Public traffic network optimization method considering station position and riding comfort

Abstract

The invention provides a public traffic network optimization method considering station positions and riding comfort, and the method employs a BTNDP model of the following equation (1): Ctotal = C1 * Ctraveler + C2 * Coperator (1). According to the method, various factors such as line design, service frequency, station positions, elastic demands and in-vehicle congestion can be considered at the same time; the bus route layout, the service frequency and the station positions can be optimized at the same time in a grid street network and a demand mode of many-to-many demand heterogeneous distribution, and the total cost of users and bus operators can be reduced to the maximum extent.

Description

Technical Field

[0001] The invention belongs to the technical field of public transportation, and in particular is a method for optimizing a public transportation network taking into account station locations and riding comfort. Background Art

[0002] The goal of public transportation agencies is to provide efficient transportation networks and services to maximize service levels and minimize costs. Given the high costs involved in building and operating rail transit lines, it is not feasible to achieve comprehensive coverage of urban rail transit services in every corner of the city. Urban bus networks are an important supplement to rail transit services. The inherent characteristics of bus services enable public transportation agencies to attract potential passengers by introducing new lines or expanding existing lines under limited budgets, thereby improving operational performance.

[0003] The objective of the Bus Network Development Problem (BTNDP) is to find the optimal bus network configuration for a target service area. Optimization is generally achieved by minimizing a combination of various costs such as user costs and operator costs. During this process, different types of variables are determined such as service frequency, station locations, route layout, bus demand, etc.

[0004] In the process of transportation planning, the existing bus optimization methods are mainly divided into the following categories:

[0005] Route design and service frequency optimization: This type of methods mainly focuses on designing bus routes and determining service frequencies to minimize total travel time and operating costs. For example, Szeto and Wu proposed a model for optimizing bus route layout and service frequency under fixed bus demand, with the goal of minimizing total travel time and the number of transfers. This type of method usually involves complex mathematical models and algorithms to find the optimal solution under various constraints.

[0006] Station location optimization: This type of approach focuses on determining the optimal location of bus stops to improve the accessibility and service efficiency of the bus system. Chien and Yang developed a scheme to calculate the station access cost for heterogeneous distribution demand, while Fan and Machemehl proposed a concept of distributed nodes to predetermine bus stops. Station location optimization needs to consider multiple factors, such as the walking distance of passengers, the coverage of stations, and the connectivity between stations.

[0007] Traffic Assignment Model (TAM): This type of method is used to simulate and analyze the travel choice behavior of public transportation users, including collinear problems and transfer problems. The model of Arbex et al. uses the multinomial logit (MNL) model to consider transfers between different lines. The research of Chriqui and Robillar pointed out that the boarding behavior of passengers is based on the purpose of minimizing the total travel time. TAM needs to process a large amount of data and parameters to accurately simulate the behavior of passengers and the operation of the system.

[0008] Elastic demand and in-vehicle crowding considerations: Some studies incorporate elastic demand and in-vehicle crowding into bus optimization models to more accurately predict bus demand and optimize bus services. Huang and Liu proposed a two-stage multinomial logit-proportional model (MNL-PM) to solve the problem of connecting bus networks and consider three travel modes: walking, car, and bus. This type of model needs to consider passengers' sensitivity to service quality and the competitive relationship between different travel modes.

[0009] Comprehensive optimization methods: In addition to the above methods, some studies have attempted to comprehensively consider multiple optimization objectives and constraints. For example, factors such as line design, service frequency, station location, and demand elasticity are combined to build a more complex and comprehensive optimization model. This type of research usually requires the combination of advanced computing technology and optimization algorithms to cope with the challenges of high dimensionality and complexity.

[0010] In general, existing bus optimization methods focus on different aspects, but few models can simultaneously consider multiple factors such as route design, service frequency, station location, elastic demand, and on-board crowding. These models that comprehensively consider multiple factors are of great significance for improving the overall efficiency and service level of the bus system. However, due to the complexity and diversity of the problems, these models still face many challenges and problems in practical applications. For example, how to accurately obtain and process large-scale travel data, how to perform real-time optimization in a dynamic and uncertain environment, and how to comprehensively consider various operational and management goals are all key issues that need to be solved. Summary of the invention

[0011] In order to solve the above technical problems, the purpose of the present invention is to provide a new BTNDP model, thereby minimizing the total cost of passengers and service providers.

[0012] The present invention provides a method for optimizing a public transportation network taking into account station locations and riding comfort, using the BTNDP model of the following equation (1):

[0013] C total =C1*C traveler +C2*C operater(1).

[0014] The traveler cost can be expressed as:

[0015] C traveler =C walk_only +C auto_only +C bus (2).

[0016] When a traveler chooses to walk to connect the starting point and the end point, the total cost of walking is:

[0017]

[0018] When a traveler chooses a car as their mode of travel, the total cost of the car is:

[0019]

[0020] For passengers who choose bus as their travel mode, the cost of bus is:

[0021]

[0022] In formula (5) It represents the walking connection cost of passengers who choose public transportation as their travel mode from the centroid i of the departure point to the bus stop r where they get on the bus; represents the “last kilometer” walking connection cost from the bus stop s where passengers get off to the destination centroid j; the second part of formula (5) represents the cost of passengers waiting for the bus, Indicates the departure interval of the first bus line; represents the departure interval of the transfer bus line (if there is a transfer); in the third part, the number of people boarding at station r for trip l and the bus service frequency Proportional, represents the percentage of passengers who choose itinerary l; the length of the route connecting bus stations r and s on itinerary l can be expressed as is predetermined and can be obtained from a given bus route m. For each trip connecting stations r and s, the maximum number of transfers is one, and transfers can only be made between different bus routes. Therefore, for any given pair of isl, there is

[0023]

[0024] The operating cost can be expressed as:

[0025]

[0026] Operating costs refer to the expenses incurred in operating bus services. These expenses come from vehicle operation, maintenance and management. In this model, all operation-related costs are aggregated and measured in terms of the mileage traveled by each vehicle. The total round-trip time for bus line k is 2Γ k / V bus , m k F k (2Γ k / V bus ) represents the total number of buses required on route k per hour.

[0027] The final objective function of the model is:

[0028]

[0029] The method described includes the following constraints for establishing bus stops:

[0030] Constraint (8) ensures that for a given origin and destination (OD point), the number of travel bus stops that passengers can choose near the origin is 1: i=1,2,…,I; j=1,2,…,J(8);

[0031] Constraint (9) means that for OD, when choosing a boarding bus stop r, at least one alighting bus stop s can be found near destination j: i=1,2,…,I; j=1,2,…,J(9);

[0032] Constraint (10) indicates that if there is no available bus stop for boarding near the starting point i, there will be no bus stop for getting off: i=1,2,…,I; j=1,2,…,J(10);

[0033] Constraints (11) and (12) are used to ensure that all boarding and final alighting bus stops are within the maximum walking distance Γ between the bus stop origin and the destination. bwmax Within the threshold:

[0034] i=1,2,…,I; j=1,2,…,J; r∈R i (11)

[0035] i=1,2,…,I; j=1,2,…,J; s∈S i (12);

[0036] Constraint (13) ensures that if a boarding station r and an alighting station s are chosen simultaneously, there is at least one trip that can connect r and s: i=1,2,…,I; j=1,2,…,J; r∈R i ,s∈S i (13);

[0037] Constraint (14) indicates that the maximum number of trips a passenger can choose is Lmax: i=1,2,…,I; j=1,2,…,J (14);

[0038] Constraint (15) states that the in-car distance of all selected routes is always less than the total walking distance: for l∈L rs ,

[0039] Constraint (16) assumes that the walking distance between the final drop-off station and the destination is less than the distance between the pick-up station and the destination:

[0040] The method described includes the following route and fleet constraints:

[0041] Constraint (17) sets the length range that route k must meet when it is included in the candidate bus route set M. The detour coefficient of a bus route is the ratio of the bus route length to the Manhattan distance between the starting point and the end point: Γ min ≤Γ k ≤Γ max , k=1,2,…,M (17);

[0042] Constraint (18) ensures that the detour coefficient of bus routes in set M is less than 1.4: Γ k ≤Γ′ k ≤1.4, k=1,2,…,M (18);

[0043] Constraint (19) ensures that the ratio of the total travel distance of the route to the straight-line distance (Manhattan distance) between i and j is less than 1.8: for l∈L rs (19);

[0044] because means that the passenger has chosen at least two stops on route k from i to j. Constraint (20) ensures that on any route in subset m, the passenger can choose at least two bus stops: k∈M m (20);

[0045] Constraint (21) limits the total number of bus routes:

[0046] Constraint (22) is a restriction on the availability of the fleet size:

[0047] In constraint (23), represents the total demand of bus line k in both directions per hour; C is the passenger capacity of the bus; F k is the hourly service frequency on bus line k, ρ is the policy load factor, and the inequality ensures that the capacity of the line is greater than or equal to the demand: k∈M m (twenty three);

[0048] Constraint (24) is a constraint on the bus service frequency in the present invention: 1≤F k ≤12,k=1,2,…,M m (twenty four).

[0049] The solution algorithm of the model is a heuristic algorithm, which consists of three main parts: initial preparation procedure (IPP), network feature determination procedure (NFDP) and solution search procedure (SSP). The decision variables include bus route selection m k , service frequency F k and bus stop / route selection related variables. The latter three variables are directly determined by the two-stage multinomial logit model proposed by NFDP. After obtaining all network characteristics, SSP will apply genetic algorithm (GA) to optimize the combination of bus routes and service frequencies to minimize the total cost for passengers and service providers. The final output result is a set of optimized bus routes as well as the characteristics of service frequency and station locations.

[0050] In the initial preparation procedure step, two procedures need to be performed: determining the candidate bus station set N and developing the candidate bus route set M. Two types of nodes are considered as candidate bus stations, including intersection nodes and distribution nodes.

[0051] The network characteristics determination procedure adopts the two-stage polynomial logarithmic bus assignment model (2-Stage MNL Model), which combines the "attractive set" and random utility maximization (RUM) strategies. The model divides the traffic assignment problem into two levels: centroid level and station level:

[0052] At the centroid level, passengers at the centroid i and heading to the centroid j have three travel options: walking, car, and bus. They must first make a decision on the travel mode. The centroid travel mode selection is based on the multinomial logit model (MNL). It is assumed that in the process of choosing a travel mode at the centroid, passengers always take the cost of the feasible route with the maximum public transportation utility nearby as the expected cost of the public transportation mode. bwmaxIn the area, all bus stops near the centroid of the departure area are considered as candidate boarding bus stops; bwmax All bus stops near the centroid of the destination in the region are also considered as candidate destination alighting bus stops;

[0053] Any route connecting the candidate boarding point and the final drop-off point is a feasible route. In addition, it is assumed that at the center point level, the passengers at the center point do not know the crowding of any bus, which means that the crowding is not considered at this stage. It is assumed that the paths of walking and car modes are the same and are determined by the Manhattan distance between the origin and the destination. The utility function of these alternatives can be expressed as follows:

[0054]

[0055] Dummy variables Indicates whether the passenger from centroid i to j chooses station r. If r is chosen, the value of this dummy variable is 1, otherwise it is 0. Since the bus service utility in the MNL model used at the centroid level is based on the feasible journey with the maximum utility value, the location of the first boarding station for passengers departing from centroid i can be obtained by Determine, among which is the utility set of all feasible bus trips for pairs ij, if the demand for OD pairs from i to j is Q ij , then the number of passengers who choose different travel modes can be determined by the following equation:

[0056]

[0057] After determining the number of passengers and the location of the boarding stations for different travel modes, the second stage of the MNL model is applied: at the station level, in order to solve the public line problem, it is assumed that bus users can obtain bus arrival information from the LCD screen installed at the station. Therefore, the perceived waiting time is determined and the utility of each trip can be estimated. The attraction set is based on the first L passengers with the shortest waiting time at the first boarding station. max If there is a transfer, it is assumed that bus users cannot predict the expected waiting time at the transfer station. Since the in-vehicle time (IVT) of the trips with concentrated attraction may be different, the share of bus users in these trips is no longer proportional to its service frequency. Instead, the second stage of the MNL model is applied to determine the number of its passengers traveling in different modes according to the utility of the trip;

[0058] In order to incorporate the in-vehicle crowding penalty in the second stage of the proposed MNL model, a one-step nonlinear time multiplication function f(ρ) is developed in Eq. (31):

[0059]

[0060] Therefore, the public transportation utility function used in the second stage can be expressed as:

[0061]

[0062] Including the in-vehicle crowding effect in the MNL model will affect the choice of bus trips. Similarly, the choice of bus trips will again affect the degree of crowding in the car, forming a feedback loop. After each iteration, the difference in bus trip choices at each station is evaluated. If the maximum difference in trip choices between two iterations is less than a threshold, these trip choices are accepted.

[0063] The beneficial effects of the present invention are as follows: a public transportation network optimization method of the present invention that takes into account station locations and riding comfort can simultaneously take into account multiple factors such as line design, service frequency, station location, elastic demand and in-vehicle congestion, and can simultaneously optimize bus line layout, service frequency, and station location under a gridded street network and a multi-to-multi demand heterogeneous distribution demand model, thereby minimizing the total cost for users and bus operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the bus line network configuration of the present invention;

[0065] Figure 2 It is the flow chart of the second stage MNL traffic assignment model of the present invention. DETAILED DESCRIPTION

[0066] The present invention is further described in detail below in conjunction with the embodiments. It should be noted that the following description is merely an example of the technical solutions claimed in the present invention and does not constitute any limitation to these technical solutions.

[0067] The present invention provides a public transportation network optimization method considering station locations and riding comfort, using the BTNDP model of the following equation (1):

[0068] C total =C1*C traveler +C2*C operater (1).

[0069] The traveler cost is defined by the cost of the travel mode chosen by each traveler. Three modes are considered in the study and the traveler cost can be expressed as:

[0070] C traveler =C walk_only +C auto_only +C bus (2).

[0071] When a traveler chooses to walk to connect the starting point and the end point, the total cost of walking is:

[0072]

[0073] When a traveler chooses a car as their mode of travel, the total cost of the car is:

[0074]

[0075] For those passengers who choose bus as a travel mode, the cost of bus can be measured by the time of walking to the bus station, the waiting time (including the waiting time of transferring buses), and the riding time. Since the bus ticket cost is transferable between user cost and bus operation cost, the ticket cost is not included in this formula. In this invention, transfers between different bus routes are considered, so we use "trip" to represent various alternative bus services. A trip connecting two designated bus stops can be a single route or a combination of two routes with one transfer.

[0076]

[0077]

[0078] In formula (5) It represents the walking connection cost of passengers who choose public transportation as their travel mode from the centroid i of the departure point to the bus stop r where they get on the bus; represents the “last kilometer” walking connection cost from the bus stop s where passengers get off to the destination centroid j; the second part of formula (5) represents the cost of passengers waiting for the bus, Indicates the departure interval of the first bus line; represents the departure interval of the transfer bus line (if there is a transfer); in the third part, the number of people boarding at station r for trip l and the bus service frequency Proportional, represents the percentage of passengers who choose itinerary l; the length of the route connecting bus stations r and s on itinerary l can be expressed as is predetermined and can be obtained from a given bus route m. For each trip connecting stations r and s, the maximum number of transfers is one, and transfers can only be made between different bus routes. Therefore, for any given pair of isl, there is

[0079] The operating cost can be expressed as:

[0080]

[0081] Operating costs refer to the expenses incurred in operating bus services. These expenses come from vehicle operation, maintenance and management. In this model, all operation-related costs are aggregated and measured in terms of the mileage traveled by each vehicle. The total round-trip time for bus line k is 2Γ k / V bus , m k F k (2Γ k / V bus ) represents the total number of buses required on route k per hour.

[0082] The final objective function of the model is:

[0083]

[0084]

[0085] Here are two sets of constraints for the above model:

[0086] Constraints for establishing bus stops:

[0087] Constraint (8) ensures that for a given origin and destination (OD point), the number of travel bus stops that passengers can choose near the origin is 1: i=1,2,…,I; j=1,2,…,J(8);

[0088] Constraint (9) means that for OD, when choosing a boarding bus stop r, at least one alighting bus stop s can be found near destination j: i=1,2,…,I; j=1,2,…,J(9);

[0089] Constraint (10) indicates that if there is no available bus stop for boarding near the starting point i, there will be no bus stop for getting off: i=1,2,…,I; j=1,2,…,J(10);

[0090] Constraints (11) and (12) are used to ensure that all boarding and final alighting bus stops are within the maximum walking distance Γ between the bus stop origin and the destination. bwmax Within the threshold:

[0091] i=1,2,…,I; j=1,2,…,J; r∈R i (11)

[0092] i=1,2,…,I; j=1,2,…,J; s∈S i (12);

[0093] Constraint (13) ensures that if a boarding station r and an alighting station s are chosen simultaneously, there is at least one trip that can connect r and s: i=1,2,…,I; j=1,2,…,J; r∈R i ,s∈S i (13);

[0094] Constraint (14) indicates that the maximum number of trips a passenger can choose is Lmax: i=1,2,…,I; j=1,2,…,J(14);

[0095] Constraint (15) states that the in-car distance of all selected routes is always less than the total walking distance: for l∈L rs ,

[0096] Constraint (16) assumes that the walking distance between the final drop-off station and the destination is less than the distance between the pick-up station and the destination:

[0097] Routing and Fleet Constraints:

[0098] Constraint (17) sets the length range that route k must meet when it is included in the candidate bus route set M. The detour coefficient of a bus route is the ratio of the bus route length to the Manhattan distance between the starting point and the end point: Γ min ≤Γ k ≤Γ max , k=1,2,…,M (17);

[0099] Constraint (18) ensures that the detour coefficient of bus routes in set M is less than 1.4: Γ k ≤Γ′ k ≤1.4, k=1,2,…,M (18);

[0100] Constraint (19) ensures that the ratio of the total travel distance of the route to the straight-line distance (Manhattan distance) between i and j is less than 1.8: for l∈L rs (19);

[0101] because means that the passenger has chosen at least two stops on route k from i to j. Constraint (20) ensures that on any route in subset m, the passenger can choose at least two bus stops: k∈M m (20);

[0102] Constraint (21) limits the total number of bus routes:

[0103] Constraint (22) is a restriction on the availability of the fleet size:

[0104] In constraint (23), represents the total demand of bus line k in both directions per hour; C is the passenger capacity of the bus; F k is the hourly service frequency on bus line k, ρ is the policy load factor, and the inequality ensures that the capacity of the line is greater than or equal to the demand: k∈M m (twenty three);

[0105] Constraint (24) is a constraint on the bus service frequency in the present invention: 1≤F k ≤12,k=1,2,…,M m (twenty four).

[0106] The solution algorithm of the model is a heuristic algorithm, which consists of three main parts: initial preparation procedure (IPP), network feature determination procedure (NFDP) and solution search procedure (SSP). The decision variables include bus route selection m k , service frequency F k and bus stop / route selection related variables. The latter three variables are directly determined by the two-stage multinomial logit model proposed by NFDP. After obtaining all network characteristics, SSP will apply genetic algorithm (GA) to optimize the combination of bus routes and service frequencies to minimize the total cost for passengers and service providers. The final output result is a set of optimized bus routes as well as the characteristics of service frequency and station locations.

[0107] In the initial preparation procedure step, two procedures need to be performed: determining the candidate bus station set N and developing the candidate bus route set M. Two types of nodes are considered as candidate bus stations, including intersection nodes and distribution nodes.

[0108] The bus allocation mode is a key component of the network characteristic determination procedure. The present invention considers three modes in the many-to-many demand model. At the centroid level, the travel demand starting from the starting point faces three optional modes, including walking, bus and car. The walking mode refers to people only choosing walking as a means of travel. The bus mode refers to passengers in the centroid area who use walking to enter and exit the bus station, and most of their trips are covered by bus services. The car mode refers to the use of cars only when traveling. The development of "attractor sets" and the application of RUM are two types of methods to solve the main obstacles of the bus allocation problem.

[0109] In most previous literature, all bus choices are considered as walking and car choices, and the multinomial logit model (MNL) can be applied. Obviously, this assumption violates the independent and identically distributed sample (iid) assumption of MNL. Many previous literatures assume that the passenger percentage of Line 1 is the ratio of the frequency of Line 1 to the total frequency of bus lines. However, in reality, the IVT of each bus line between two stations may be different, so the assumption of the same in-vehicle time may be violated. To overcome the above problems, the network characteristic determination procedure adopts a two-stage multinomial logarithmic bus assignment model (2-Stage MNL Model), which combines the "attractive set" and random utility maximization (RUM) strategies. The model divides the traffic assignment problem into two levels: centroid level and station level:

[0110] At the centroid level, passengers at the centroid i and heading to the centroid j have three travel options: walking, car, and bus. They must first make a decision on the travel mode. The centroid travel mode selection is based on the multinomial logit model (MNL). It is assumed that in the process of choosing a travel mode at the centroid, passengers always take the cost of the feasible route with the maximum public transportation utility nearby as the expected cost of the public transportation mode. bwmax In the area, all bus stops near the centroid of the departure area are considered as candidate boarding bus stops; bwmax All bus stops near the centroid of the destination in the region are also considered as candidate destination alighting bus stops;

[0111] Any route connecting the candidate boarding point and the final drop-off point is a feasible route. In addition, it is assumed that at the center point level, the passengers at the center point do not know the crowding of any bus, which means that the crowding is not considered at this stage. It is assumed that the paths of walking and car modes are the same and are determined by the Manhattan distance between the origin and the destination. The utility function of these alternatives can be expressed as follows:

[0112]

[0113] Table 1 Variable symbols of travel mode choice model

[0114]

[0115]

[0116] Table 2 Travel mode selection coefficient

[0117]

[0118] Dummy variables Indicates whether the passenger from centroid i to j chooses station r. If r is chosen, the value of this dummy variable is 1, otherwise it is 0. Since the bus service utility in the MNL model used at the centroid level is based on the feasible journey with the maximum utility value, the location of the first boarding station for passengers departing from centroid i can be obtained by Determine, among which is the utility set of all feasible bus trips for pairs ij, if the demand for OD pairs from i to j is Q ij , then the number of passengers who choose different travel modes can be determined by the following equation:

[0119]

[0120]

[0121] After determining the number of passengers and the location of the boarding stations for different travel modes, the second stage of the MNL model is applied: at the station level, in order to solve the public line problem, it is assumed that bus users can obtain bus arrival information from the LCD screen installed at the station. Therefore, the perceived waiting time is certain and the utility of each trip can be estimated. Figure 1 The bus service configuration at the station level is shown. The attraction set is based on the first L with the shortest waiting time at the first boarding station. max If there is a transfer, it is assumed that bus users cannot predict the expected waiting time at the transfer station. Since the in-vehicle time (IVT) of the trips with concentrated attraction may be different, the share of bus users in these trips is no longer proportional to its service frequency. Instead, the second stage of the MNL model is applied to determine the number of its passengers traveling in different modes according to the utility of the trip;

[0122] In the present invention, in-vehicle congestion is considered as a negative utility in the second stage of the MNL model. The additional value of the ride time is expressed as a congestion penalty. There are three forms of congestion penalties used: time multiplier, monetary value per unit time, and monetary value per ride. Since the time multiplier is more transferable in different contexts, it has become the main method. At present, the mathematical formula form representing the time multiplier function is still under study. However, the linear function form has been criticized by many literatures. In essence, the relationship between congestion penalty and congestion level is nonlinear. In addition, when the congestion level reaches a certain threshold, the perception of congestion will deteriorate sharply.

[0123] In order to incorporate the in-vehicle crowding penalty in the second stage of the proposed MNL model, a one-step nonlinear time multiplication function f(ρ) is developed in Eq. (31):

[0124]

[0125] Therefore, the public transportation utility function used in the second stage can be expressed as:

[0126]

[0127] Including the in-vehicle crowding effect in the MNL model will affect the choice of bus trips. Similarly, the choice of bus trips will again affect the degree of crowding in the car, forming a feedback loop. After each iteration, the difference in bus trip choices at each station is evaluated. If the maximum difference in trip choices between two iterations is less than a threshold, these trip choices are accepted.

[0128] Table 3 Parameters used in this paper

[0129]

[0130]

[0131]

[0132]

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for optimizing a public transportation network taking into account station locations and riding comfort, characterized in that The BTNDP model is adopted as follows: C total =C1*C traveler +C2*C operater (1)。 2. The method according to claim 1, characterized in that The traveler cost can be expressed as: C traveler =C walk_only +C auto_only +C bus (2).

3. The method according to claim 1, characterized in that When a traveler chooses to walk to connect the starting point and the end point, the total cost of walking is: When a traveler chooses a car as their mode of travel, the total cost of the car is: For passengers who choose bus as their travel mode, the cost of bus is: In formula (5) It represents the walking connection cost of passengers who choose public transportation as their travel mode from the centroid i of the departure point to the bus stop r where they get on the bus; represents the "last kilometer" walking connection cost from the bus stop s where passengers get off to the destination centroid j; the second part of formula (5) represents the cost of passengers waiting for the bus, Indicates the departure interval of the first bus line; represents the departure interval of the transfer bus line (if there is a transfer); in the third part, the number of people boarding at station r for trip l and the bus service frequency Proportional, represents the percentage of passengers who choose itinerary l; the length of the route connecting bus stations r and s on itinerary l can be expressed as is predetermined and can be obtained from a given bus route m. For each trip connecting stations r and s, the maximum number of transfers is one, and transfers can only be made between different bus routes. Therefore, for any given pair of isl, there is 4. The method according to claim 1, characterized in that The operating cost can be expressed as: Operating costs refer to the expenses incurred in operating bus services. These expenses come from vehicle operation, maintenance and management. In this model, all operation-related costs are aggregated and measured in terms of the mileage traveled by each vehicle. The total round-trip time for bus line k is 2Γ k / V bus , m k F k (2Γ k / V bus ) represents the total number of buses required on route k per hour.

5. The method according to claim 1, characterized in that The final objective function of the model is:

6. The method according to any one of claims 1 to 5, characterized in that The constraints for establishing bus stops include the following: Constraint (8) ensures that for a given origin and destination (OD point), the number of travel bus stops that passengers can choose near the origin is 1: Constraint (9) means that for OD, when choosing a boarding bus stop r, at least one alighting bus stop s can be found near destination j: Constraint (10) indicates that if there is no available bus stop for boarding near the starting point i, there will be no bus stop for getting off: Constraints (11) and (12) are used to ensure that all boarding and final alighting bus stops are within the maximum walking distance Γ between the bus stop origin and the destination. bwmax Within the threshold: Constraint (13) ensures that if a boarding station r and an alighting station s are chosen simultaneously, there is at least one trip that can connect r and s: Constraint (14) indicates that the maximum number of trips a passenger can choose is Lmax: Constraint (15) states that the in-car distance of all selected routes is always less than the total walking distance: Constraint (16) assumes that the walking distance between the final drop-off station and the destination is less than the distance between the pick-up station and the destination:

7. The method according to any one of claims 1 to 5, characterized in that The following route and fleet constraints are included: Constraint (17) sets the length range that route k must meet when it is included in the candidate bus route set M. The detour coefficient of a bus route is the ratio of the bus route length to the Manhattan distance between the starting point and the end point: Γ min ≤Γ k ≤Γ max , k=1,2,…,M(17); Constraint (18) ensures that the detour coefficient of bus routes in set M is less than 1.4: Γ k ≤Γ′ k ≤1.4, k=1,2,…,M(18); Constraint (19) ensures that the ratio of the total travel distance of the route to the straight-line distance (Manhattan distance) between i and j is less than 1.8: because means that the passenger has chosen at least two stops on route k from i to j. Constraint (20) ensures that on any route in subset m, the passenger can choose at least two bus stops: Constraint (21) limits the total number of bus routes: Constraint (22) is a restriction on the availability of the fleet size: In constraint (23), represents the total demand of bus line k in both directions per hour; C is the passenger capacity of the bus; F k is the hourly service frequency on bus line k, ρ is the policy load factor, and the inequality ensures that the capacity of the line is greater than or equal to the demand: Constraint (24) is a constraint on the bus service frequency in the present invention: 1≤F k ≤12,k=1,2,…,M m (twenty four).

8. The method according to any one of claims 1 to 5, characterized in that The solution algorithm of the model is a heuristic algorithm, which consists of three main parts: initial preparation procedure (IPP), network feature determination procedure (NFDP) and solution search procedure (SSP). The decision variables include bus route selection m k , service frequency F k and bus stop / route selection related variables. The latter three variables are directly determined by the two-stage multinomial logit model proposed by NFDP. After obtaining all network characteristics, SSP will apply genetic algorithm (GA) to optimize the combination of bus routes and service frequencies to minimize the total cost for passengers and service providers. The final output result is a set of optimized bus routes as well as the characteristics of service frequency and station locations.

9. The method according to claim 8, characterized in that In the initial preparation procedure step, two procedures need to be performed: determining the candidate bus station set N and developing the candidate bus route set M. Two types of nodes are considered as candidate bus stations, including intersection nodes and distribution nodes.

10. The method according to claim 8, characterized in that The network characteristics determination procedure adopts the two-stage polynomial logarithmic bus assignment model (2-Stage MNL Model), which combines the "attractive set" and random utility maximization (RUM) strategies. The model divides the traffic assignment problem into two levels: centroid level and station level: At the centroid level, passengers at the centroid i and heading to the centroid j have three travel options: walking, car, and bus. They must first make a decision on the travel mode. The centroid travel mode selection is based on the multinomial logit model (MNL). It is assumed that in the process of choosing a travel mode at the centroid, passengers always take the cost of the feasible route with the maximum public transportation utility nearby as the expected cost of the public transportation mode. bwmax In the area, all bus stops near the centroid of the departure area are considered as candidate boarding bus stops; bwmax All bus stops near the centroid of the destination in the region are also considered as candidate destination alighting bus stops; Any route connecting the candidate boarding point and the final drop-off point is a feasible route. In addition, it is assumed that at the center point level, the passengers at the center point do not know the crowding of any bus, which means that the crowding is not considered at this stage. It is assumed that the paths of walking and car modes are the same and are determined by the Manhattan distance between the origin and the destination. The utility function of these alternatives can be expressed as follows: Dummy variables Indicates whether the passenger from centroid i to j chooses station r. If r is chosen, the value of this dummy variable is 1, otherwise it is 0. Since the bus service utility in the MNL model used at the centroid level is based on the feasible journey with the maximum utility value, the location of the first boarding station for passengers departing from centroid i can be obtained by Determine, among which is the utility set of all feasible bus trips for pairs ij, if the demand for OD pairs from i to j is Q ij , then the number of passengers who choose different travel modes can be determined by the following equation: After determining the number of passengers and the location of the boarding stations for different travel modes, the second stage of the MNL model is applied: at the station level, in order to solve the public line problem, it is assumed that bus users can obtain bus arrival information from the LCD screen installed at the station. Therefore, the perceived waiting time is determined and the utility of each trip can be estimated. The attraction set is based on the first L passengers with the shortest waiting time at the first boarding station. max If there is a transfer, it is assumed that bus users cannot predict the expected waiting time at the transfer station. Since the in-vehicle time (IVT) of the trips with concentrated attraction may be different, the share of bus users in these trips is no longer proportional to its service frequency. Instead, the second stage of the MNL model is applied to determine the number of its passengers traveling in different modes according to the utility of the trip; In order to incorporate the in-vehicle crowding penalty in the second stage of the proposed MNL model, a one-step nonlinear time multiplication function f(ρ) is developed in Eq. (31): Therefore, the public transportation utility function used in the second stage can be expressed as: Including the in-vehicle crowding effect in the MNL model will affect the choice of bus trips. Similarly, the choice of bus trips will again affect the degree of crowding in the car, forming a feedback loop. After each iteration, the difference in bus trip choices at each station is evaluated. If the maximum difference in trip choices between two iterations is less than a threshold, these trip choices are accepted.

Images