A customized public transport candidate route generation method, system and device based on hypergraph dynamic modeling and a storage medium

Abstract

The application provides a customized public transport candidate route generation method, system, device and storage medium based on hypergraph dynamic modeling, which adopts a multi-objective collaborative optimization architecture, simultaneously considers path cost, vehicle fixed cost and passenger satisfaction, avoids the deterioration of operation indicators caused by single-objective optimization, and generates a route scheme that is more in line with the commercial operation requirements of customized public transport; secondly, the method realizes unified correlation modeling of demand, road network and environment multidimensional elements through hypergraph dynamic modeling, breaks through the limitation of traditional simple graph twoway correlation, can integrate dynamic constraints such as historical traffic and adverse weather into the route planning process, and generates a route scheme that is more in line with the actual road traffic conditions; the method realizes dynamic allocation of multi-objective optimization weights through a meta-learning network, replaces the traditional artificial experience parameter adjustment mode, and has stronger adaptability in large-scale complex road networks.

Description

Technical Field

[0001] This application relates to the field of intelligent public transportation planning technology, specifically to a method, system, device, and storage medium for generating customized bus candidate routes based on hypergraph dynamic modeling. Background Technology

[0002] With the continuous growth of personalized travel demands from urban residents, customized bus services have become an important supplement to the public transportation system. Route generation for customized buses is a core link connecting user needs with actual operations. Its core objective is to generate bus route plans that balance operating costs, service quality, and traffic feasibility based on aggregated user travel demands.

[0003] Current customized bus route generation technology has the following shortcomings: 1) Existing technologies mostly adopt a single-objective optimization architecture, focusing solely on minimizing the total route distance. This easily leads to problems such as shortened routes but a doubling of transfers and soaring empty-load rates, resulting in a severe deterioration of actual operational indicators and an inability to meet the commercial operation requirements of customized buses. 2) Existing route modeling often uses traditional simple graph structures, only achieving pairwise relationships between nodes. The generated routes only achieve theoretical path optimization, completely detached from actual road conditions and unforeseen environmental constraints, making them unsuitable for practical operation. 3) The weight coefficients of existing optimization models rely entirely on manual experience for configuration, resulting in extremely poor adaptability to large-scale complex road networks. This leads to slow convergence speed, solution quality heavily influenced by initial parameters, and significant fluctuations in optimization performance across different scenarios.

[0004] Therefore, how to design a method, system, device, and storage medium for generating customized bus candidate routes based on hypergraph dynamic modeling that can take into account route distance and commercial needs, and conform to actual road traffic conditions and environmental constraints, is a technical problem that has not yet been solved in the existing technology. Summary of the Invention

[0005] Therefore, this application proposes a method, system, device, and storage medium for generating customized bus candidate routes based on hypergraph dynamic modeling, which can take into account both route distance and commercial needs, and conform to actual road traffic conditions and environmental constraints, so as to improve or solve at least one of the above-mentioned technical problems.

[0006] This application mainly includes the following aspects: In a first aspect, embodiments of this application provide a method for generating customized bus candidate routes based on hypergraph dynamic modeling, the method comprising: Acquire a set of spatiotemporal demand clusters and supporting basic data for route planning, and perform standardized preprocessing on both to construct a customized basic dataset for bus route planning. The set of spatiotemporal demand clusters includes the spatial attributes, temporal attributes, demand intensity attributes, and correlation attributes of each spatiotemporal demand cluster. The supporting basic data for route planning includes urban road network topology data, road construction and closure event data, severe weather data, road segment toll cost data, historical traffic flow data, and bus vehicle operation parameters associated with the spatiotemporal demand clusters. Based on the customized bus route planning dataset, demand nodes, road nodes, environmental event nodes and hyperedge mapping rules are defined to construct an initial hypergraph topology structure, realizing cross-dimensional unified association of demand, road network and environment; By constructing a dynamic weight function for each hyperedge and assigning weights, a weighted dynamic hypergraph structure is generated, enabling a quantitative representation of the priority and passage constraints of different demand clusters. Based on the weighted dynamic hypergraph structure, a two-layer graph convolutional network is constructed and inter-layer convolution operations are performed by inputting the hypergraph adjacency matrix and the node attribute matrix to extract the spatial association, demand intensity and environmental constraint features of the nodes, and generate a set of hypergraph node embedding vectors. Based on the set of hypergraph node embedding vectors, a meta-learning network containing two fully connected layers and a normalized exponential function layer is constructed to complete the nonlinear mapping and normalization of node features to multi-objective weights, and output dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thereby realizing the adaptive dynamic allocation of multi-objective optimization weights. Based on the customized bus route planning dataset, the weighted dynamic hypergraph structure, and the dynamic weight coefficients, a multi-objective optimization overall objective function is constructed and hard constraints are injected to form a complete multi-objective mixed integer programming model. The multi-objective mixed integer programming model is solved using an integer programming solver to obtain multiple initial route schemes. The core operation indicators of each initial route scheme are calculated and compliance verification is completed. After invalid schemes are eliminated, a pool of customized bus candidate route schemes is output.

[0007] According to one embodiment of the present invention, the spatial attributes include: cluster center coordinates, geographic coverage boundary, and a list of associated road nodes; the temporal attributes include: a set of timestamps and time distribution characteristics; the demand intensity attributes include: cluster demand quantity, demand density, and priority weight; and the association attributes include original demand data reference links and environmental event tags.

[0008] According to one embodiment of the present invention, based on the customized bus route planning basic dataset, defining demand nodes, road nodes, environmental event nodes, and hyperedge mapping rules, constructing an initial hypergraph topology, and realizing cross-dimensional unified association of demand, road network, and environment includes the following steps: Define the demand node, the road node, and the environmental event node; wherein, the demand node corresponds to the spatiotemporal demand cluster, the road node corresponds to the intersection and road segment endpoint in the urban road network topology data, and the environmental event node corresponds to the road construction and road closure event data and the event node affecting traffic in the severe weather data; Each demand node is mapped to an independent hyperedge, and each hyperedge connects all the road nodes covered by the virtual site corresponding to the demand node, as well as all the environmental event nodes within the time distribution range of the demand node, thereby realizing a cross-dimensional unified association between demand, road network and environment. Based on the definitions of the demand nodes, road nodes, and environmental event nodes, as well as the hyperedge mapping rules, the initial hypergraph topology is constructed, clarifying the relationships between all nodes and the coverage of hyperedges.

[0009] According to one embodiment of the present invention, the steps of generating a weighted dynamic hypergraph structure by constructing a dynamic weight function for each hyperedge and assigning weights to it, thereby realizing the quantitative representation of the priority and passage constraints of different demand clusters, include: Construct the dynamic weight function, which is as follows: ; in, For the current planning period; The base travel time corresponding to the demand node; This is the time sensitivity coefficient; The cluster demand of the demand node corresponding to the superedge; The average traffic congestion index of the super-edge covered road nodes is obtained through the historical traffic flow data; This is the demand weighting coefficient. It is the traffic weighting coefficient, and and The sum of is 1; Calculate the dynamic weight value of each hyperedge; the higher the dynamic weight value, the higher the priority of the corresponding spatiotemporal demand cluster, the better the passage conditions, and the more likely it is to be covered in the route planning. The dynamic weight values ​​are assigned to the corresponding hyperedges to construct the weighted dynamic hypergraph structure.

[0010] According to one embodiment of the present invention, based on the weighted dynamic hypergraph structure, the step of constructing a two-layer graph convolutional network and performing inter-layer convolution operations by inputting the hypergraph adjacency matrix and node attribute matrix to extract the spatial association, demand intensity, and environmental constraint features of nodes, and generating a set of hypergraph node embedding vectors includes: Extract the numbering order of all nodes and the attribute dimensions of the hypergraph nodes in the weighted dynamic hypergraph structure, and construct a two-layer graph convolutional network; set the input layer dimension of the graph convolutional network to be completely matched with the attribute dimensions of the hypergraph nodes, set the first hidden layer dimension of the graph convolutional network to 64 dimensions, the second output layer dimension to 32 dimensions, and set the activation function of both layers of the graph convolutional network to be a linear rectified function, thereby completing the initial configuration of the two-layer graph convolutional network; Based on the initial hypergraph topology and the weighted dynamic hypergraph structure, the association relationships of all nodes and the dynamic weights of the corresponding hyperedges are statistically analyzed to generate a hypergraph adjacency matrix. Based on the customized bus route planning dataset, all nodes' full-dimensional standardized attributes are extracted and arranged in the same order as the hypergraph adjacency matrix to generate a node attribute matrix. The hypergraph adjacency matrix and the node attribute matrix are input into the initialized two-layer graph convolutional network. Inter-layer convolution is performed in the first graph convolutional layer to fuse the association information between the node's own attributes and its first-order neighboring nodes, generating a 64-dimensional intermediate feature vector. This intermediate feature vector is then input into the second graph convolutional layer to perform inter-layer convolution, fusing the global association information between second-order neighboring nodes, thereby generating a corresponding 32-dimensional node embedding vector for each node. ; All node embedding vectors are arranged in the same numbering order as the hypergraph adjacency matrix and the node attribute matrix. The nodes are arranged in an ordered manner to obtain the set of embedding vectors for the hypergraph nodes.

[0011] According to one embodiment of the present invention, based on the hypergraph node embedding vector set, a meta-learning network comprising two fully connected layers and a normalized exponential function layer is constructed to complete the nonlinear mapping and normalization processing from node features to multi-objective weights, and output dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights. The steps for achieving adaptive dynamic allocation of multi-objective optimization weights include: The meta-learning network is constructed, and the core weight calculation formula of the meta-learning network is as follows: ; in, It is a multilayer perceptron used to implement nonlinear mapping between node features and weights; This is a normalized exponential function used to probabilistically assign initial weights; The set of embedding vectors for the hypergraph nodes is input into the meta-learning network. First, the node features are nonlinearly mapped through two fully connected layers, transforming the high-dimensional node features into preliminary multi-objective weights. Then, the normalized exponential function layer performs normalization processing, outputting the dynamic weight coefficients optimized for multi-objective goals. These dynamic weight coefficients include path cost weights. Vehicle fixed cost weight And passenger satisfaction penalty weighting And the sum of the three is 1.

[0012] According to one embodiment of the present invention, the overall objective function of the multi-objective optimization is: ; in, The set of feasible paths For demand clusters To demand cluster The aforementioned road segment toll cost data, For decision variables, representing whether to choose arrive The path; For vehicle types, For the first The number of vehicles of this type deployed, The first of the bus vehicle operation parameters Fixed cost per vehicle for this type of vehicle; For a set of demand clusters, For the first The actual number of passengers served by each demand cluster The rated passenger capacity of a single vehicle; , , The dynamic weighting coefficient; The hard constraints include: demand coverage constraints, vehicle capacity constraints, total transport capacity constraints, route length constraints, and time window constraints.

[0013] Secondly, embodiments of this application also provide a customized bus candidate route generation system based on hypergraph dynamic modeling, the system comprising: The data acquisition module is used to acquire spatiotemporal demand clusters and basic data supporting route planning, and to perform standardized preprocessing on both to build a basic dataset for customized bus route planning. The Hypergraph Topology module is used to define demand nodes, road nodes, environmental event nodes, and hyperedge mapping rules, construct the initial hypergraph topology structure, and realize cross-dimensional unified association of demand, road network, and environment. The hypergraph weight calculation module is used to construct a dynamic weight function for each hyperedge and complete the weight assignment, generating a weighted dynamic hypergraph structure to realize the quantitative representation of the priority and passage constraints of different requirement clusters. The hypergraph feature extraction module is used to construct a two-layer graph convolutional network and input the hypergraph adjacency matrix and node attribute matrix to perform inter-layer convolution operations to extract the spatial association, demand intensity and environmental constraint features of nodes and generate a set of hypergraph node embedding vectors. The weight allocation module is used to construct a meta-learning network containing two fully connected layers and a normalized exponential function layer. It completes the nonlinear mapping and normalization of node features to multi-objective weights and outputs dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thereby realizing the adaptive dynamic allocation of multi-objective optimization weights. The multi-objective mixed-integer programming module is used to construct the overall objective function for multi-objective optimization and inject hard constraints to form a complete multi-objective mixed-integer programming model. The scheme generation module is used to solve the multi-objective mixed integer programming model using an integer programming solver to obtain multiple initial route schemes, calculate the core operation indicators of each initial route scheme and complete compliance verification, and output a customized bus candidate route scheme pool after eliminating invalid schemes.

[0014] Thirdly, embodiments of this application also provide a computer device, the computer device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the steps of the method described above when executing the instructions.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described above.

[0016] The technical solution provided in this application has the following advantages: 1. The customized bus candidate route generation method based on hypergraph dynamic modeling proposed in this application adopts a multi-objective collaborative optimization architecture, which takes into account path cost, vehicle fixed cost and passenger satisfaction at the same time, avoiding the problem of deterioration of operation indicators caused by single-objective optimization, and the generated route plan is more in line with the needs of commercial operation of customized buses.

[0017] 2. The customized bus candidate route generation method based on hypergraph dynamic modeling in this application realizes the unified association modeling of multiple dimensions of demand, road network and environment through hypergraph dynamic modeling, which breaks through the limitation of pairwise association of traditional simple graphs. It can integrate dynamic constraints such as historical traffic and severe weather into the route planning process, and the generated route plan is more in line with the actual road traffic conditions.

[0018] 3. The customized bus candidate route generation method based on hypergraph dynamic modeling proposed in this application realizes the dynamic allocation of multi-objective optimization weights through meta-learning network, which replaces the traditional manual experience parameter tuning method. It can adjust the priority according to the demand intensity and traffic status of different scenarios, and has stronger adaptability in large-scale complex road networks, thus improving the convergence speed and solution quality of the model.

[0019] 4. The customized bus candidate route generation method based on hypergraph dynamic modeling in this application realizes the quantitative representation of the priority and traffic constraints of different spatiotemporal demand clusters through the weighted dynamic hypergraph structure. It can reflect the time urgency, demand and traffic difficulty of different spatiotemporal demand clusters, and ensure that areas with high timeliness, high demand and good traffic conditions are covered by routes first, thereby improving the allocation efficiency of public transportation resources.

[0020] 5. The method for generating customized bus candidate routes based on hypergraph dynamic modeling in this application obtains a set of hypergraph node embedding vectors through inter-layer convolution operations, automatically capturing implicit relationships in large-scale complex road networks that are difficult to define manually, and significantly improving the comprehensiveness and accuracy of feature representation.

[0021] 6. The customized bus candidate route generation method based on hypergraph dynamic modeling in this application outputs multiple candidate schemes instead of a single optimal scheme, which enables operators to flexibly select the most suitable route scheme according to the operational goals, temporary adjustment needs or resource allocation at different times, significantly improving the system's adaptability in dynamic traffic environments.

[0022] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is the overall flowchart of the customized bus candidate route generation method based on hypergraph dynamic modeling in this application.

[0025] Figure 2 This is the architecture diagram of the customized bus candidate route generation system based on hypergraph dynamic modeling in this application.

[0026] Figure 3 This is a schematic diagram of the user terminal interface for this application.

[0027] Figure 4 This is another screenshot of the user terminal interface for this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0029] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0030] The present application will now be described in detail with reference to the accompanying drawings and embodiments.

[0031] Example like Figure 1 As shown, this embodiment provides a method for generating customized bus candidate routes based on hypergraph dynamic modeling, including the following steps: S101: Obtain the spatiotemporal demand cluster set and the supporting basic data for route planning, and perform standardized preprocessing on both to construct a customized bus route planning basic dataset; wherein, the spatiotemporal demand cluster set includes the spatial attributes, temporal attributes, demand intensity attributes, and correlation attributes of each spatiotemporal demand cluster; the supporting basic data for route planning includes urban road network topology data, road construction and road closure event data, severe weather data, road segment toll cost data, historical traffic flow data, and bus vehicle operation parameters associated with the spatiotemporal demand clusters; S102: Based on the customized bus route planning basic dataset, define demand nodes, road nodes, environmental event nodes and hyperedge mapping rules, construct the initial hypergraph topology, and realize cross-dimensional unified association of demand, road network and environment; S103: By constructing a dynamic weight function for each hyperedge and assigning weights, a weighted dynamic hypergraph structure is generated, enabling a quantitative representation of the priority and passage constraints of different demand clusters. S104: Based on the weighted dynamic hypergraph structure, a two-layer graph convolutional network is constructed and inter-layer convolution operations are performed by inputting the hypergraph adjacency matrix and the node attribute matrix to extract the spatial association, demand intensity and environmental constraint features of the nodes, and generate a set of hypergraph node embedding vectors. S105: Based on the set of embedded vectors of the hypergraph nodes, a meta-learning network containing two fully connected layers and a normalized exponential function layer is constructed to complete the nonlinear mapping and normalization of node features to multi-objective weights, and output dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thereby realizing the adaptive dynamic allocation of multi-objective optimization weights. S106: Based on the customized bus route planning basic dataset, the weighted dynamic hypergraph structure, and the dynamic weight coefficients, construct a multi-objective optimization overall objective function and inject hard constraints to form a complete multi-objective mixed integer programming model; S107: Use an integer programming solver to solve the multi-objective mixed integer programming model to obtain multiple initial route schemes. Calculate the core operation indicators of each initial route scheme and complete compliance verification. After eliminating invalid schemes, output a pool of customized bus candidate route schemes.

[0032] The following details each step of the method for generating customized bus candidate routes based on hypergraph dynamic modeling: S101: Obtain the spatiotemporal demand cluster set and the supporting basic data for route planning, and perform standardized preprocessing on both to construct a customized bus route planning basic dataset; wherein, the spatiotemporal demand cluster set includes the spatial attributes, temporal attributes, demand intensity attributes, and correlation attributes of each spatiotemporal demand cluster; the supporting basic data for route planning includes urban road network topology data, road construction and road closure event data, severe weather data, road segment toll cost data, historical traffic flow data, and bus vehicle operation parameters associated with the spatiotemporal demand clusters; In one possible implementation, the spatial attributes include: cluster center coordinates, geographic coverage boundaries, and a list of associated road nodes; the temporal attributes include: a set of timestamps and time distribution characteristics; the demand intensity attributes include: cluster demand quantity, demand density, and priority weight; and the association attributes include original demand data reference links and environmental event tags.

[0033] Specifically, through user terminals (such as...) Figure 3 and Figure 4 (As shown) Obtain the customized bus travel demand data. Each piece of customized bus travel demand data includes the unique demand identifier, the latitude and longitude coordinates of the boarding point, the latitude and longitude coordinates of the alighting point, the user's expected boarding time window, and the user's expected alighting time window; Obtain the multimodal environment data that matches the customized bus travel demand data. The multimodal environment data includes urban road network topology data, weather warning data, historical traffic status data, and special event identifier data for holidays and school opening seasons. The spatial attributes include: cluster center coordinates, geographic coverage boundary, and a list of associated road nodes. The cluster center coordinates are the average of the latitude and longitude coordinates of all standardized spatiotemporal objects within the cluster. The geographic coverage boundary is calculated by first determining the Euclidean distance from each standardized spatiotemporal object within the cluster to the cluster center coordinates, then taking the maximum distance as the coverage radius. A circular area with the cluster center coordinates as the center and the coverage radius as the radius is the geographic coverage boundary of the cluster. Alternatively, a convex hull algorithm can be used to connect the latitude and longitude coordinates of all standardized spatiotemporal objects within the cluster into a minimum closed polygon, which serves as the precise geographic coverage boundary of the cluster. The list of associated road nodes is formed by matching the original collected urban road network topology data with the cluster center coordinates and geographic coverage boundary, extracting all accessible urban road intersections and road segments within the geographic coverage boundary, and deduplicating them to create the list of associated road nodes for the cluster.

[0034] The time attributes include: a timestamp set and time distribution characteristics. The timestamp set is formed by directly extracting the timestamps of all standardized spatiotemporal objects within the corresponding cluster, removing duplicates, and sorting them chronologically. The timestamps for the boarding point spatiotemporal demand cluster are the midpoint timestamps of the corrected user's expected boarding time window, and the timestamps for the alighting point spatiotemporal demand cluster are the midpoint timestamps of the corrected user's expected alighting time window. The time distribution characteristics are calculated based on the timestamp set and include three core components: first, the timestamp mean, which is the average of all timestamps in the set, representing the core travel time of users in the cluster; second, the timestamp variance, used to measure the dispersion of user travel time within the cluster, with smaller variance indicating more concentrated user travel time; and third, the peak travel time window, which involves frequency statistics on the timestamp set, selecting consecutive time periods where the number of users exceeds 60% of the total number of people in the cluster as the peak travel time window for that cluster.

[0035] The demand intensity attributes include: cluster demand, demand density, and priority weight. Cluster demand directly represents the total number of standardized spatiotemporal objects within a cluster, corresponding to the total number of passengers that the virtual station can serve. Demand density is calculated by dividing the cluster demand by the area of ​​the cluster's geographical coverage boundary, yielding the demand per unit geographical area, which measures the density of demand in the region. Priority weight is calculated based on cluster demand and peak travel time windows, with the cluster demand as the base weight. For example, if the peak travel time window falls during weekday morning and evening rush hours, the weight is multiplied by a coefficient of 1.5; a higher weight indicates a higher priority for the cluster's route coverage.

[0036] The associated attributes include original demand data reference links and environmental event tags. Original demand data reference links: Each standardized spatiotemporal object within the cluster corresponds to a single original customized public transport travel demand data entry through a unique demand identifier. Storage reference links for the corresponding original demand data are bound to all objects within the cluster. These links are then aggregated to form a set of original demand data reference links for the cluster, allowing for the tracing of the complete original information for each demand. Environmental event tags: Multimodal environmental data corresponding to all standardized spatiotemporal objects within the cluster are extracted and aggregated to form environmental event tags for the cluster. These tags include weather, historical traffic conditions, and special events such as holidays and school opening seasons within the corresponding spatiotemporal range.

[0037] For example, the standardized preprocessing includes: using the cluster center of the spatiotemporal demand cluster as a virtual station for route planning, and matching the road network nodes corresponding to the virtual station; and organizing urban road network topology data, road construction and road closure event data, severe weather data, road segment toll cost data, historical traffic flow data, and public transport vehicle operation parameters into standardized parameters.

[0038] For example, the severe weather data refers to all weather phenomena that have a significant impact on road traffic, specifically including: heavy rain, heavy snow, heavy fog, hail, and other weather that reduces road traffic capacity or increases safety risks.

[0039] S102: Based on the customized bus route planning basic dataset, define demand nodes, road nodes, environmental event nodes and hyperedge mapping rules, construct the initial hypergraph topology, and realize cross-dimensional unified association of demand, road network and environment; In one possible implementation, step S102 includes the following steps: Step 1021: Define the demand node, the road node, and the environmental event node; wherein, the demand node corresponds to the spatiotemporal demand cluster, the road node corresponds to the intersection and road segment endpoint in the urban road network topology data, and the environmental event node corresponds to the road construction and road closure event data and the event node affecting traffic in the severe weather data; Step 1022: Map each demand node as an independent hyperedge, and each hyperedge connects all the road nodes covered by the virtual site corresponding to the demand node, as well as all the environmental event nodes within the time distribution range of the demand node, thereby realizing cross-dimensional unified association of demand, road network and environment; Step 1023: Based on the definitions of the demand node, the road node, and the environmental event node, as well as the hyperedge mapping rules, construct the initial hypergraph topology structure and clarify the association relationship of all nodes and the coverage of hyperedges.

[0040] S103: By constructing a dynamic weight function for each hyperedge and assigning weights, a weighted dynamic hypergraph structure is generated, enabling a quantitative representation of the priority and passage constraints of different demand clusters. In one possible implementation, step S103 includes the following steps: Step 1031: Construct the dynamic weight function, wherein the dynamic weight function is: ; in, For the current planning period; The base travel time corresponding to the demand node; This is the time sensitivity coefficient; The cluster demand of the demand node corresponding to the superedge; The average traffic congestion index of the super-edge covered road nodes is obtained through the historical traffic flow data; This is the demand weighting coefficient. It is the traffic weighting coefficient, and and The sum of is 1; Step 1032: Calculate the dynamic weight value of each hyperedge; the higher the dynamic weight value, the higher the priority of the corresponding spatiotemporal demand cluster, the better the passage conditions, and the more likely it is to be covered in the route planning. Step 1033: Assign the dynamic weight values ​​to the corresponding hyperedges to construct the weighted dynamic hypergraph structure.

[0041] The customized bus candidate route generation method based on hypergraph dynamic modeling proposed in this application realizes the quantitative representation of the priority and traffic constraints of different spatiotemporal demand clusters through a weighted dynamic hypergraph structure. It can reflect the time urgency, demand and traffic difficulty of different spatiotemporal demand clusters, ensuring that areas with high timeliness, high demand and good traffic conditions are prioritized for route coverage, thereby improving the allocation efficiency of public transportation resources.

[0042] S104: Based on the weighted dynamic hypergraph structure, a two-layer graph convolutional network is constructed and inter-layer convolution operations are performed by inputting the hypergraph adjacency matrix and the node attribute matrix to extract the spatial association, demand intensity and environmental constraint features of the nodes, and generate a set of hypergraph node embedding vectors. The method for generating customized bus candidate routes based on hypergraph dynamic modeling proposed in this application achieves unified correlation modeling of multiple dimensions of demand, road network, and environment through hypergraph dynamic modeling. It breaks through the limitations of pairwise correlation of traditional simple graphs and can integrate dynamic constraints such as historical traffic and severe weather into the route planning process, resulting in route schemes that are more in line with actual road traffic conditions.

[0043] In one possible implementation, step S104 includes the following steps: Step 1041: Extract the numbering order of all nodes and the attribute dimension of the hypergraph nodes in the weighted dynamic hypergraph structure, and construct a two-layer graph convolutional network; set the input layer dimension of the graph convolutional network to be completely matched with the attribute dimension of the hypergraph nodes, set the first hidden layer dimension of the graph convolutional network to 64 dimensions, the second output layer dimension to 32 dimensions, and set the activation function of both layers of the graph convolutional network to be a linear rectified function, thereby completing the initial configuration of the two-layer graph convolutional network; Step 1042: Based on the initial hypergraph topology and the weighted dynamic hypergraph structure, calculate the association relationships of all nodes and the dynamic weights of the corresponding hyperedges, and generate a hypergraph adjacency matrix; Step 1043: Based on the customized bus route planning basic dataset, extract the full-dimensional standardized attributes of all nodes, arrange them in the same node order as the hypergraph adjacency matrix, and generate a node attribute matrix; Step 1044: Input the hypergraph adjacency matrix and the node attribute matrix into the initialized two-layer graph convolutional network. Perform inter-layer convolution operation through the first graph convolutional layer to fuse the association information between the node's own attributes and first-order neighbor nodes, generating a 64-dimensional intermediate feature vector. Input the intermediate feature vector into the second graph convolutional layer to perform inter-layer convolution operation to fuse the global association information between second-order neighbor nodes, thereby generating a corresponding 32-dimensional node embedding vector for each node. ; Step 1045: Embed all the nodes into vectors according to the numbering order consistent with the hypergraph adjacency matrix and the node attribute matrix. The nodes are arranged in an ordered manner to obtain the set of embedding vectors for the hypergraph nodes.

[0044] The proposed method for generating customized bus candidate routes based on hypergraph dynamic modeling obtains a set of hypergraph node embedding vectors through inter-layer convolution operations, automatically capturing implicit relationships in large-scale complex road networks that are difficult to define manually, and significantly improving the comprehensiveness and accuracy of feature representation.

[0045] S105: Based on the set of embedded vectors of the hypergraph nodes, a meta-learning network containing two fully connected layers and a normalized exponential function layer is constructed to complete the nonlinear mapping and normalization of node features to multi-objective weights, and output dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thereby realizing the adaptive dynamic allocation of multi-objective optimization weights. In one possible implementation, step S105 includes the following steps: Step 1051: Construct the meta-learning network. The core weight calculation formula of the meta-learning network is as follows: ; in, It is a multilayer perceptron used to implement nonlinear mapping between node features and weights; This is a normalized exponential function used to probabilistically assign initial weights; Step 1052: Input the set of embedding vectors of the hypergraph nodes into the meta-learning network, and first complete the nonlinear mapping of node features through the two fully connected layers to transform the high-dimensional node features into multi-objective preliminary weights; Step 1053: The normalization process is then performed through the normalized exponential function layer to output the dynamic weight coefficients for multi-objective optimization; the dynamic weight coefficients include path cost weights. Vehicle fixed cost weight And passenger satisfaction penalty weighting And the sum of the three is 1.

[0046] Among them, path cost weight Prioritize the economic efficiency of control paths. The higher the weight, the more the algorithm tends to choose routes with low fuel consumption and short distances; vehicle fixed cost weight. Prioritize controlling vehicle fixed costs. The higher the value, the more the algorithm tends to reduce the total number of vehicles; passenger satisfaction penalty weight. Prioritize passenger satisfaction. The higher the value, the more the algorithm tends to avoid overcrowding and improve passenger comfort.

[0047] When road operating costs change, such as due to rising oil prices, the route cost weight is automatically increased. Prioritize controlling the total operating cost of the route; when the cost of purchasing, leasing, or maintaining buses increases, the weight of fixed costs for vehicles will be automatically increased. Prioritize reducing the total number of vehicles deployed and lowering vehicle operation costs; when operating subsidies increase and overall cost pressure decreases, automatically increase the penalty weight for passenger satisfaction. Prioritize improving service quality; at the same time, avoid extreme weighting of a single objective by using a normalized exponential function to ensure a balanced adaptation of multiple objectives.

[0048] The proposed method for generating candidate routes for customized buses based on hypergraph dynamic modeling adopts a multi-objective collaborative optimization architecture, which takes into account path cost, vehicle fixed cost and passenger satisfaction. This avoids the problem of deterioration of operating indicators caused by single-objective optimization, and the generated route plan is more in line with the needs of commercial operation of customized buses.

[0049] The proposed method for generating customized bus candidate routes based on hypergraph dynamic modeling in this application achieves dynamic allocation of multi-objective optimization weights through meta-learning networks, replacing the traditional manual experience-based parameter tuning method. It can adjust priorities according to the demand intensity and traffic conditions of different scenarios, and has stronger adaptability in large-scale complex road networks, improving the convergence speed and solution quality of the model.

[0050] S106: Based on the customized bus route planning basic dataset, the weighted dynamic hypergraph structure, and the dynamic weight coefficients, construct a multi-objective optimization overall objective function and inject hard constraints to form a complete multi-objective mixed integer programming model; In one possible implementation, the overall objective function of the multi-objective optimization is: ; in, The set of feasible paths For demand clusters To demand cluster The aforementioned road segment toll cost data, For decision variables, representing whether to choose arrive The path; For vehicle types, For the first The number of vehicles of this type deployed, The first of the bus vehicle operation parameters Fixed cost per vehicle for this type of vehicle; For a set of demand clusters, For the first The actual number of passengers served by each demand cluster The rated passenger capacity of a single vehicle; , , The dynamic weighting coefficient; The hard constraints include: demand coverage constraints, vehicle capacity constraints, total transport capacity constraints, route length constraints, and time window constraints.

[0051] Specifically, the constraints are as follows: Demand Coverage Constraint: All generated routes must cover at least 80% of high-priority demand clusters, which are demand nodes with a superedge weight greater than 0.5; Vehicle Capacity Constraint: The total number of passengers on a single route must not exceed the total rated passenger capacity of the deployed vehicles, and overloading is prohibited; Total Capacity Constraint: The total number of vehicles deployed on all routes must not exceed the preset maximum number of vehicles; Route Length Constraint: The total mileage of a single route must not exceed the preset maximum route mileage threshold; Time Window Constraint: The total travel time of a single route must match the time distribution window of the corresponding demand cluster and must not exceed the user's expected travel time window.

[0052] Based on a customized bus route planning dataset, a weighted dynamic hypergraph structure, and dynamic weight coefficients, this application constructs a multi-objective optimization overall objective function and injects hard constraints such as demand coverage constraints, vehicle capacity constraints, total transport capacity constraints, route length constraints, and time window constraints to form a complete multi-objective mixed integer programming model. The model defines the boundaries of the solution from multiple dimensions such as service scope, operational safety, resource constraints, operational efficiency, and passenger experience, ensuring the compliance of the generated solution.

[0053] S107: Use an integer programming solver to solve the multi-objective mixed integer programming model to obtain multiple initial route schemes. Calculate the core operation indicators of each initial route scheme and complete compliance verification. After eliminating invalid schemes, output a pool of customized bus candidate route schemes.

[0054] For example, the Gurobi integer programming solver is used to solve the model, with the upper limit of the solution iterations set to 1000 times and the solution accuracy threshold set to 0.01. After solving the problem, we obtain the solution set of the model. Each solution corresponds to a complete route plan, which includes the route direction, the stopping order of demand clusters, and the number of vehicles deployed on the route. Calculate the core operational indicators for each initial route plan, including total route mileage, estimated total operating cost, average route load factor, demand coverage, and total travel time. Link the operational indicators to the route plan. Perform compliance verification, eliminate invalid plans, and output a pool of customized bus candidate route plans.

[0055] The proposed method for generating customized bus candidate routes based on hypergraph dynamic modeling outputs multiple candidate schemes instead of a single optimal scheme, enabling operators to flexibly select the most suitable route scheme according to different operational goals, temporary adjustment needs, or resource allocation, thus significantly improving the system's adaptability in dynamic traffic environments.

[0056] like Figure 2 As shown, based on the same application concept, this application embodiment also provides a customized bus candidate route generation system based on hypergraph dynamic modeling, including: a data acquisition module, used to acquire a set of spatiotemporal demand clusters and basic data supporting route planning, and to perform standardized preprocessing on both to construct a basic dataset for customized bus route planning; a hypergraph topology module, used to define demand nodes, road nodes, environmental event nodes and hyperedge mapping rules, construct an initial hypergraph topology structure, and realize cross-dimensional unified association of demand, road network and environment; a hypergraph weight calculation module, used to construct a dynamic weight function for each hyperedge and complete weight assignment, generating a weighted dynamic hypergraph structure, and realizing the quantitative representation of the priority and traffic constraints of different demand clusters; and a hypergraph feature extraction module, used to construct a two-layer graph convolutional network and input the hypergraph adjacency matrix and node attribute matrix to perform inter-layer convolution operations to extract the spatial features of nodes. The system generates a set of hypergraph node embedding vectors based on association, demand intensity, and environmental constraints. A weight allocation module constructs a meta-learning network consisting of two fully connected layers and a normalized exponential function layer. This network performs nonlinear mapping and normalization of node features to multi-objective weights, outputting dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thus achieving adaptive dynamic allocation of multi-objective optimization weights. A multi-objective mixed-integer programming module constructs the overall multi-objective optimization objective function and injects hard constraints to form a complete multi-objective mixed-integer programming model. A scheme generation module solves the multi-objective mixed-integer programming model using an integer programming solver, obtaining multiple initial route schemes. It calculates the core operational indicators for each initial route scheme, performs compliance checks, and outputs a pool of customized bus candidate route schemes after eliminating invalid schemes.

[0057] Based on the same concept, embodiments of this application also provide a computer device, including at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by a processor, the instructions being executed by the at least one processor to cause the at least one processor to implement the above-described method when executing the instructions.

[0058] Based on the same concept, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method.

[0059] In this embodiment, the computer program, when run by the processor, can also execute other machine-readable instructions to perform other methods as described in the embodiments. For details on the specific execution steps and principles, please refer to the description of the embodiments, which will not be repeated here.

[0060] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0061] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0062] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0063] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0064] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0065] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for generating customized bus candidate routes based on hypergraph dynamic modeling, characterized in that, include: Acquire a set of spatiotemporal demand clusters and supporting basic data for route planning, and perform standardized preprocessing on both to construct a customized basic dataset for bus route planning. The set of spatiotemporal demand clusters includes the spatial attributes, temporal attributes, demand intensity attributes, and correlation attributes of each spatiotemporal demand cluster. The supporting basic data for route planning includes urban road network topology data, road construction and closure event data, severe weather data, road segment toll cost data, historical traffic flow data, and bus vehicle operation parameters associated with the spatiotemporal demand clusters. Based on the customized bus route planning dataset, demand nodes, road nodes, environmental event nodes and hyperedge mapping rules are defined to construct an initial hypergraph topology structure, realizing cross-dimensional unified association of demand, road network and environment; By constructing a dynamic weight function for each hyperedge and assigning weights, a weighted dynamic hypergraph structure is generated, enabling a quantitative representation of the priority and passage constraints of different demand clusters. Based on the weighted dynamic hypergraph structure, a two-layer graph convolutional network is constructed and inter-layer convolution operations are performed by inputting the hypergraph adjacency matrix and the node attribute matrix to extract the spatial association, demand intensity and environmental constraint features of the nodes, and generate a set of hypergraph node embedding vectors. Based on the set of hypergraph node embedding vectors, a meta-learning network containing two fully connected layers and a normalized exponential function layer is constructed to complete the nonlinear mapping and normalization of node features to multi-objective weights, and output dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thereby realizing the adaptive dynamic allocation of multi-objective optimization weights. Based on the customized bus route planning dataset, the weighted dynamic hypergraph structure, and the dynamic weight coefficients, a multi-objective optimization overall objective function is constructed and hard constraints are injected to form a complete multi-objective mixed integer programming model. The multi-objective mixed integer programming model is solved using an integer programming solver to obtain multiple initial route schemes. The core operation indicators of each initial route scheme are calculated and compliance verification is completed. After invalid schemes are eliminated, a pool of customized bus candidate route schemes is output.

2. The method for generating customized bus candidate routes based on hypergraph dynamic modeling according to claim 1, characterized in that, The spatial attributes include: cluster center coordinates, geographic coverage boundary, and a list of associated road nodes; the temporal attributes include: a set of timestamps and time distribution characteristics; the demand intensity attributes include: cluster demand quantity, demand density, and priority weight; and the association attributes include original demand data reference links and environmental event tags.

3. The method for generating customized bus candidate routes based on hypergraph dynamic modeling according to claim 2, characterized in that, Based on the aforementioned customized bus route planning dataset, the steps for defining demand nodes, road nodes, environmental event nodes, and hyperedge mapping rules, constructing an initial hypergraph topology, and achieving cross-dimensional unified association between demand, road network, and environment include: Define the demand node, the road node, and the environmental event node; wherein, the demand node corresponds to the spatiotemporal demand cluster, the road node corresponds to the intersection and road segment endpoint in the urban road network topology data, and the environmental event node corresponds to the road construction and road closure event data and the event node affecting traffic in the severe weather data; Each demand node is mapped to an independent hyperedge, and each hyperedge connects all the road nodes covered by the virtual site corresponding to the demand node, as well as all the environmental event nodes within the time distribution range of the demand node, thereby realizing a cross-dimensional unified association between demand, road network and environment. Based on the definitions of the demand nodes, road nodes, and environmental event nodes, as well as the hyperedge mapping rules, the initial hypergraph topology is constructed, clarifying the relationships between all nodes and the coverage of hyperedges.

4. The method for generating customized bus candidate routes based on hypergraph dynamic modeling according to claim 3, characterized in that, By constructing a dynamic weight function for each hyperedge and assigning weights, a weighted dynamic hypergraph structure is generated. The steps to achieve a quantitative representation of the priority and access constraints of different demand clusters include: Construct the dynamic weight function, which is as follows: ； in, For the current planning period; The base travel time corresponding to the demand node; This is the time sensitivity coefficient; The cluster demand of the demand node corresponding to the superedge; The average traffic congestion index of the super-edge covered road nodes is obtained through the historical traffic flow data; This is the demand weighting coefficient. It is the traffic weighting coefficient, and and The sum of is 1; Calculate the dynamic weight value of each hyperedge; the higher the dynamic weight value, the higher the priority of the corresponding spatiotemporal demand cluster, the better the passage conditions, and the more likely it is to be covered in the route planning. The dynamic weight values ​​are assigned to the corresponding hyperedges to construct the weighted dynamic hypergraph structure.

5. The method for generating customized bus candidate routes based on hypergraph dynamic modeling according to claim 4, characterized in that, Based on the weighted dynamic hypergraph structure, the steps of constructing a two-layer graph convolutional network and performing inter-layer convolution operations by inputting the hypergraph adjacency matrix and node attribute matrix to extract the spatial association, demand intensity, and environmental constraint features of nodes, and generating a set of hypergraph node embedding vectors include: Extract the numbering order of all nodes and the attribute dimensions of the hypergraph nodes in the weighted dynamic hypergraph structure, and construct a two-layer graph convolutional network; set the input layer dimension of the graph convolutional network to be completely matched with the attribute dimensions of the hypergraph nodes, set the first hidden layer dimension of the graph convolutional network to 64 dimensions, the second output layer dimension to 32 dimensions, and set the activation function of both layers of the graph convolutional network to be a linear rectified function, thereby completing the initial configuration of the two-layer graph convolutional network; Based on the initial hypergraph topology and the weighted dynamic hypergraph structure, the association relationships of all nodes and the dynamic weights of the corresponding hyperedges are statistically analyzed to generate a hypergraph adjacency matrix. Based on the customized bus route planning dataset, all nodes' full-dimensional standardized attributes are extracted and arranged in the same order as the hypergraph adjacency matrix to generate a node attribute matrix. The hypergraph adjacency matrix and the node attribute matrix are input into the initialized two-layer graph convolutional network. Inter-layer convolution is performed in the first graph convolutional layer to fuse the association information between the node's own attributes and its first-order neighboring nodes, generating a 64-dimensional intermediate feature vector. This intermediate feature vector is then input into the second graph convolutional layer to perform inter-layer convolution, fusing the global association information between second-order neighboring nodes, thereby generating a corresponding 32-dimensional node embedding vector for each node. ; All node embedding vectors are arranged in the same numbering order as the hypergraph adjacency matrix and the node attribute matrix. The nodes are arranged in an ordered manner to obtain the set of embedding vectors for the hypergraph nodes.

6. The method for generating customized bus candidate routes based on hypergraph dynamic modeling according to claim 5, characterized in that, Based on the hypergraph node embedding vector set, a meta-learning network is constructed, consisting of two fully connected layers and a normalized exponential function layer. This network completes the nonlinear mapping and normalization of node features to multi-objective weights, outputting dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights. The steps for adaptive dynamic allocation of multi-objective optimization weights include: The meta-learning network is constructed, and the core weight calculation formula of the meta-learning network is as follows: ； in, It is a multilayer perceptron used to implement nonlinear mapping between node features and weights; This is a normalized exponential function used to probabilistically assign initial weights; The set of embedding vectors for the hypergraph nodes is input into the meta-learning network. First, the node features are nonlinearly mapped through two fully connected layers, transforming the high-dimensional node features into preliminary multi-objective weights. Then, the normalized exponential function layer performs normalization processing, outputting the dynamic weight coefficients optimized for multi-objective goals. These dynamic weight coefficients include path cost weights. Vehicle fixed cost weight And passenger satisfaction penalty weighting And the sum of the three is 1.

7. The method for generating customized bus candidate routes based on hypergraph dynamic modeling according to claim 6, characterized in that: The overall objective function for the multi-objective optimization is: ； in, The set of feasible paths For demand clusters To demand cluster The aforementioned road segment toll cost data, For decision variables, representing whether to choose arrive The path; For vehicle types, For the first The number of vehicles of this type deployed, The first of the bus vehicle operation parameters Fixed cost per vehicle for this type of vehicle; For a set of demand clusters, For the first The actual number of passengers served by each demand cluster The rated passenger capacity of a single vehicle; , , The dynamic weighting coefficient; The hard constraints include: demand coverage constraints, vehicle capacity constraints, total transport capacity constraints, route length constraints, and time window constraints.

8. A customized bus candidate route generation system based on hypergraph dynamic modeling, characterized in that, include: The data acquisition module is used to acquire spatiotemporal demand clusters and basic data supporting route planning, and to perform standardized preprocessing on both to build a basic dataset for customized bus route planning. The Hypergraph Topology module is used to define demand nodes, road nodes, environmental event nodes, and hyperedge mapping rules, construct the initial hypergraph topology structure, and realize cross-dimensional unified association of demand, road network, and environment. The hypergraph weight calculation module is used to construct a dynamic weight function for each hyperedge and complete the weight assignment, generating a weighted dynamic hypergraph structure to realize the quantitative representation of the priority and passage constraints of different requirement clusters. The hypergraph feature extraction module is used to construct a two-layer graph convolutional network and input the hypergraph adjacency matrix and node attribute matrix to perform inter-layer convolution operations to extract the spatial association, demand intensity and environmental constraint features of nodes and generate a set of hypergraph node embedding vectors. The weight allocation module is used to construct a meta-learning network containing two fully connected layers and a normalized exponential function layer. It completes the nonlinear mapping and normalization of node features to multi-objective weights and outputs dynamic weight coefficients including path cost weights, vehicle fixed cost weights, and passenger satisfaction penalty weights, thereby realizing the adaptive dynamic allocation of multi-objective optimization weights. The multi-objective mixed-integer programming module is used to construct the overall objective function for multi-objective optimization and inject hard constraints to form a complete multi-objective mixed-integer programming model. The scheme generation module is used to solve the multi-objective mixed integer programming model using an integer programming solver to obtain multiple initial route schemes, calculate the core operation indicators of each initial route scheme and complete compliance verification, and output a customized bus candidate route scheme pool after eliminating invalid schemes.

9. A computer device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the steps of the method according to any one of claims 1 to 7 when executing the instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images