Supersaturated subway network connection bus service design and vehicle scheduling optimization method

By building a spatio-temporal network and optimization algorithms, designing bus service and vehicle scheduling solutions, the problems of congestion during peak subways and low utilization rate of bus lanes are solved, and passenger travel efficiency is improved and operating costs are reduced.

CN119067275APending Publication Date: 2024-12-03BEIJING JIAOTONG UNIV
View PDF 0 Cites 5 Cited by

Patent Information

Application Number
CN202410967328.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The urban subway system is congested due to oversaturation during peak hours, which affects passenger travel experience. At the same time, the utilization rate of bus lanes is low, resulting in low operational efficiency, making it difficult to effectively alleviate the contradiction between subway capacity supply and passenger demand.

Method used

By building a passenger-distribution space-time network and a bus vehicle scheduling space-time network, combining Lagrangian slack and column generation algorithm, a collaborative optimization model for connecting bus services and vehicle scheduling is designed to generate the best bus services and vehicle scheduling solutions, and optimize subway passenger flow Control strategies and bus shuttle services to achieve the optimal matching of time-varying passenger flow needs.

Benefits of technology

Effectively alleviate congestion in the subway system, shorten the total travel time of passengers, improve the service level and operational efficiency of the bus system, reduce operating costs, and improve passenger satisfaction and overall performance of the public transportation system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067275A_ABST
    Figure CN119067275A_ABST
Patent Text Reader

Abstract

The invention provides a supersaturated subway network connection bus service design and vehicle scheduling optimization method. The method comprises the following steps: respectively constructing a passenger distribution space-time network and a bus scheduling space-time network; according to passenger flow distribution and road conditions, alternative connection routes, feasible time headway and time table adjustment parameters are set, and a service arc in a passenger distribution space-time network, a task arc in a bus scheduling space-time network and an initial scheduling scheme set are generated; establishing a connection bus service design and vehicle scheduling collaborative optimization model; decomposing the connection bus service design and vehicle scheduling collaborative optimization model through a Lagrange relaxation technology, and constructing a series of pricing sub-problems based on a column generation algorithm to dynamically generate a bus service and vehicle scheduling scheme; and according to a calculation result, determining an optimal bus service driving scheme, a vehicle scheduling scheme and a refined passenger flow control strategy. According to the method, the subway passenger flow control strategy and the bus connection service can be optimized at the same time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of urban public transportation organization, and in particular to an oversaturated subway network shuttle bus service design and vehicle dispatch optimization method. Background Art

[0002] Urban public transportation plays an important role in residents' daily commuting. Continuously improving the public transportation system will not only help optimize the urban spatial layout and alleviate urban traffic congestion, but also effectively improve the quality of life of residents, and become a key way to promote the construction of green smart cities and promote sustainable urban development. As the backbone of the public transportation system, urban rail transit has been favored by passengers for its high punctuality, safety, reliability and strong carrying capacity. However, with the increasing demand of passengers, the subway system is also facing serious oversaturation problems. In order to cope with the extreme congestion of subway platforms during peak hours in the morning and evening, the subway operation management department usually takes passenger flow control measures at crowded stations, and by limiting the number of people entering the station, the passenger gathering level on the platform is maintained within a safe range to ensure operational safety. According to the Beijing Transportation Development Annual Report, in 2020, 96 stations in the entire Beijing rail transit network implemented normalized flow control measures, accounting for 21.3% of the total number of subway network stations. Although the flow control measures have alleviated the congestion of the subway to a certain extent, they have also brought many inconveniences to some passengers who were forced to stay outside the station, seriously affecting their travel experience. Therefore, during subway peak hours, how to organize efficient transportation to alleviate the contradiction between subway capacity supply and passenger demand, and provide better service guarantees for passengers while ensuring operational safety has become a difficult problem that the subway system needs to solve urgently.

[0003] On the other hand, as another important part of the urban public transportation system, buses, with their flexible and maneuverable characteristics, can cover areas that are difficult to serve by the subway network, providing residents with a variety of travel options. Under the active promotion of the bus priority policy, many large cities have effectively reduced the impact of road congestion on buses by setting up bus lanes, thereby ensuring the efficient operation of the bus system. Taking Beijing as an example, since the first bus lane in the country was opened on Chang'an Avenue in Beijing, the construction of bus lanes in Beijing has continued to be vigorously promoted, and the total length has now exceeded 1,000 kilometers. With the gradual implementation of the bus priority strategy, the overall level of bus services has been significantly improved. However, with the adjustment of urban planning, the increase in the number of private cars, and the emergence of various travel modes such as rail transit, online car-hailing, and shared bicycles, the bus lanes in some areas have a low utilization rate, which has caused a certain degree of inconvenience to social vehicles, and it is urgent to improve their operating efficiency through strengthened management. Summary of the invention

[0004] In view of this, the present invention provides a method for designing bus service for an oversaturated subway network and optimizing vehicle scheduling to solve the above problems.

[0005] The present invention provides a method for designing a shuttle bus service for an oversaturated subway network and optimizing vehicle scheduling, comprising: constructing a passenger allocation space-time network based on subway line network data, train timetable data, and operation data of ground buses between stations, and constructing a bus vehicle scheduling space-time network based on bus station locations and bus scheduling related data; setting alternative connecting routes, feasible headways, and schedule adjustment parameters according to passenger flow distribution and road conditions, and generating service arcs in the passenger allocation space-time network and task arcs in the bus vehicle scheduling space-time network and an initial scheduling solution set; constructing a passenger allocation space-time network based on the passenger allocation space-time network and the bus vehicle scheduling space-time network; and A collaborative optimization model of shuttle bus service design and vehicle scheduling considering the subway passenger flow control strategy is established; a heuristic solution algorithm based on the combination of Lagrangian relaxation and column generation is adopted, and the collaborative optimization model of shuttle bus service design and vehicle scheduling is decomposed with the help of Lagrangian relaxation technology. At the same time, based on the column generation algorithm framework, a series of pricing sub-problems are constructed to dynamically generate bus service plans and vehicle scheduling plans; according to the calculation results, the optimal bus service operation plan, vehicle scheduling plan and refined passenger flow control strategy are determined, and the real-time passenger flow distribution is visualized to assist operation managers in making dynamic decisions.

[0006] The oversaturated subway network shuttle bus service design and vehicle scheduling optimization method of the present invention takes into account the time-varying passenger flow demand, based on the modeling method of the space-time network, and takes the minimization of passenger travel cost and bus operation cost as the optimization goal, and establishes an integer linear programming model. The model can simultaneously optimize the subway passenger flow control strategy and bus shuttle service; at the same time, it also explores an integrated optimization method, which comprehensively considers the decision-making problems at different stages in the bus operation planning process, so as to achieve the optimal trade-off between service level and operation cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. By reading the detailed description of the following implementation, the advantages and benefits of the solutions become clear to those skilled in the art. The drawings are only used to illustrate the preferred implementation and are not considered to be limitations of the present invention. In the drawings:

[0008] Figure 1 The present invention is a flowchart of a method for designing a shuttle bus service for an oversaturated subway network and optimizing vehicle scheduling according to an embodiment of the present invention.

[0009] Figure 2 A schematic diagram of a rail transit network according to an embodiment of the present invention.

[0010] Figure 3 Schematic diagram of passenger space-time network according to an embodiment of the present invention.

[0011] Figure 4 This is a schematic diagram of the distribution results of passengers in the space-time network without considering connecting buses according to an embodiment of the present invention.

[0012] Figure 5 This is a schematic diagram of the distribution results of passengers in a space-time network when connecting buses according to an embodiment of the present invention.

[0013] Figure 6 This is a schematic diagram of a public transportation space-time network according to an embodiment of the present invention.

[0014] Figure 7 FIG. 1 is a schematic diagram of a service arc in a passenger network according to an embodiment of the present invention.

[0015] Figure 8 A schematic diagram of a pricing network for a service solution according to an embodiment of the present invention.

[0016] Fig. 9 A schematic diagram of a pricing subnetwork for a service solution according to an embodiment of the present invention.

[0017] Fig.10 Schematic diagram of time-varying passenger flow demand of subway line 1 according to an embodiment of the present invention.

[0018] Fig.11 Schematic diagram of time-varying passenger flow demand of subway line 2 according to an embodiment of the present invention.

[0019] Fig.12 This is a schematic diagram of passenger flow distribution of subway line 1 without considering connecting buses according to an embodiment of the present invention.

[0020] Fig.13 This is a schematic diagram of passenger flow distribution of subway line 2 without considering connecting buses according to an embodiment of the present invention.

[0021] Fig.14 This is a schematic diagram of passenger flow distribution of subway line 1 when connecting to a bus according to an embodiment of the present invention.

[0022] Fig.15 This is a schematic diagram of passenger flow distribution of subway line 2 when connecting to a bus according to an embodiment of the present invention.

[0023] Fig.16 The figure is a schematic diagram of a vehicle dispatching scheme under an unbalanced timetable according to an embodiment of the present invention.

[0024] Fig.17The figure is a schematic diagram of a vehicle dispatching scheme under a balanced timetable according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and in detail below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in the field based on the embodiments in the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0026] Figure 1 A schematic flow chart of a method for designing a bus service connecting an oversaturated subway network and optimizing vehicle scheduling is provided in an embodiment of the present invention. Figure 1 As shown, this embodiment mainly includes:

[0027] S101, constructing a passenger allocation spatiotemporal network based on subway line network data, train timetable data, and inter-station ground bus operation data, and constructing a bus vehicle dispatch spatiotemporal network based on bus station locations and bus dispatch related data;

[0028] For example, Figure 2 A small rail transit network with 2 lines and 7 stations is given, where Station 3 and Station 6 are the same physical transfer stations belonging to different lines. When the subway system is severely congested, a feeder bus service is introduced to determine the optimal feeder route (such as candidate routes 1 and 2), and dispatch vehicles from nearby parking lots (such as parking lots 1 and 2) to perform transportation tasks.

[0029] Figure 3 The spatiotemporal network of passenger allocation based on the travel demand from station 2 to station 7 is shown. The spatiotemporal network can be represented as a directed acyclic graph. in Represents a collection of nodes, In order to describe the changes in time-distance-mode of passengers during their entire journey from the initial station to the final station, we first divide the entire time axis into The time is discretized into a series of time points, that is, definition represents the travel mode set, where 1 and 2 represent the subway and shuttle bus respectively. At the subway level, consider the queue area outside the station and the waiting area on the platform of all stations, respectively. and Represented by, and the same physical transfer stations belonging to different lines are regarded as different nodes, then It can represent the set of all locations in the subway layer, and then expand into a series of spatiotemporal nodes according to discrete time points. In addition, in order to ensure that the shuttle bus service can effectively alleviate the passenger congestion of the subway, the present invention assumes that the bus station is set near the subway station. Therefore, the spatiotemporal nodes of the bus layer correspond to the nodes of the queuing area outside the station on the subway layer.

[0030] Figure 6 The spatiotemporal network of bus scheduling constructed based on the direct bus connecting route from Station 2 to Station 7 is shown. In this network, the node set covers the spatiotemporal locations corresponding to the depot and the starting and ending stations of all candidate connecting routes (i.e., Station 2 and Station 7).

[0031] With the help of spatiotemporal network modeling technology, a passenger assignment and vehicle scheduling problem model was constructed, and parallel computing technology was introduced to significantly improve the solution efficiency of the model.

[0032] S102. According to passenger flow distribution and road conditions, alternative connecting routes, feasible headways and schedule adjustment parameters are set, and service arcs in the passenger allocation space-time network and task arcs in the bus dispatch space-time network and a set of initial dispatch plans are generated.

[0033] For example, the node set of the passenger allocation spatiotemporal network can be defined as The arc from node (i, t, m) to node (j, t′, m′) can be defined as a six-tuple (i, j, t, t′, m, m′). It is divided into 8 categories, including: queuing arc Pitstop Arc Waiting Arc Metro Service Arc Subway transfer arc Bus Service Arc Mode conversion arc and virtual arc Specifically, after arriving at the subway station, passengers may need to queue outside the station due to passenger flow control measures. This queuing activity can be represented by a queuing arc that connects the spatiotemporal nodes of adjacent times in each queuing area outside the station on the subway layer. Subsequently, the process of passengers entering the platform from outside the station to wait for the train is represented by an entry arc that connects the nodes of the queuing area outside the station and the nodes of the platform waiting area. The waiting activities of passengers before leaving by boarding the train are characterized by waiting arcs that connect adjacent time nodes on the subway platform. In order to ensure operational safety, the number of waiting passengers on the platform needs to be controlled within a safety threshold, so the capacity of the waiting arc is set to the number of passengers that the platform can accommodate, C platform The subway service arc can be constructed based on a pre-given train schedule (Assumption 1), and the corresponding arc capacity is set to the number of passengers C that the train can accommodate. trainConsidering the existence of multiple subway lines, subway transfer arcs are used to connect the same physical transfer stations belonging to different lines.

[0034] Similarly, the bus service arc follows the travel trajectory of the bus and is closely related to the bus connection routes and schedules. is the set of all candidate connecting routes, where each route refers to a specific driving path with a congested subway station as a bus stop. These routes can be constructed based on the actual road conditions (such as the layout of bus lanes) and the operation of the subway system (such as the spatiotemporal distribution of passenger flow demand). To represent each bus connection route The applicable average headway set, combined with the given feeder bus operation period, can be used to derive the number of buses on the route during the operation period. In order to determine the specific spatiotemporal service arc, it is also necessary to clarify the arrival and departure times of each bus on the route at the stations along the way, that is, the bus schedule. Generally speaking, schedules are mainly divided into two types: balanced schedules and unbalanced schedules. A balanced schedule plan can be derived by strictly following the pre-specified departure intervals. In contrast, by adjusting the arrival and departure times of vehicles at each station, an unbalanced schedule plan that is more in line with the distribution of passenger flow demand can be formulated, thereby improving service operation efficiency. When generating bus service plans, the spatiotemporal distribution characteristics of passenger flow demand are fully considered, and an unbalanced schedule compilation strategy is adopted to ensure that bus services can accurately match the travel needs of passengers, thereby significantly improving the service level and passenger satisfaction of the bus system.

[0035] Preferably, the arc set of the bus scheduling space-time network includes task arcs, exit arcs, entry arcs, empty arcs and waiting arcs. Each task arc represents a complete transportation service. Once the passenger's travel trajectory involves the bus service arc in the passenger network, it is necessary to dispatch a vehicle to perform the corresponding transportation task to ensure the feasibility of the service arc. When generating a vehicle scheduling plan, a flexible cross-line vehicle scheduling strategy is adopted. By rationally allocating vehicle resources, the bus operating cost is effectively reduced and the economic benefits of the public transportation system are improved.

[0036] The key technology of the present invention is to convert passenger travel trajectories and vehicle driving trajectories into space-time trajectories by means of space-time network modeling technology, and then convert the passenger allocation problem and vehicle scheduling problem in the network into a path selection problem in the space-time network, and further use the Lagrangian relaxation method to convert the complex original problem into a series of easy-to-solve shortest path problems. In a space-time network, each node represents a spatial position at a specific time. In order to accurately describe the travel trajectory of passengers in a multimodal transportation network, the present invention specifically introduces a mode dimension to identify the specific travel mode chosen by the passenger. In view of this, each arc in the constructed space-time network corresponds to a state transition including location, time and transportation mode.

[0037] S103. Based on the passenger allocation spatiotemporal network and the bus vehicle scheduling spatiotemporal network, a collaborative optimization model of shuttle bus service design and vehicle scheduling is established that takes into account the subway passenger flow control strategy.

[0038] For example, while considering the subway passenger flow control strategy, the shuttle bus service design and vehicle scheduling were collaboratively optimized to achieve an improvement in the overall performance of the multimodal public transportation system. The constructed model is concise and easy to understand, has good expansion space, and can be flexibly adjusted and optimized according to actual needs to adapt to more complex practical application scenarios.

[0039] S104. A heuristic solution algorithm based on the combination of Lagrangian relaxation and column generation is used to decompose the collaborative optimization model of shuttle bus service design and vehicle scheduling with the help of Lagrangian relaxation technology. At the same time, based on the column generation algorithm framework, a series of pricing sub-problems are constructed to dynamically generate bus service plans and vehicle scheduling plans.

[0040] Exemplarily, the solution of the model mainly includes three parts: explaining the decomposition method of the original model; designing an efficient solution algorithm based on Lagrangian relaxation and column generation; and conducting example verification.

[0041] From the analysis of the constructed model, it can be seen that the shuttle bus service design, as a key link, integrates the multi-commodity flow model for solving passenger allocation and the set covering model for solving vehicle scheduling, which makes the original model quite complex and suitable for decomposition with the help of Lagrangian relaxation technology.

[0042] The core idea of ​​the heuristic solution algorithm combining Lagrangian relaxation and column generation is to obtain Lagrangian multipliers by solving the Lagrangian dual problem to approximate the dual variables of the linear relaxation of the restrictive master problem. Using this dual information, valuable vehicle scheduling plans and bus service plans are iteratively generated under the framework of column generation to update the initial column set. Next, based on the updated columns and multipliers, the heuristic solution process based on Lagrangian relaxation is restarted to further construct a high-quality feasible solution.

[0043] S105. According to the calculation results, the optimal bus service operation plan, vehicle dispatch plan and refined passenger flow control strategy are determined, and the real-time passenger flow distribution is visualized to assist operation management personnel in making dynamic decisions.

[0044] The oversaturated subway network shuttle bus service design and vehicle scheduling optimization method of the present invention takes into account the time-varying passenger flow demand, based on the modeling method of the space-time network, and takes the minimization of passenger travel cost and bus operation cost as the optimization goal, and establishes an integer linear programming model. The model can simultaneously optimize the subway passenger flow control strategy and bus shuttle service; at the same time, it also explores an integrated optimization method, which comprehensively considers the decision-making problems at different stages in the bus operation planning process, so as to achieve the optimal trade-off between service level and operation cost.

[0045] In another implementation of the present invention, the set of candidate bus connection service solutions is represented as:

[0046]

[0047] in, is the set of all candidate connecting routes, Indicates each bus connecting route The applicable average headway set, represents the average headway time applicable to each bus connecting route, Based on the average headway setting The set of all feasible timetable options;

[0048] The set of service arcs in the passenger allocation space-time network is expressed as:

[0049]

[0050] in, Indicates each connection service plan The corresponding service arc set, the corresponding arc capacity is set to the maximum passenger capacity of the bus C bus .

[0051] For example, each service plan corresponds to a set of bus service arcs, the lengths of which are determined based on the pre-set interval travel time and minimum stop time. Indicates each connection service plan The corresponding service arc set, the corresponding arc capacity is set to the maximum passenger capacity of the bus C bus , then the bus service arc set can be defined as In order to establish a connection between the subway layer and the bus layer, mode transition arcs are used to represent passengers' transfer activities from subway to bus.

[0052] In order to describe the travel trajectory of passengers, it is necessary to specify clear starting and ending nodes for each passenger in the network and construct a feasible space-time path. represents the set of passengers, where each passenger At time point p Arrive at the starting station o p , and plan to go to the destination p In the constructed spatiotemporal network, the starting node of passenger p can be directly specified as (o p ,at p ,1). However, due to potential passenger congestion, each passenger may involve multiple travel trajectories, so its exact final destination node is difficult to determine in advance. In order to ensure the integrity of the space-time path assigned to the passenger, an additional virtual arrival node (d p ,T,1), and connect it with all possible actual arrival nodes. The arcs formed are classified as virtual arcs, and the corresponding arc costs are set to zero. For other types of arcs, their costs are determined based on the time required for the actual activity. However, for queuing arcs, in addition to considering the actual waiting time, it is also necessary to consider the dissatisfaction and inconvenience that passengers may experience due to long queues outside the station. Therefore, the cost of the queuing arc is expressed as ω·(t′-t), where ω is the penalty for queuing activities. In summary, the arcs in the network The arc cost c i,j,t,t′,m,m′ It can be defined as:

[0053]

[0054] Based on the above network structure, we can further explain the key role of bus shuttle services in alleviating subway system congestion. Consider the following oversaturation scenario: three passengers depart from station 2 at the same time (time point 1) and plan to go to the destination station 7. Assume that the maximum carrying capacity of subway trains, platforms, and buses is 1 person. Figure 4 and Figure 5 The passenger flow allocation results in a single subway system and a multi-modal public transportation system are shown respectively. Figure 4It can be seen that in a single subway system, in order to ensure that the number of stranded passengers on the platform is within a safe range, passenger 3 is affected by passenger flow control and has to queue outside the station until the waiting area on the platform is reopened at time point 3 before he can enter the station to wait for the train. In addition, due to limited train capacity, passenger 3 can only take the third train, and the final total travel time is 8 time units. Obviously, in a single subway system, there are significant differences in the level of passenger service. For passengers waiting on the platform and queuing outside the station, the travel experience is relatively poor.

[0055] To optimize passengers’ travel experience, Figure 5 A direct bus connection route from Station 2 to Station 7 has been added. In this multimodal public transportation system, if passenger 3 chooses to transfer to a bus after arriving outside the subway station, his total travel time is 5 time units. It can be seen that by flexibly using public transportation, passengers can avoid long waits and stays outside the station, thereby greatly shortening the overall travel time. Although the bus travel speed is slow in actual operation, in the case of an oversaturated subway system, the introduction of bus connection services can effectively alleviate the situation of passengers staying outside the station, significantly reduce the total travel time of passengers, and improve passenger travel satisfaction.

[0056] In another implementation of the present invention, the task arc set in the bus scheduling spatiotemporal network is expressed as:

[0057]

[0058] in, Indicates each service plan The collection of all quest arcs below.

[0059] For example, since the service arc is closely related to the bus service plan, To represent each service plan The set of all task arcs under the above conditions, where each task is represented by a four-tuple space-time arc in the bus scheduling space-time network, and then the task arc set can be defined as The exit arc and entry arc can intuitively describe the use of vehicle resources, where the exit arc is the connection between the bus station’s exit node and the starting point of the task arc, and the entry arc is the connection between the task arc’s end point and the bus station’s return node. An empty arc is added between any two compatible task arcs (i.e., after the previous task is completed, the vehicle can reach its starting station before the next task starts) to indicate the connection between the two. In order to simplify the network structure and avoid redundant connections between compatible tasks, only the empty arc between each task arc’s end node and the first compatible task’s start node of other stations, and the empty arc between each task arc’s start node and the last compatible task’s end node of other stations are retained. At the same time, in order not to lose any connection between compatible task arcs, a waiting arc is set between two adjacent nodes at each station, indicating that the vehicle is waiting on the roadside to prepare for subsequent transportation tasks.

[0060] The cost of the bus network arc can be set as a variable service operating cost proportional to the travel time. At the same time, an additional fixed cost is imposed on each exit arc as the cost of vehicle use. These two costs together constitute the total cost of bus vehicle scheduling. Based on this network structure, we define For each bus stop, Construct the corresponding bus scheduling spatiotemporal network and further generate all feasible vehicle scheduling solutions Each of these plans The cost of k is the sum of the arc costs involved.

[0061] Taking into account the subway passenger flow control strategy and the coordination of shuttle bus service design and vehicle scheduling, it is planned to implement refined passenger flow control on an oversaturated subway network to reduce subway platform congestion and ensure safe and smooth subway operation; secondly, design efficient shuttle bus services as a supplement to the subway system's capacity to provide affected passengers with flexible and convenient alternative travel options; finally, by allocating vehicle resources from multiple bus stations and flexibly deploying transportation tasks on different shuttle routes to reduce bus operating costs, the operating efficiency of the entire public transportation system will be comprehensively improved, and a high-quality, multi-modal coordinated public transportation system will be constructed to provide passengers with better quality and more efficient travel services.

[0062] In another implementation of the present invention, it also includes: defining a binary decision variable If passenger p is assigned to the arc (i,j,t,t′,m,m′) in the space-time network, otherwise

[0063] Define the binary decision variable y k , if vehicle scheduling scheme k is selected, y k=1, otherwise y k =0.

[0064] Define the binary decision variable z r,h,l , if the bus service solution (r,h,l) is selected as part of the feeder bus service design result, z r,h,l =1, otherwise z r,h,l =0.

[0065] Among the above decision variables, and k They are used to characterize the path choices of passengers and buses respectively. The internal connection and mutual influence between the two are expressed through the coupling variable z r,h,l Establish.

[0066] Based on the above decision variables, network flow balance constraints, capacity constraints and bus service feasibility constraints are formulated in the passenger space-time network.

[0067] In another implementation of the present invention, network flow balance constraints are used to ensure that each passenger The path feasibility is as follows:

[0068]

[0069] Among them, (o p ,at p ,1) and (d p ,T,1) respectively represent the starting point and virtual destination of passenger p in the network, represents the set of all passengers.

[0070] Due to the limited capacity of the infrastructure, it is necessary to restrict the number of passengers that can be assigned to some arcs in the passenger space-time network.

[0071] In order to control the level of passenger gathering on the subway platform, the number of passengers on the waiting arc must not exceed the safety threshold that the platform can accommodate. The platform capacity constraint is expressed as:

[0072]

[0073] In addition, the number of passengers in the car, that is, the cumulative number of passengers on each subway service arc, must not exceed the maximum passenger capacity of the subway train. The subway capacity constraint is expressed as:

[0074]

[0075] The number of passengers that can be assigned to each bus service arc should not exceed the number of passengers that the bus can carry. It is worth noting that since the design of the shuttle bus service is not given in advance, it is necessary to introduce an additional binary indicator parameter To describe whether the potential bus service arc is available. The bus capacity constraint is expressed as:

[0076]

[0077] Among them, the binary indicator parameter It is used to describe whether the potential bus service arc is available. If the service plan (r, h, l) corresponds to the potential bus service arc (i, j, t, t′, m, m′), then otherwise when When , the number of passengers that the corresponding bus service arc can accommodate is equal to the bus capacity, and the other irrelevant arcs do not allow any passengers to choose. When , it means that the connecting bus service is not opened and all bus service arcs are not feasible.

[0078] The feasibility constraints of bus service are used to ensure the feasibility of the bus service plan. Each candidate connecting route r can only choose one headway setting and one schedule plan at most. The specific constraints are as follows:

[0079]

[0080] In order to better characterize the coupling relationship between the bus service plan and the vehicle scheduling plan, a binary indicator parameter is introduced. Describes the correspondence between the service plan and the task arc in the bus dispatch network. This parameter can establish the coupling relationship between the vehicle dispatch plan variables and the bus service plan variables. Specifically, once the service plan is determined, the corresponding transportation task must be performed by a vehicle, and the corresponding constraint is expressed as:

[0081]

[0082] in, It is a common indicator parameter in the set covering problem constraints, indicating whether the vehicle scheduling solution k can cover the task arc (i, j, t, t′).

[0083] In another implementation of the present invention, the shuttle bus service design and vehicle scheduling collaborative optimization model is expressed as:

[0084]

[0085] Among them, c k is the operating cost of vehicle scheduling scheme k, c i,j,t,t′,m,m′ is the passenger travel cost of arc (i, j, t, t′, m, m′), ω1 and ω2 are the weight coefficients of passenger travel cost and vehicle scheduling cost respectively, and the decision variable They are represented by vectors x, y, and z respectively.

[0086] For example, the optimization goal is to develop a comprehensive optimization strategy based on the global perspective of the multimodal public transportation system to improve the service level, ensure operational safety, and effectively reduce operating costs. The above goals can be achieved by minimizing the travel cost of passengers and optimizing the vehicle scheduling cost. The key decision-making issues in the bus operation planning process, including route design, headway setting, timetable compilation and vehicle scheduling, are integrated, and an integer linear programming model is constructed based on the modeling method of the space-time network.

[0087] In another implementation of the present invention, the Lagrangian relaxation problem is as follows:

[0088]

[0089] In the relaxation problem, the coupling constraints between different decision variables no longer exist. The Lagrangian relaxation problem is decomposed into three tractable subproblems, including the passenger assignment subproblem, the vehicle scheduling subproblem, and the bus service design subproblem. The objective function is as follows:

[0090]

[0091] Among them, α i,j,t,t′,m,m′ , β i,j,t,t′ , γ i,j,t,t′,m,m′ and δ i,j,t,t′,m,m′ is the Lagrange multiplier, {β i,j,t,t′} (i,j,t,t′)∈Ω , They are represented by vectors α, β, γ, and δ respectively, and the arc (i, j, t, t′, m, m′) in the passenger allocation space-time network is denoted as τ, and the arc (i, j, t, t′) in the bus scheduling space-time network is denoted as τ′.

[0092] For example, by introducing the Lagrange multiplier α i,j,t,t′,m,m′ and β i,j,t,t′ The relaxation involves the bus service design variable z r,h,l Considering that capacity constraints (2) and (3) contain interactions between passengers, which are difficult to handle, we further introduce the multiplier γ i,j,t,t′,m,m′ and δ i,j,t,t′,m,m′ To relax these two types of constraints respectively. Based on the heuristic solution algorithm combining Lagrangian relaxation with column generation, the original model is transformed into a series of independent and easy-to-solve sub-problems. At the same time, a pricing problem is constructed to dynamically generate bus service plans and vehicle scheduling plans, and the optimization software GUROBI and the proposed heuristic algorithm are used to effectively solve the model.

[0093] In another implementation of the present invention, the passenger assignment subproblem is expressed as:

[0094]

[0095] in, is the generalized arc cost, expressed as:

[0096]

[0097] This subproblem only contains the passenger flow balance constraint, which transforms the passenger allocation subproblem into a series of path subproblems related to passenger allocation. The passenger routing subproblem can be defined as follows:

[0098]

[0099]

[0100] It should be understood that the path selection subproblem for each passenger is essentially a typical shortest path problem, whose core goal is to find an optimal path for each passenger in the space-time network, aiming to minimize the sum of the selected arc costs. This problem can be solved by a labeling algorithm. In particular, since these shortest path problems share the same network structure and do not affect each other, passengers can be grouped according to their arrival time, starting station, and destination, and then the shortest path solution can be assigned to different threads, using parallel computing technology and labeling algorithms for efficient solution. This processing method can significantly improve computing efficiency and can better cope with the increase in the number of passengers.

[0101] In another implementation of the present invention, the vehicle scheduling subproblem can be transformed into an unconstrained 0-1 integer linear programming problem, expressed as:

[0102]

[0103]

[0104] in, represents the generalized scheduling cost.

[0105] Preferably, the optimal solution to this subproblem can be obtained based on the decision variable y in the objective function (14): k The coefficients of are derived, namely:

[0106]

[0107] The optimal objective function value can be calculated by the following formula:

[0108]

[0109] In another implementation of the present invention, the bus service design sub-problem is expressed as:

[0110]

[0111] Proposition 1: Sub-problem (L Z ) is fully unimodular.

[0112] Proof: Let the coefficient matrix of constraint (5) be an R×B matrix A. Obviously, all elements of A belong to {0,1}, and each column contains exactly one non-zero element. Let the n-dimensional square matrix A′ be an arbitrary submatrix of A. The following proves by mathematical induction that the value of the determinant of each submatrix A′ belongs to {0,1,-1}. If n=1, the value of the determinant of A′ is equal to 0 or 1. Assume that when n=R-1, the value of the determinant of A′ belongs to {0,1,-1}. If n=R, rearrange A′ into the following form:

[0113]

[0114] Among them, v is an R-1 dimensional row vector composed of 0 or 1, 0 is an R-1 dimensional all-zero column vector, and A″ is an R-1 dimensional square matrix. According to the assumption, the value of the determinant of the square matrix A″ belongs to {0,1,-1}. It is not difficult to find that after expanding the matrix by the first column, the determinant value of A′ is equal to the determinant value of A″. Therefore, the proposition holds when n=R, and the proof is complete. All order sub-expressions of A are equal to 0, 1 or -1, indicating that the matrix A is fully unimodular.

[0115] Since the right-hand side terms of constraint (5) are all integers, the problem (L Z ) are all integers. Note that the relaxed subproblems are linear programming models that can be solved using the commercial solver GUROBI.

[0116] In addition, the service design sub-problem (L Z ) can also be further decomposed into a series of independent bus schedule compilation sub-problems. For each candidate connecting route Sub-questions It can be defined as follows:

[0117]

[0118] In the formula, the generalized cost It is not difficult to find that the sub-problem obtained after decomposition is essentially a special knapsack problem, which can be solved by judging the coefficients of the decision variables, namely:

[0119]

[0120] The corresponding optimal objective function value can be calculated by the following formula:

[0121]

[0122] In addition to decomposing the original model with the help of the above-mentioned Lagrangian relaxation technique, the model can also be modified by fixing the values ​​of certain decision variables. Since the uncertainty of the bus service arcs in the passenger assignment network and the task arcs in the bus scheduling network are affected by the bus service design scheme, once the service design variable z is determined, the passenger network and bus network can be reconstructed based on the feasible service arcs and task arcs, and then the passenger assignment problem (P) can be solved independently under the fixed service scheme. X ) and the multi-depot vehicle scheduling problem (P Y Specifically, the problem (P X ) aims to assign a suitable travel path to each passenger to minimize the total travel cost of the passenger while satisfying the capacity constraints of subway platforms, trains and buses in the multimodal transportation system, that is:

[0123]

[0124] stConstraints(1)-(4).

[0125] Problem (P Y ) aims to find a set of optimal bus scheduling solutions to perform all transportation tasks with the minimum operating cost, as shown below:

[0126]

[0127] stConstraints(6).

[0128] It should be understood that the solution algorithm based on Lagrangian relaxation and column generation is divided into two main stages, and the specific steps are described as follows: First, the first stage can be regarded as a preprocessing stage based on the column generation algorithm, which aims to generate high-quality column sets and Lagrangian multipliers. Before the start of this stage, the service design scheme under the equilibrium timetable is taken as the initial input, and the corresponding vehicle scheduling problem is solved for each scheme to construct the initial column set of the scheduling scheme. At the same time, the Lagrangian multiplier is initialized to a zero vector, and the initial upper bound is obtained by solving the passenger assignment problem under a single subway system (Step Ⅰ); then, the subgradient algorithm is used to solve the Lagrangian dual problem based on the restricted column to obtain the optimal Lagrangian multiplier and Lagrangian lower bound (Step Ⅱ); the obtained Lagrangian multiplier is used as an approximate dual variable to solve the pricing problem of the bus service plan and the vehicle scheduling plan in turn (Step Ⅲ); a column with a negative test number is generated and added to the main problem (Step Ⅳ); the lower bound of the original problem is estimated based on the updated column, and a new upper bound is generated. During the iteration process, if a new column is found or the number of iterations does not reach the threshold given in advance, the above steps are repeated, otherwise, go to the second stage.

[0129] The second stage can be regarded as an improvement stage based on Lagrangian relaxation, which aims to further explore high-quality feasible solutions based on the columns and Lagrangian multipliers updated in the first stage. In each iteration, the Lagrangian relaxation problem is first solved to obtain the lower bound and bus service design results (Step V); then, based on the fixed bus service plan, the passenger assignment problem and the multi-depot vehicle scheduling problem are solved respectively to construct a feasible solution (Step VI); the subgradient algorithm is also used to update the Lagrangian multipliers (Step VII). Once the number of iterations or the relative error between the upper and lower bounds reaches a given threshold, the algorithm terminates and outputs the best feasible solution obtained in the solution process.

[0130] (1) Phase 1

[0131] Since the integrated optimization problem of shuttle bus service design and vehicle scheduling adopts the modeling method of set covering model, this method needs to enumerate all feasible service solutions and vehicle scheduling solutions, resulting in an abnormally large number of decision variables (columns). Although the strategy of designing a limit on the number of candidate shuttle routes and headways can improve the problem to a certain extent, it is still very difficult to enumerate all feasible timetable solutions. In particular, as the number of task arcs increases, the number of corresponding vehicle scheduling solutions will show an exponential growth, making the problem more and more complicated. In order to effectively meet these challenges, the present invention adopts a column generation technology, which gradually expands the feasible domain by solving a series of simplified problems containing only part of the columns in turn, and using dual variables to iteratively generate new columns, until no new columns are generated and the algorithm converges. In the context of column generation, this simplified problem is usually called a restrictive main problem, and the problem of generating new columns is called a pricing subproblem.

[0132] (1.1) Restricted main problem

[0133] Generally speaking, for an integer programming problem, the simplex method can be used to solve its linear relaxation problem to obtain the dual solution. However, considering the computational complexity, this method is not applicable to this problem. Recalling the discussion on model decomposition in the previous section, the passenger assignment problem (L X ) contains a series of independent shortest path problems, and the coefficient matrix of each path subproblem is fully unimodular. At the same time, the bus service design subproblem (L Z ) is also fully unimodular. In addition, the vehicle scheduling subproblem (L Y ) is an unconstrained 0-1 linear programming problem. Therefore, the three Lagrangian relaxation subproblems all have integer properties, that is, the solutions of their relaxation problems are the same as the solutions of the original problems. Then the dual problem of the linear relaxation of the original problem is equivalent to the Lagrangian dual problem, so the approximate dual solution can be obtained by solving the Lagrangian dual problem. The dual problem (LD) can be defined as:

[0134]

[0135] Obviously, the above problem is not differentiable and is usually solved using a subgradient algorithm. In each iteration, only the Lagrangian relaxation problem needs to be solved and the Lagrangian multipliers are updated in combination with the feasible solution obtained from the external column generation framework. Since each iteration inherits the multipliers obtained in the previous iteration and the three Lagrangian relaxation subproblems can be solved efficiently, the entire subgradient optimization process can be completed in a relatively short time. The steps of the subgradient algorithm (Algorithm 1) for solving the restricted master problem are as follows:

[0136] Step 1: Initialize the Lagrange multiplier and upper and lower bounds, and set the number of iterations n = 1;

[0137] Step 2: Solve the Lagrange subproblem based on the current column set and obtain the Lagrange multiplier and lower bound;

[0138] Step 3: Update the Lagrange multiplier according to the relaxed solution and upper and lower bounds;

[0139] Step 4: Convergence test. The algorithm terminates when the relative error between the upper and lower bounds is less than the set threshold or the maximum number of iterations is reached, otherwise n=n+1 is set and the algorithm returns to step 2.

[0140] (1.2) Pricing sub-problem

[0141] After solving the Lagrange dual problem, the optimal multiplier will be used as an approximate dual solution to price out new columns with negative test numbers. Lagrange multiplier α τ , β τ′ , γ τ and δ τ They correspond to constraints (4), (6), (2) and (3) respectively. Define λ r is the approximate dual variable of constraint (5), and its value can be calculated according to the following formula:

[0142]

[0143] After that, each service plan The number of tests It can be calculated by the following formula:

[0144]

[0145] Service Plan Pricing Subproblem (PP Z ) aims to determine the optimal arrival and departure times of all services at each stop, given candidate connecting routes and headway settings. In a passenger network, the schedule of each service corresponds to a space-time path connecting the route's start and end stations. Conversely, by constructing a series of compatible service arcs, the schedule of the service can be accurately derived. Connecting routes Setting the time distance to the front vehicle The timetables of all services under the above condition together constitute a feasible timetable solution l. Therefore, the pricing problem aims to find a bus service solution that can reduce the value of the objective function of the main problem. so that Or prove that no such solution exists.

[0146] In order to construct a feasible timetable solution, the following three types of constraints need to be considered: (i) Time window constraints. In order to ensure that bus services are evenly distributed during the operating period, time window constraints should be imposed to specify the adjustment range of each service timetable to ensure that the optimized arrival and departure times meet the predetermined time limits. (ii) Stop time constraints. Usually, the timetable can be adjusted by extending the stop time at the starting station or the intermediate stop. Setting stop time constraints can ensure that the vehicle stop time is within the preset range. (iii) Flow balance constraints. Since the timetable compilation aims to find a feasible compatible space-time path, the flow balance constraint of each intermediate stop must be guaranteed.

[0147] The following will explain in detail how to build a customized spatiotemporal network for this pricing problem. s,i ,LT s,i ] represents the departure time window of service s at station i, where the station is determined according to the given connecting route and the number of services is estimated by the headway. In fact, once the time window of the starting station is given, the time windows of subsequent stations can be derived. Figure 7 Two bus services in the passenger network are shown, and the feasible areas for their schedule adjustments are marked with different colors. Taking service s as an example, first determine its estimated departure time dt at the starting station (station 1) according to the equilibrium schedule solution s , on this basis, a perturbation is applied You can get the departure time range of the service, including the earliest departure time and the latest departure time Next, we continue to derive the earliest and latest departure time constraints of the service at subsequent stations. Taking station 3 as an example, the earliest departure time ET of service s is s,3 Equal to its earliest departure time ET at the previous station s,2 , interval travel time rt 2,3 and the minimum stop time at the current station The sum of As for its latest departure time, it is necessary to check whether there is a conflict between the current service and the subsequent service. Without considering the impact of other services, the latest departure time LT of service s at station 3 s,3 Equal to its latest departure time LT at the previous station s,2 , interval travel time rt 2,3 、The minimum stop time at the current station and maximum additional stop time In addition, to ensure a safe departure interval between adjacent services, the latest departure time of service s should not exceed the earliest departure time ET of its subsequent service. s+1,3Minimum safe departure interval h min Therefore, the latest departure time of service s at station 3 is LT s,3 It is equal to the latest possible departure time of the service under the premise of maintaining a safe departure interval, that is:

[0148]

[0149] In summary, the calculation formula for the earliest and latest departure times of each service at each station can be summarized as follows:

[0150]

[0151]

[0152] In the formula, rt i,i′ represents the travel time between stations i and i', h min is the minimum safe departure interval between adjacent services, and They represent the minimum stop time and the maximum additional stop time at station i respectively.

[0153] In order to transform the timetable compilation problem considering the stop time constraint into the shortest path problem, the present invention is particularly aimed at the pricing problem (PP Z ) built a series of spatiotemporal networks like Figure 8 As shown in Figure 1. Each node represents a feasible departure time and is connected to the available departure nodes of the subsequent station. The cost of each arc can be determined by the dual variable according to its originating node, that is,

[0154] Note that the dual variable β τ′ It is only related to the start and end time of the transportation task corresponding to each service, which is difficult to directly consider into the network structure. For example, if service s is sent from station 1 at time t1 and arrives at station 4 at time t2 to complete the transportation task, then the product term will be taken into account in the calculation of the test number, otherwise it is irrelevant to the calculation of the test number. Therefore, it is necessary to enumerate all feasible start and end time combinations of each service, and further construct the corresponding pricing subnetwork based on the relevant arcs under each combination. like Fig. 9 As shown. Through this enumeration method, the dual variable β τ′ and λ ris considered as a constant and then explicitly represented in the network. Although this approach will increase the size of the problem, since each subnetwork is independent of each other, it can be solved efficiently with the help of parallel computing technology. At this point, all relevant constraints and dual variables have been incorporated into the spatiotemporal network structure, and the pricing problem for the service plan has been successfully transformed into a series of independent shortest path problems, which can be solved in parallel using the labeling algorithm.

[0155] The pricing problem of vehicle scheduling plan (PP Y ) aims to determine the optimal vehicle driving trajectory based on a given parking lot. The number of tests It can be calculated by the following formula:

[0156]

[0157] In the formula, c k Represents the operating cost of vehicle scheduling solution k. Generally speaking, there are two main strategies for finding a scheduling solution with the smallest test number. The simplest is to use a depth-first search strategy for each parking lot to enumerate all feasible scheduling solutions and identify solutions with negative test numbers. However, as the number of transportation tasks increases, the number of feasible vehicle scheduling solutions will grow exponentially, and it is difficult to fully enumerate all solutions. Therefore, the present invention adopts another more efficient strategy, which is to convert the generation of vehicle scheduling solutions into solving the shortest path problem in the vehicle scheduling network after updating the arc costs. For ease of explanation, the bus scheduling network is represented as The network includes task arc Ω, exit arc Entry Arc Empty arc And the waiting arc For each arc The corresponding costs can be defined as follows:

[0158]

[0159] Among them, c f and c v represent the fixed cost of vehicle use and the variable cost per unit time of vehicle operation, respectively. Therefore, the test number of the vehicle scheduling scheme is the sum of all relevant arc costs, and the scheduling scheme with the lowest test number will be added to the main problem. Similar to the other pricing problem, this problem can also be solved efficiently by using a parallel labeling algorithm for different parking lots. In addition, it should be noted that since the values ​​of the dual variables are approximated by Lagrange multipliers, the columns in the restricted main problem may also have negative test numbers. In order to prevent the repeated generation of the same columns, it is necessary to appropriately adjust the dual variables to ensure that the columns in the main problem have non-negative test numbers.

[0160] During each column generation iteration, the lower bound of the original problem needs to be estimated. The calculation formula is as follows:

[0161]

[0162] Where L(α,β,γ,δ) represents the Lagrangian lower bound obtained during the iteration process. and are the sets of columns added to the main problem. In addition, based on the updated columns and the relaxed solution of bus service design, the upper bound can be recalculated using Algorithm 3 and the solver GUROBI. In summary, the steps of the preprocessing algorithm based on column generation (Algorithm 2) can be summarized as follows:

[0163] Step 1: Initialization. Generate the initial column set and Algorithm 3 is used to solve the passenger assignment problem in a single subway system and obtain the optimal target value as the initial upper bound UB cg . Initialize the lower bound LB cg =-∞ and Lagrange multipliers α, β, γ, δ are zero vectors. Let the number of iterations n = 1.

[0164] Step 2: Solve the restricted master problem. Solve the Lagrangian dual problem (LD) based on the current column. Obtain the Lagrangian lower bound, relaxed solution, and Lagrangian multipliers.

[0165] Step 3: Solve the pricing subproblem. Calculate the approximate dual variable λ of constraint (5) r . Solve the pricing problem of bus service plans and vehicle scheduling plans and obtain the corresponding test numbers.

[0166] Step 4: Column management. Add the columns with negative test numbers to the restricted main problem, that is,

[0167] Step 5: Convergence test. Based on the relaxed solution, use Algorithm 3 and the solver GUROBI to calculate the feasible solution and estimate the lower bound of the original problem. When no new columns are generated or the maximum number of iterations is reached, the algorithm terminates and goes to the second stage, otherwise set n = n + 1 and return to step 2.

[0168] (2) Second stage

[0169] In the first stage, the calculation of the upper bound is mainly used to update the step size of the subgradient algorithm, and the quality of the resulting feasible solution is often difficult to guarantee. Therefore, in the second stage, a heuristic method based on Lagrangian relaxation is used to further construct a high-quality feasible solution. This stage is similar to the method used to solve the restricted master problem. The main difference is that a feasible solution will be generated in each iteration, and the truncated column generation algorithm is used to further improve the quality of the solution to the vehicle scheduling problem. Given that the first stage has provided high-quality Lagrangian multipliers and columns as initial parameters, the calculation time of the second stage will be greatly reduced, and only a small number of iterations are required to converge to a high-quality solution.

[0170] (2.1) Feasible solution

[0171] Although the Lagrangian relaxation problem can provide a good lower bound for the original problem, the obtained solution is often infeasible. Therefore, the present invention proposes a heuristic method for constructing a feasible solution based on the bus service plan in the relaxed solution. Since the capacity constraints (2)-(4) are not considered in the relaxed problem, the obtained passenger allocation result may contain infeasible bus service arcs or violate the established arc capacity constraints. In order to generate a feasible allocation solution, it is first necessary to reconstruct the passenger allocation spatiotemporal network. The bus service arcs in the network are based on the relaxed subproblem (L Z ) is used to determine the bus service plan obtained in . Then, the feasibility of the passenger time-space paths generated in the relaxed solution is evaluated one by one according to the time sequence of passenger arrival. For those feasible paths that meet the constraints, they are retained, and for those that are not feasible, the shortest path algorithm is restarted to find new feasible paths for the relevant passengers. The specific steps of the algorithm (Algorithm 3) for obtaining feasible passenger allocation results are as follows:

[0172] Step 1: Based on the given bus service plan variables Filter out feasible bus service arcs and reconstruct the passenger allocation space-time network. Generate passenger loading order S: Sort passengers according to their arrival time. If there are simultaneous arrivals, sort them according to the rule of increasing travel cost in the relaxed solution. Initialize the cumulative passenger flow f on each arc τ .

[0173] Step 2: Traverse the loading sequence S and obtain the relaxation solution for passenger p Perform the following steps:

[0174] (a) Check whether the arc capacity constraint is violated: (i) For arc like Then set the arc cost c τ =∞; (ii) for the arc like Then set the arc cost cτ =∞; (iii) for the arc like Then set the arc cost c τ =∞. If no arc cost is modified, go to (c).

[0175] (b) Reconstruct the space-time network based on the modified arc costs, use the labeling method to find a new shortest path for the passenger, and update the passenger assignment result

[0176] (c) For arc Update cumulative passenger flow

[0177] Similarly, once the service plan is given, all transportation tasks can be screened out, and the original problem can be simplified to a vehicle scheduling problem with multiple parking lots, which can then be solved using the truncated column generation algorithm. The specific algorithm steps are as follows:

[0178] Step 1: Based on the given bus service plan variables After selecting feasible task arcs, reconstruct the bus scheduling space-time network. Use the vehicle scheduling plan generated in stage 1 as the initial set of the algorithm. Set the number of iterations n = 1 and n0 = 1. Set the algorithm parameters y min and ε * .

[0179] Step 2: If and Go to step 6.

[0180] Step 3: Based on the current column The vehicle scheduling problem (P Y ) to obtain the solution of the original problem And the solution of the dual problem {π τ′} τ′∈Ω , and retain the optimal objective function value

[0181] Step 4: For each parking lot Perform the following steps:

[0182] (a) Update the arc cost according to the dual solution:

[0183] (b) Use the labeling method to solve the pricing subproblem, that is, solve the shortest path problem in the vehicle scheduling network and retain the cost of the shortest path as the test number.

[0184] Step 5: If Then add the vehicle dispatching plan corresponding to the shortest path into the main problem, then set n=n+1 and return to step 3.

[0185] Step 6: If If both are integers, the algorithm is terminated, otherwise the rounding operation y is performed k* =1,k*∈{k|y k ≥y min}, let n0=n,n=n+1 and return to step 3.

[0186] (2.2) Multiplier update

[0187] This study adopts the widely used subgradient algorithm to update the Lagrange multiplier to obtain a better lower bound. In particular, the present invention makes some modifications to the standard subgradient algorithm to improve the algorithm performance. Define the vector represents the subgradient of the associated relaxed constraint, and its components can be calculated according to the following formula:

[0188]

[0189] Afterwards, the multiplier is updated according to the following formula:

[0190]

[0191] In the formula, the superscript n represents the current iteration number, the parameter μ is the step size adjustment factor, and the value range is usually set to (0,2], UB * and LB * are the best upper and lower bounds obtained by the algorithm, respectively. In particular, the present invention designs different update step sizes based on the characteristics of each type of multiplier, and the calculation results show that this strategy can ensure faster convergence of the algorithm. In summary, the steps of the improved algorithm (Algorithm 5) based on Lagrangian relaxation are as follows:

[0192] Step 1: Initialization. Use the updated column set and Lagrange multiplier obtained in stage 1 as the initial parameters of the algorithm and initialize the global lower bound LB lr =LB cg , the global upper bound UB lr =UB cg , let the number of iterations n=1.

[0193] Step 2: Solve the Lagrangian relaxation problem.

[0194] (a) Use the parallel labeling algorithm to solve the passenger assignment subproblem (L X ), and obtain the relaxed solution and the corresponding objective function value L X (α,γ,δ).

[0195] (b) Solve the vehicle scheduling subproblem (L) by calculating the coefficients of the decision variables Y ), and the relaxed solutions are obtained according to formulas (15) and (16) respectively: and the corresponding objective function value L Y (β).

[0196] (c) Solve the bus service design subproblem (L) by calculating the coefficients of the decision variables Z ), and the relaxed solutions are obtained according to formulas (20) and (21) respectively: and the corresponding objective function value L Z (α,β).

[0197] Step 3: Update the global lower bound. Calculate the lower bound L(α, β, γ, δ) in the current iteration according to formula (9), and update the global lower bound LB lr =max{LB lr ,L(α,β,γ,δ)}.

[0198] Step 4: Construct a feasible solution.

[0199] (a) Based on the relaxation solution Algorithm 3 is used to solve the passenger assignment problem (P X ), and obtain the feasible solution x and the corresponding objective function value

[0200] (b) Based on the relaxation solution Algorithm 4 is used to solve the multi-depot vehicle scheduling problem (P Y ), and obtain the feasible solution y and the corresponding objective function value

[0201] Step 5: Update the global upper bound. Calculate the upper bound in the current iteration. And update the global upper bound

[0202] Step 6: Multiplier update. Calculate the step size and use the subgradient algorithm to update the Lagrange multiplier.

[0203] Step 7: Convergence test. The algorithm terminates when the relative error between the upper and lower bounds is less than the set threshold or the maximum number of iterations is reached, otherwise n=n+1 is set and the algorithm returns to step 2.

[0204] The subway passenger flow control strategy combined with shuttle buses proposed in the present invention aims to use buses as a capacity supplement for the subway system, and provide additional travel options for queuing passengers at passenger flow control stations, so as to reduce the inconvenience caused to passengers by passenger flow control measures, thereby improving the utilization rate of bus lanes while alleviating the congestion of the subway system. It is understood that there are relatively few studies on providing alternative travel modes (such as bus services) for oversaturated subway networks. Usually, subways and buses are often regarded as two independent modes of transportation, and there is a certain competitive relationship. However, through proper collaborative optimization, they can be integrated into a coordinated multi-modal public transportation system. Therefore, the present invention focuses on how to design efficient bus shuttle services to supplement the capacity of the urban rail transit network, while ensuring operational safety, and improving the service quality of the entire public transportation system.

[0205] In order to explain the present invention more clearly, the present invention is further described below in conjunction with preferred examples and drawings.

[0206] A series of numerical examples are designed, and the heuristic algorithm combining Lagrangian relaxation and column generation proposed in this paper and the commercial solver GUROBI are used to verify the invention. Figure 2 The subway network with two bidirectional rail transit lines shown in the figure has 4 stations in each direction of line 1, represented by {1,2,3,4}, and line 2 contains 3 stations, represented by {5,6,7}, where station 3 and station 6 are the same physical transfer station. In addition, two parking lots near the rail transit network are considered to be used to dispatch shuttle buses to perform transportation tasks. The study time domain is set to 50 minutes, and the entire study time domain is discretized into a series of time points in units of one minute. In order to simulate the oversaturation phenomenon of the subway system during peak hours, it is assumed that the number of passengers arriving at each station per unit time follows a normal distribution. Fig.10 and Fig.11 The time-varying passenger flow demand of Line 1 and Line 2 is shown respectively, with a total of 848 passengers. The passenger distribution ratio of each OD pair is detailed in Table 1. According to the spatiotemporal distribution characteristics of passenger flow, it is considered to operate shuttle buses during the operation period [15,25] to alleviate the congestion of the subway system. It is worth noting that the bus operation period lags slightly behind the peak period of passenger arrival. The reason is that congestion often appears after the initial peak period of passenger arrival. At the same time, the transfer activities between multiple lines also cause a deviation between the actual peak period of the subway system and the peak period of passenger arrival. Subsequent experiments have also confirmed this point, and the calculation results show that this operation period setting has better service performance. In actual operation, the operation period of the shuttle bus can be synchronized with the passenger flow control period of the subway system.

[0207] Table 1 Passenger flow demand ratio of different OD pairs

[0208]

[0209]

[0210] Regarding the experimental parameters, the departure interval of Line 1 and Line 2 is set to 6 minutes, and each train runs based on the scheduled travel time (150 seconds) and stop time (30 seconds). To ensure operational safety and service quality, it is assumed that the subway platform can accommodate a maximum of 35 passengers, and the maximum passenger capacity of subway trains and buses is 70 and 14, respectively. In terms of the design of the shuttle bus service, a total of ten alternative shuttle routes and three headway settings (3 minutes, 5 minutes and 8 minutes, respectively) were considered, and the adjustment parameters of the timetable were set accordingly (1 minute, 2 minutes and 3 minutes, respectively). In addition, the bus safety interval is set to 1 minute, and the maximum additional stop time is set to 2 minutes. The fixed cost of vehicle use is 25, the unit time variable cost of vehicle operation is 1, and the penalty weight of passenger queuing is 2.

[0211] In terms of algorithm parameter settings, if no new columns are generated or the maximum number of iterations is reached The first stage of the algorithm terminates. During each column generation iteration, the maximum number of iterations and the acceptable upper and lower bounds of the relative error of the subgradient algorithm are 150 and 1% respectively. When the algorithm enters the second stage, if the maximum number of iterations is reached Or the relative error between the upper and lower bounds is less than the given threshold 1%, then the second stage of the algorithm terminates. The parameters of the truncated column generation algorithm are y min = 0.8 and ε * = 0.01. Unless otherwise stated, the above algorithm parameters remain unchanged in the following experiments.

[0212] Based on the above case description, we first directly use GUROBI to solve the model to explore the actual performance of the various operation strategies proposed in this paper, including the introduction of feeder bus services and the use of non-balanced timetables, so as to verify the effectiveness of the proposed method. In order to reveal the oversaturation phenomenon of the subway network, we first give a set of candidate service solutions without considering feeder bus services (i.e., is empty) of the subway operation status. Fig.12 and Fig.13 The passenger flow distribution of each station on the two lines is shown in detail, where the light gray and dark gray rectangles represent the number of passengers queuing outside the subway station and waiting on the platform, respectively, and the number of passengers in the train is represented by straight lines of different shades to reflect the degree of train congestion. In this operation scenario, the optimal objective function value is 11758, indicating that the total travel cost of all passengers is 11758.

[0213] As can be seen from the figure, when the subway system is used as the transport capacity supply, each station has different degrees of oversaturation. In particular, the congestion problem of Station 1, Station 2, Station 3 and Station 6 is particularly serious. Specifically, although Station 1 is the starting station and the train capacity is sufficient to accommodate the arriving passengers, due to the limitation of platform capacity, some passengers still need to queue outside the station. When the train arrives at Station 2, the remaining train capacity cannot meet the accumulated passenger demand during the departure interval, resulting in a large number of passengers stranded on the platform. For transfer stations 3 and 6, the presence of transfer passenger flow further aggravates the congestion. Once these transfer passengers arrive, the limited platform capacity will be quickly saturated. At this time, in order to ensure safety, the subsequent arriving passengers need to wait outside the station first. In addition, it can be seen from the figure that the peak period of the subway system starts from the 15th minute and gradually recovers until about 25 minutes, which is consistent with the pre-set bus operation period.

[0214] In order to alleviate the extreme congestion of the subway system, the proposed shuttle bus service strategy is further introduced. After 3373.09 seconds of calculation, the optimal solution with an objective function value of 9224 and a relative error of 0 is obtained. The corresponding passenger flow distribution results are as follows: Fig.14 and Fig.15 Compared with the case where the multimodal public transportation system is not considered, the introduction of feeder buses as an additional capacity supplement to the subway system can reduce the total travel cost of passengers by about 26% (from 11758 to 8696). Fig.12 and Fig.14 It can be seen that after the introduction of the shuttle bus service strategy, the congestion on the subway platform and the queues outside the station have been significantly improved. Almost all passengers can complete their travel without obstacles, effectively avoiding delays caused by congestion.

[0215] In the operational scenario of introducing shuttle buses, a total of six shuttle routes were designed and opened, and 11 buses were dispatched from nearby parking lots to perform transportation tasks, including 5 buses dispatched from parking lot 1 and 6 buses dispatched from parking lot 2. In order to further explore the effect of shuttle bus services under different operating modes, based on the optimized shuttle routes and the corresponding headway settings, the operating performance under the equilibrium schedule was further studied (i.e., the disturbance parameters and additional stop time were set to 0). The comparison results are shown in Table 2. It can be seen that compared with the fixed headway operation mode, the passenger travel cost and bus dispatch cost under the non-equilibrium schedule strategy were reduced by 4.02% and 8.17% respectively, and the service level and operational efficiency were improved.

[0216] Table 2 Comparison of results of different operation modes

[0217]

[0218] Table 3 lists the optimized connecting routes and their schedules under different departure interval modes. It can be seen from the table that under the balanced schedule mode, the adjacent services on each connecting bus route strictly follow the preset headway time. In contrast, the unbalanced schedule mode shows higher flexibility, allowing vehicles to flexibly adjust the arrival and departure times at the stopover stations according to the time distribution characteristics of passenger flow demand, thereby better coordinating with the subway train schedule, enhancing the service efficiency of the multimodal transportation system, and reducing passenger travel costs. In addition, this flexible schedule adjustment strategy also improves operational efficiency to a certain extent. To illustrate this point, Fig.16 and Fig.17 The transport tasks corresponding to each connecting route under the two modes are shown in the form of a Gantt chart, in which the serial number of the vehicle performing the task is marked, and the parking lot to which the vehicle belongs is distinguished by different colors. It is worth noting that although the optimized service plan involves 14 transport tasks, by introducing a flexible cross-line vehicle scheduling strategy and an unbalanced schedule strategy, only 11 vehicles are needed to complete these services. In contrast, under the balanced schedule mode, due to the limitation of fixed departure intervals, Fig.16 The train 11 will be merged by adjusting the timetable Fig.17 The transport tasks performed by vehicle 1 and vehicle 12. Therefore, an additional vehicle (i.e., vehicle 12) is required in the balanced schedule mode to ensure the feasibility of the service solution, which leads to an increase in operating costs.

[0219] Table 3. Shuttle routes and timetables under different operation modes

[0220]

[0221] In order to further verify the effectiveness of the proposed algorithm, the commercial solver GUROBI and the heuristic algorithm based on Lagrangian relaxation and column generation (LR&CG) are used to solve instances of different scales. The results are shown in Table 4. 1 represents the optimal solution obtained by the solver, and the upper bound 2 Represents the best feasible solution obtained by the algorithm, relative error 1 Represents the relative error of the upper and lower bounds obtained by the solver, the relative error 2 Represents the relative error between the upper bound obtained by the algorithm and the estimated lower bound, the deviation 1 = (serial computing time - parallel computing time) / serial computing time, which indicates the efficiency gap between the algorithm serial computing and parallel computing. Deviation 2 = (algorithm parallel computing time - solver solution time) / solver solution time, which indicates the gap between the algorithm and the solver in solution time. Relative error 3 =(upper bound 2 -Upper Bound 1 ) / Upper bound2 , which represents the gap between the algorithm and the solver in terms of the optimal solution obtained. Given that the non-balanced timetable strategy will derive a large number of candidate timetable schemes and vehicle scheduling schemes, this will significantly increase the complexity of the problem. In order to simplify the problem, an additional set of experiments are introduced, which are based on the balanced departure interval mode. In order to illustrate the problem scale of each instance, a triple is used to describe the experimental parameter setting, including the operating mode, the study time domain, and the number of passengers (for example, U-50-848 represents a 50-minute study time domain and 848 passengers under the non-balanced timetable, while E-50-848 represents the corresponding experimental setting under the balanced departure interval mode).

[0222] Table 4 Comparison of results of two solution methods

[0223]

[0224]

[0225] According to the above table, for the numerical experiments under the balanced departure interval mode, when the case size is relatively small, the solver can still find the optimal solution within an acceptable time. However, with the increase of the study time domain and the number of passengers, the solver performance drops sharply, and it is even unable to provide a feasible solution for a simple instance with only 80 time points and 7936 passengers within 3600 seconds. In contrast, the proposed heuristic algorithm based on Lagrangian relaxation and column generation can always obtain solutions with a relative error of less than 5% corresponding to all instances in a shorter time. In general, the proposed algorithm can obtain a solution with only about 1% deviation from the optimal solution generated by the solver while shortening the solution time by about 94.02%, which proves the effectiveness of the algorithm.

[0226] Compared with the balanced schedule strategy, the introduction of the unbalanced schedule strategy significantly increases the complexity of the problem. For the two instances "E-50-848" and "U-50-848" with the same parameters, the difference in the calculation time of the solver is as high as 20.4 times (i.e., from 157.39 seconds to 3373.09 seconds), while the calculation time of the proposed algorithm only increases by about 1.8 times (i.e., from 6.70 seconds to 18.63 seconds). When the problem scale is further expanded, the solver cannot return a feasible solution due to memory limitations that make it impossible to enumerate all feasible columns. In contrast, despite the relatively complex problem, the proposed algorithm can find high-quality solutions within an acceptable time, which is only about 2.10% deviation from the optimal solution generated by the solver. In addition, compared with serial calculation, the adopted parallel computing technology can reduce the solution time by 32.53%, proving the effectiveness of the acceleration strategy.

[0227] In summary, the model disclosed in the present invention has good applicability in alleviating the extreme congestion problem of the subway system by operating feeder bus services during peak hours.

[0228] First, the collaborative transportation organization method for the multi-modal public transportation system during subway peak hours mentioned in the present invention is designed based on the idea of ​​system optimization, which can maximize the safety of subway system operations while improving the passenger transportation service level and bus operation efficiency. Secondly, the heuristic algorithm designed by the present invention based on the combination of Lagrangian relaxation and column generation can quickly obtain high-quality solutions. In fact, for large-scale instances, the commercial solver GUROBI can no longer handle the model, but the method proposed by the present invention can still obtain a better solution within a limited time.

[0229] Thus far, specific embodiments of the present invention have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired results. Additionally, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing may be advantageous.

[0230] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0231] In the description of the present invention, the terms "first" and "second" are only used to facilitate the description of different components or names, and cannot be understood as indicating or implying a sequential relationship, relative importance, or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0232] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0233] It should be noted that although the specific embodiments of the present invention are described in detail in conjunction with the accompanying drawings, it should not be understood as limiting the scope of protection of the present invention. Within the scope described in the claims, various modifications and variations that can be made by those skilled in the art without creative work still belong to the scope of protection of the present invention.

[0234] The examples of the embodiments of the present invention are intended to concisely illustrate the technical features of the embodiments of the present invention so that those skilled in the art can intuitively understand the technical features of the embodiments of the present invention, and are not intended to be improper limitations of the embodiments of the present invention.

[0235] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for designing bus service and optimizing vehicle dispatching for an oversaturated subway network, characterized in that: include: The passenger distribution spatiotemporal network is constructed based on the subway network data, train timetable data and the operation data of ground buses between stations, and the bus vehicle dispatch spatiotemporal network is constructed based on the location of bus stations and bus dispatch related data; According to passenger flow distribution and road conditions, alternative connecting routes, feasible headways and schedule adjustment parameters are set, and service arcs in the passenger allocation space-time network and task arcs in the public transport vehicle scheduling space-time network and a set of initial scheduling plans are generated; According to the passenger allocation spatiotemporal network and the bus vehicle scheduling spatiotemporal network, a shuttle bus service design and vehicle scheduling collaborative optimization model considering the subway passenger flow control strategy is established; A heuristic solution algorithm based on Lagrangian relaxation combined with column generation is adopted to decompose the collaborative optimization model of shuttle bus service design and vehicle scheduling with the help of Lagrangian relaxation technology. At the same time, based on the column generation algorithm framework, a series of pricing sub-problems are constructed to dynamically generate bus service plans and vehicle scheduling plans. Based on the calculation results, the optimal bus service operation plan, vehicle scheduling plan and refined passenger flow control strategy are determined. At the same time, the real-time passenger flow distribution is visualized to assist operation managers in making dynamic decisions.

2. The method according to claim 1, characterized in that The set of candidate bus connection service solutions is expressed as: in, is the set of all candidate connecting routes, Indicates each bus connecting route The applicable average headway set, represents the average headway time applicable to each bus connecting route, Based on the average headway setting The set of all feasible timetable options; The service arc set in the passenger allocation space-time network is expressed as: in, Indicates each connection service plan The corresponding service arc set, the corresponding arc capacity is set to the maximum passenger capacity of the bus C bus .

3. The method according to claim 2, characterized in that The task arc set in the bus scheduling spatiotemporal network is expressed as: in, Indicates each service plan The collection of all quest arcs below.

4. The method according to claim 2, characterized in that: Also includes: Defining binary decision variables If passenger p is assigned to the arc (i,j,t,t′,m,m′) in the space-time network, otherwise Define the binary decision variable y k , if vehicle scheduling scheme k is selected, y k =1, otherwise y k =0; Define the binary decision variable z r,h,l , if the bus service solution (r,h,l) is selected as part of the feeder bus service design result, z r,h,l =1, otherwise z r,h,l =0; Among the above decision variables, and k They are used to characterize the path choices of passengers and buses respectively. The internal connection and mutual influence between the two are expressed through the coupling variable z r,h,l Establish; Based on the above decision variables, network flow balance constraints, capacity constraints and bus service feasibility constraints are respectively formulated in the passenger space-time network.

5. The method according to claim 4, characterized in that The network flow balance constraint is used to ensure that each passenger The path feasibility is as follows: Among them, (o p ,at p ,1) and (d p ,T,1) respectively represent the starting point and virtual destination of passenger p in the network, represents the set of all passengers; The capacity constraint is used to limit the number of passengers that can be allocated on some arcs in the passenger space-time network. The platform capacity constraint is expressed as: The subway capacity constraint is expressed as: The bus capacity constraint is expressed as: Among them, the binary indicator parameter It is used to describe whether the potential bus service arc is available. If the service plan (r, h, l) corresponds to the potential bus service arc (i, j, t, t′, m, m′), then otherwise when When , the number of passengers that the corresponding bus service arc can accommodate is equal to the bus capacity, and the other irrelevant arcs do not allow any passengers to choose. When , it means that the connecting bus service is not available and all bus service arcs are not feasible; The feasibility constraints of the public transportation service are used to ensure the feasibility of the public transportation service plan. The specific constraints are as follows: in, It is a common indicator parameter in the set covering problem constraints, indicating whether the vehicle scheduling solution k can cover the task arc (i, j, t, t′).

6. The method according to claim 5, characterized in that The shuttle bus service design and vehicle scheduling collaborative optimization model is expressed as: Among them, c k is the operating cost of vehicle scheduling scheme k, c i,j,t,t′,m,m′ is the passenger travel cost of arc (i, j, t, t′, m, m′), ω1 and ω2 are the weight coefficients of passenger travel cost and vehicle scheduling cost respectively, and the decision variable They are represented by vectors x, y, and z respectively.

7. The method according to claim 5, characterized in that The Lagrangian relaxation problem is as follows: The Lagrangian relaxation problem is decomposed into three tractable sub-problems, including the passenger assignment sub-problem, the vehicle scheduling sub-problem and the bus service design sub-problem. The objective function is as follows: Among them, α i,j,t,t′,m,m′ , β i,j,t,t′ , γ i,j,t,t′,m,m′ and δ i,j,t,t′,m,m′ is the Lagrange multiplier, {β i,j,t,t′ } (i,j,t,t′)∈Ω , They are represented by vectors α, β, γ, and δ respectively, and the arc (i, j, t, t′, m, m′) in the passenger allocation space-time network is denoted as τ, and the arc (i, j, t, t′) in the bus scheduling space-time network is denoted as τ′.

8. The method according to claim 7, characterized in that The passenger assignment subproblem is expressed as: in, represents the generalized arc cost; The passenger assignment subproblem is transformed into a series of path subproblems related to passenger assignment. The passenger routing subproblem can be defined as follows:

9. The method according to claim 7, characterized in that: The vehicle scheduling subproblem is expressed as: in, represents the generalized dispatch cost; The optimal objective function value can be calculated by the following formula:

10. The method according to claim 7, characterized in that The bus service design sub-problem is expressed as:

Citation Information

Cited By

  • Airport ferry vehicle scheduling organization method considering service delay

    CN119886666A

  • Bus turnover plan optimization method considering bus number task allocation fairness

    CN119940849A

  • Cooperative operation optimization method for bus and subway integrated network

    CN121072915A

  • Online car-hailing and bus cooperative scheduling method and system based on cloud scheduling platform

    CN122551604A

  • A method and system for collaborative scheduling of online car-hailing and public transportation based on a cloud scheduling platform

    CN122551604B