one trip and the start of the next trip for the same vehicle. The directed edges in the DAG satisfy the following conditions: the sum of the end time of the previous trip and the time between the end of the previous trip and the start time of the next trip must be earlier than the departure time of the next trip; and the time difference between the departure time of the next trip and the end time of the previous trip must meet the trip connection time threshold.

7. A method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network, as described in claim 6, is characterized in that... The method for constructing the global network of the region to be studied is as follows: A directed acyclic graph (DAG) is constructed in the study area. The nodes in the DAG represent each trip within the entire study area. The directed edges in the DAG refer to the directed connections between the end of one trip and the start of the next trip for the same vehicle. The directed edges in the DAG satisfy the following conditions: the sum of the end time of the previous trip and the time between the end of the previous trip and the start time of the next trip must be earlier than the departure time of the next trip; the time difference between the departure time of the next trip and the end time of the previous trip must meet the trip connection time threshold. Each directed edge in the DAG is assigned a weight, making the DAG a directed weighted graph.

8. A method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network, as described in claim 6 or 7, characterized in that... The method to obtain the minimum number of vehicles for each shared network or global network is as follows: For a directed acyclic graph of a shared network or a global network, the minimum path coverage method is used to find the minimum number of consecutive paths that cover all travel nodes. The minimum number of consecutive paths represents the minimum number of vehicles required. The minimum path coverage method divides all nodes in the directed acyclic graph into several paths and satisfies that each node belongs to only one path, and all nodes are covered by the path.

9. A method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network, as described in claim 8, is characterized in that... The method for finding the minimum number of continuous paths using the minimum path cover method is to use the Hopcroft-Karp algorithm with bisection matching to solve the directed acyclic graph.

10. A ride-hailing regional minimum fleet deployment system based on a ride-sharing network, characterized in that, The method for deploying a minimum fleet of ride-hailing vehicles in a specific area based on a ride-sharing network, as described in any one of claims 1 to 9, is implemented using each step of the method.

Images