Bus route optimization method considering travel behavior complexity

Abstract

The invention provides a bus route optimization method considering travel behavior complexity, which comprises the step of acquiring bus station information of an existing bus network system, and further comprises the following steps of: constructing a bus route multi-target optimization model according to a bus network target; optimizing the bus route multi-objective optimization model by using a multi-objective particle swarm algorithm; and inputting the bus station information into the optimized bus route multi-objective optimization model and outputting a bus route optimization scheme. The invention provides a bus route optimization method considering travel behavior complexity, two objectives of reducing displacement and social cost are considered at the same time, a bus network structure multi-objective optimization model is established, ticket prices and departure frequencies of different routes are optimized while route design is optimized, and the bus route optimization efficiency is improved. A bus route feasible solution adjustment strategy, a crossover operator based on route similarity and a multi-target particle swarm algorithm are organically combined, and an optimization algorithm solving model is constructed.

Description

Technical Field

[0001] The invention relates to the technical field of bus route optimization, and in particular to a bus route optimization method taking into account the complexity of travel behavior. Background Art

[0002] The design and planning of bus networks have a significant impact on the performance of transportation systems. In response to this problem, scholars have proposed corresponding methods based on different perspectives. The volatility of traffic flow is an important variable affecting the performance of bus networks, and the changes in traffic flow are rooted in the travel behavior of the majority of social members.

[0003] By combing through the existing technical methods, it is not difficult to find two problems. First, in most methods for bus network design, emission reduction is not often used as the Pareto optimal decision-making goal. The coordination relationship between "efficiency" and "emission reduction" lacks a quantitative measurement method. Secondly, the network planning process uses relatively direct strong assumptions to deal with social members at the micro level. For example, it is assumed that travelers have homogeneous travel behavior mechanisms, equal travel lengths, and do not consider the mutual influence between different travel modes. Existing studies have shown that, based on the individual full-chain travel carbon footprint of travelers as the statistical standard, cars, conventional buses, rail transit, etc. all have different carbon emissions. Vehicle types and the busy state of the road network are also important factors affecting emissions. In addition, urban public transportation has typical social public product attributes, with a large number of participants. The interactive emergence of cluster behaviors of a large number of social members also brings uncertainty to the system. It can be seen that the scientific planning of bus networks is a complex optimization process with multivariate hyperparameters, and the above two problems reduce the flexibility and applicability of line optimization schemes in actual scenarios. Therefore, these problems deserve further improvement. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a bus route optimization method that takes into account the complexity of travel behavior, considers both the reduction of displacement and the reduction of social costs, establishes a multi-objective optimization model for the bus network structure, optimizes the ticket prices and departure frequencies of different routes while optimizing the route design, organically combines the bus route feasible solution adjustment strategy, the crossover operator based on route similarity and the multi-objective particle swarm algorithm, and constructs an optimization algorithm solution model.

[0005] The present invention provides a bus route optimization method considering the complexity of travel behavior, including obtaining bus station information of an existing bus line network system, and further comprising the following steps:

[0006] Step 1: Construct a multi-objective optimization model for bus routes based on the traffic network objectives;

[0007] Step 2: Use a multi-objective particle swarm algorithm to optimize the multi-objective optimization model of the bus route;

[0008] Step 3: Input the bus stop information into the optimized bus route multi-objective optimization model and output the bus route optimization plan.

[0009] Preferably, the bus network system G = (V, A), consisting of N bus stops, V N represents the set of stations, A represents the set of road sections between stations (a∈A), R represents the set of bus routes, each route consists of stations and road sections, D represents the OD demand matrix between bus routes and stations, D i,j is the daily average travel demand between stations i, j, i, j∈V N , between the starting and ending points i and j, the traveler can choose a private car c or a bus line r. In the daily travel activity time period [0, T], the traveler has s departure time options, expressed as [τ1=0,τ2=T / S,τ3=2T / S,…,τ S =(S-1)T / S], for Let q i,j (t,τ) represents the number of people who choose to travel at time τ on the tth day between stations i and j. represents the number of people who choose to depart at time τ on day t between stations i and j and choose travel mode m, then

[0010] In any of the above solutions, preferably, the traffic network target includes:

[0011] (1) Minimize the social costs of the bus network;

[0012] (2) Minimize carbon emissions from the bus network;

[0013] The optimization variable is X = (x, p, f), where x represents the route matrix, each row of the matrix represents the stop plan of a route, x determines the basic bus route system R', p = (p1, ..., p R′ ) represents the fare column vector, f=(f1,…,f R′ ) represents the departure frequency column vector, represents the utility of choosing travel mode m on day t and departing at time τ, let represents the bus operating cost, Represents the per capita carbon emissions of new energy buses and traditional energy cars.

[0014] In any of the above solutions, it is preferred that the bus route multi-objective optimization model is:

[0015]

[0016] in, represents the passenger flow constraint, x∈Ψ represents the feasible constraint of the bus route, p r,min ≤p r ≤p r,max represents the fare constraint, f r,min ≤f r ≤f r,max represents the capacity constraint, Represents the social welfare constraints of public transportation.

[0017] In any of the above schemes, preferably, step 2 includes setting the particle swarm size to N p , the optimization variable corresponding to particle i' is the triple X i′ =(x i′ ,p i′ ,f i′ ), where x i′ 、p i′ 、f i′ They correspond to the route matrix, fare combination, and departure frequency combination represented by particle i' respectively.

[0018] In any of the above schemes, preferably, step 2 comprises the following sub-steps:

[0019] Step 21: Set the global non-dominated solution set A g , set a non-dominated solution set A for each particle i' in the population i′ , build triple X i′ =(x i′ ,p i′ ,f i′ ), initialize the feasible solution and ensure x i′ feasibility;

[0020] Step 22: Calculate X i′ =(x i′ ,p i′ ,f i′ )The final OD matrix corresponding to;

[0021] Step 23: Calculate the objective function value, given by q m (t,τ) calculates the objective function value corresponding to individual i′ The solution F corresponding to each particle * (X i′ ) are added to the non-dominated solution set A i′ With A g , add the rule: for the non-dominated solution set A (A = A i′ ,Ag ) solution A a , if there is Then delete A in A a ; if F * (X i′ ) is not dominated by any solution in A, then F * (X i′ ) add A, where Indicates a dominant relationship, that is And in F1 * and There is at least one that satisfies F * (X i′ )<A a ;

[0022] Step 24: Calculate constraint violation values in, <x>It means if x≤0, then <x>=0, otherwise <x>=|x|;

[0023] Step 25: Calculate solution set A i′ With A g The crowding distance of each group of solutions in Indicates A g Objective function In descending order, the crowding distance of the i′th group of solutions is:

[0024] Step 26: The crowding distance corresponding to individual i′ is Select excellent individuals according to probability, in A i′ With A g In the equation, the probability of the solution corresponding to individual i′ being selected is In A i′ The individual selected from A is denoted as i″. g The individual selected is denoted as i″′;

[0025] Step 27: Crossover operator based on line similarity;

[0026] Step 28: Set the mutation probability to p′ mu , individual X i′ The line matrix x i′ According to the probability p′ mu Regenerate. If the newly generated line matrix has sites that cannot be connected to the current line set, then x i′ Keep unchanged, the price column vector p i′ and the departure frequency column vector f i′ According to the probability p′ mu Re-randomize the value within the constraints;

[0027] Step 29: Check whether the termination condition is met. If the termination condition is met, stop. Otherwise, return to step 22.

[0028] In any of the above solutions, preferably, the termination condition is that the Pareto front formed by the optimal solution cannot be further advanced.

[0029] In any of the above solutions, preferably, the method for initializing the feasible solution comprises the following sub-steps:

[0030] Step 2101: In x i′ When establishing the first bus route, a station in the road network is randomly selected as the starting point;

[0031] Step 2102: For a bus route, traverse the network of stations that have a road connection with the i-1th station and have not been used, and randomly select one of them as the i-th station; if such a station does not exist, replace the currently formed route (s′1, s′2, …, s′ i-1 ) is flipped to become (s′ i-1 ,…,s′2,s′1), repeat this step until at least one traversal condition is met, and the site set is stored in x in the form of a row vector. i′ ;

[0032] Step 2103: For the kth line, randomly select a station from the generated lines 1 to k as the starting point, and repeat step 212; when the number of lines has reached the maximum value, if the line matrix x i′ All stations in the road network have been included, the algorithm stops, x i′ That is a feasible line matrix;

[0033] Step 2104, for an infeasible route, when x i′ When it is an infeasible solution, the line matrix x i′ If all the stations in the road network are not included, then the set of stations that no bus line passes through is represented as (s′1, s′2, …, s′ m ), the current line set is represented as (r′1,r′2,…,r′ k );

[0034] Step 2105: Select a set of lines (r′1, r′2, …, r′ k ) is the current line r′ i ;

[0035] Step 2106: Select a site set (s′1, s′2, …, s′ m ) is the current missing site s′ j ;

[0036] Step 2107: If line r′ i Contains the site s′ j Connected sites s′ x , go to step 2108, otherwise, go to step 2109;

[0037] Step 2108: Line r′ i is divided into two parts r′ r,1 and r′ i,2 ;

[0038] If the site s′ x In r′ i,1 At the end of the line, flip the line, station s′ x Become the head end and connect r′ to the new tail end i,2 The stations in the form a new line r″ i,1 ;

[0039] If the site s′ x In r′ i,2 At the beginning of the line, turn the line over, station s′ x Become the tail end and connect r′ at the new head end i,1 The stations in the form a new line r″ i,2 ;

[0040] Comparison line r″ i,1 and r″ i,2 The length of r′ i =max{r″ i,1 ,r″ i,2 }, replace s′ j From (s′1,s′2,…,s′ m ) is removed, and go to step 2105;

[0041] Step 2109: If the current line r′ i Not (r′1,r′2,…,r′ k ), let i=i+1, go to step 2107, otherwise, go to step 2110;

[0042] Step 2110: traverse the line matrix x i′ If there are still sites that cannot access the current line set, then x i′ If it is an infeasible solution, discard it and start again to generate an initial feasible solution.

[0043] In any of the above solutions, preferably, the traversal condition includes:

[0044] 1) The degree of the stations at both ends of the line is 1;

[0045] 2) Continuing to add sites will form a loop;

[0046] 3) The line length reaches the set maximum value.

[0047] In any of the above solutions, preferably, step 22 includes the following sub-steps:

[0048] Step 221: Based on the given x i′ , search for feasible routes and reasonable transfer routes between each OD pair.

[0049] Step 222: For a given p i′ With f i′ , so that the travel time between each station is the sum of the zero flow time of each section, the number of iterations k′=0, and the passenger flow OD matrix between stations is calculated

[0050] Step 223: Pass Calculate the traffic volume of the road section and Then update the travel time to get

[0051] Step 224: Calculate departure time utility;

[0052] Step 225: Update the passenger flow OD matrix between sites to obtain

[0053] Step 226: Set termination conditions If this condition is met, the iteration stops. That is the current feasible solution X i′ The corresponding OD matrix, otherwise the iteration number k′=k′+1, re-execute step 223.

[0054] In any of the above solutions, preferably, step 122 includes establishing a system including N for each OD pair in the bus network. agent The cell space Ω of each traveler cell , each traveler agent individual (x, y) in this space has its own cognitive structure, let Represents the cellular neural network weight matrix (n layers) of individual (x, y), let represents the probability that an individual (x, y) chooses travel mode m between stations i and j. The decision evolution process of the traveler (x, y) can be expressed as:

[0055] Step 2221: Each traveler initializes its own neural network weight matrix

[0056] Step 2222: On day t, for each travel agent between stations i and j, the net utility of each travel mode is calculated as Input its own neural network, through the weight matrix The probability of choosing different travel modes is obtained through calculation processing. Traveler agent based Choose a travel method;

[0057] Step 2223: Each traveler agent obtains its own travel experience

[0058] Step 2224: Each traveler in the cellular space searches for the travel utility (experience) value TE within the neighborhood x,y The largest source of information, denoted as (x * ,y * ), let p cr represents the interaction intensity, the relationship between the traveler (x, y) and the information source (x * ,y * ) to conduct social interaction and update the neural network weight matrix:

[0059]

[0060] Step 2225: The decision neural network weight matrix of the traveler is calculated according to the probability p mu A mutation occurs, and each traveler has a probability p mu Re-randomly initialize the neural network weight matrix;

[0061] Step 2226: Check whether the learning evolution process meets the stopping condition. If it does, stop. Otherwise, return to step 2222. The flow of travel mode m (m∈{r,c}) selected at the departure time τ between stations i and j is expressed as

[0062] In any of the above solutions, it is preferred that the net utilities of the different travel modes include the net utility of public transportation activities and the net utility of private car activities.

[0063] In any of the above solutions, preferably, the net utility of the public transportation activity is expressed as:

[0064]

[0065] In equilibrium, travelers cannot increase their net utility by changing their departure time, i.e. Then the net utility of choosing public transportation between stations i and j can be expressed as:

[0066]

[0067] Among them, α represents the travel time cost coefficient, e represents the early arrival time cost coefficient, l represents the late arrival time cost coefficient, τ e represents the last departure time of arriving at the work place early, Λ is the departure interval, and p r is the fare of route r, and η is the congestion coefficient.

[0068] In any of the above schemes, preferably, the net utility of the private car activity is expressed as

[0069]

[0070] In equilibrium, travelers cannot increase their net utility by changing their departure time, i.e. Because the earliest and latest travelers skip the queue, there is And those who choose to travel by private car can also use To express it, we can establish the equation system:

[0071]

[0072] Solve the equations to find τ 1,i,j and τ′ 1,i,j The expression of And calculate the net utility of choosing a private car between stations i and j:

[0073]

[0074] In any of the above solutions, preferably, step 224 includes setting T e , τ * 、T l They represent the early arrival time, the required on-time arrival time, and the late arrival time. represents the travel time of travel mode m at time τ on day t between stations i and j. The utility of travelers choosing travel mode m at time τ between stations i and j is expressed as:

[0075]

[0076] Among them, λ1,λ2 represent the loss aversion coefficients of different profit and loss areas, let θ represent the utility perception coefficient, and the logit model can be used to represent the number of people who choose to depart at time τ between stations i and j: Update activity utility for different travel modes.

[0077] In any of the above schemes, preferably, step 27 comprises: i′ , X i″ and X i″′ Perform crossover operation, the method is:

[0078] For individual X i′ The line matrix x in i′ :

[0079] (1) Traverse the line matrix x i′ For each line in, calculate the relationship between each line r′ and x i′ The average similarity of other routes in the statistic (average similarity: compare the number of common sites with other routes pairwise and take the average value), and find the route r′ with the greatest similarity to other routes m .

[0080] (2) Randomly select X i″ or X i″′ , with X i″ Take, for example, traverse its line matrix x i″ For each line in, calculate the relationship between r″ and x i′ The average similarity of each route in , find the route r″ with the minimum similarity m or r″′ m .

[0081] (3) The matrix x i′ The line r′ in m Replace with r″ m or r″′ m , to examine the feasibility of the matrix, if the x after replacing the line i′ If it is still an infeasible matrix, then directly use x i″ or x i″′ Replace x i′ ;

[0082] For individual X i′ The price column vector p in i′ and the departure frequency column vector f i′ :

[0083] Let μ p,i′ =(p i″ +p i″′ ) / 2,σ p =|p i″ -p i″′ |,μ f,i′ =(f i″ +f i″′ ) / 2,σ f =|f i″ -f i″′ |, then the intersection result of "individual optimal" and "global optimal" is:

[0084] The present invention proposes a bus route optimization method that takes into account the complexity of travel behavior. Under the displacement level of existing new energy buses, when planning bus routes, the number of passing stations is appropriately increased, thereby increasing the line coverage length, which helps to reduce the system displacement, improve the average cross-sectional flow of the line network, and reduce the maximum cross-sectional flow of the line network, thereby achieving the operational effects of reducing emissions, balancing the capacity between stations, encouraging public transportation and alleviating congestion. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 The figure is a schematic diagram of an embodiment of a cross-route mechanism of a bus route optimization method considering the complexity of travel behavior according to the present invention.

[0086] Figure 2 The present invention is a flowchart of an embodiment of a model optimization solution of a bus route optimization method considering the complexity of travel behavior according to the present invention.

[0087] Figure 3 A schematic diagram of an embodiment of a Mandl example road network of a bus route optimization method considering the complexity of travel behavior according to the present invention.

[0088] Figure 4 It is a schematic diagram of an embodiment of a Pareto optimal solution of a bus route optimization method considering the complexity of travel behavior according to the present invention.

[0089] Figure 5 The diagram is a schematic diagram of an embodiment of the flow state corresponding to a typical route scheme of a cluster travel evolution model of a bus route optimization method considering the complexity of travel behavior according to the present invention.

[0090] Figure 6 The diagram is a schematic diagram of an embodiment of the flow state corresponding to a typical route scheme of the logit model of the bus route optimization method considering the complexity of travel behavior according to the present invention.

[0091] Figure 7 A schematic diagram of an example road network of an embodiment of a method for optimizing public transportation routes taking into account the complexity of travel behavior according to the present invention.

[0092] Figure 8 A distribution diagram of another embodiment of a Pareto optimal solution of a bus route optimization method considering the complexity of travel behavior according to the present invention. DETAILED DESCRIPTION

[0093] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0094] Embodiment 1

[0095] This embodiment discloses a bus route optimization method considering the complexity of travel behavior, including:

[0096] 1. Scene Description and Input

[0097] Consider a bus network system G = (V, A), which consists of N bus stops, V N represents the set of stations, A represents the set of road sections between stations (a∈A). R is the set of bus routes (including transfer routes), each route consists of stations and road sections, D represents the OD demand matrix between bus routes and stations, D i,j is the daily average travel demand between stations i, j, i, j∈V N , between the starting and ending points i and j, the traveler can choose a private car (symbol c) or a bus line r (if it is a transfer line, then the maximum transfer is 1 time, r∈R). In the daily travel activity time period [0,T], the traveler has S departure time options, expressed as [τ1=0,τ2=T / S,τ3=2T / S,…,τ S =(S-1)T / S], for Let q i,j (t,τ) represents the number of people who choose to travel at time τ on the tth day between stations i and j. represents the number of people who choose to depart at time τ on day t between stations i and j and choose travel mode m, then It can be seen that differences in line design schemes will significantly change the flow pattern of the overall bus network system, thereby producing different socioeconomic indicators.

[0098] Model input: bus station information (including passenger flow data between stations, road connections between different stations, and congestion-free travel time), and coefficients related to utility.

[0099] Therefore, the value of the present invention is: under certain constraints on the number of lines and line lengths, considering the complexity of travel behavior of traveler groups, scientifically planning and comprehensively implementing key variables such as the specific structure and direction of bus lines, departure frequency, etc., generating line direction plans, and minimizing the social costs and emissions generated by the overall bus network system.

[0100] 2. Model Construction

[0101] The objectives of the urban traffic management department are as follows: (1) F1 * : Minimize the social costs of the bus network; (2) Minimize the carbon emissions generated by the bus network. The optimization variable is X = (x, p, f), where x represents the route matrix, each row of the matrix represents the stop plan of a route, x determines the basic bus route system R '(excluding transfers), p = (p1, ..., p R′ ) represents the fare column vector (the fare of each basic route), f=(f1,…,f R′ ) represents the departure frequency column vector (the departure frequency of each basic line), represents the utility of choosing travel mode m on day t and departing at time τ. In addition, considering that most cities are currently implementing full coverage of new energy buses, let represents the bus operating cost, Represents the per capita carbon emissions of new energy buses and traditional energy cars. Establish a multi-objective optimization model:

[0102]

[0103] in, represents the passenger flow constraint, x∈Ψ represents the feasibility constraint of the bus route (see Section 4.1 for details), p r,min ≤p r ≤p r,max represents the fare constraint, f r,min ≤f r ≤f r,max represents the capacity constraint, Represents the social welfare constraint of public transportation (total revenue cannot be too large).

[0104] 3. Implementation steps of bus route optimization method

[0105] In view of the multi-objective optimization problem of the bus network structure mentioned above, a multi-objective particle swarm algorithm is used to solve it. Since the optimization variables in the present invention contain a large number of discrete integer variables, the population initialization method and crossover operator of the existing multi-objective particle swarm algorithm need to be redesigned.

[0106] Set the particle swarm size to N p , the optimization variable corresponding to particle i′ is the triple X i′ =(x i′ ,p i′ ,f i′ ), where x i′ , p i′ , f i′ They correspond to the route matrix, fare combination, and departure frequency combination represented by particle i' respectively. The route design method of the prior art is organically combined with the present invention to design a new bus route optimization algorithm.

[0107] Step 1: Set the global non-dominated solution set A g , set a non-dominated solution set A for each particle i′ in the population i′ , build triple X i′ =(x i′ ,p i′ ,f i′ ), initialize the feasible solution and ensure x i′ feasibility.

[0108] The feasible solution is initialized as follows:

[0109] Step 1.1: In x i′ When establishing the first bus route, a station in the road network is randomly selected as the starting point (the first station).

[0110] Step 1.2: For a bus route, traverse the network of stations that have a road connection to the i-1th station and have not been used, and randomly select one of them as the i-th station. If such a station does not exist, replace the currently formed route (s′1, s′2, …, s′ i-1 ) is flipped to become (s′ i-1 ,…,s′2,s′1) (at this time, the i-1th site is s′1), and repeat this step until one or more of the following conditions are met:

[0111] ①The degree of the stations at both ends of the line is 1.

[0112] ②Continue to add sites to form a loop.

[0113] ③The line length reaches the set maximum value.

[0114] At this time, the site set is stored in x as a row vector i′ ,(x i′ Add a row).

[0115] Step 1.3: For the kth line, randomly select a station from the generated lines 1 to k as the starting point and repeat step 1.2. When the number of lines has reached the maximum value, if the line matrix x i′ All the stations in the road network have been included (that is, all stations have at least one bus line passing through), the algorithm stops, x i′ That is a feasible line matrix.

[0116] For infeasible routes, do the following:

[0117] Step 1.4: When x i′ When it is an infeasible solution, the line matrix x i′ If all the stations in the road network are not included, then the set of stations that no bus line passes through is represented as (s1′, s2′, …, s′ m ), the current line set is represented as (r1′,r2′,…,r k ′).

[0118] Step 1.5: Select a set of lines (r′1, r′2, …, r′ k ) is the current line r′ i .

[0119] Step 1.6: Select a set of sites (s′1, s′2, …, s′ m ) is the current missing site s′ j .

[0120] Step 1.7: If line r′ i Contains the site s′ j Connected sites s′ x , go to step 1.8, otherwise, go to step 1.9.

[0121] Step 1.8: Obviously, the line r′ i is divided into two parts, r′ i,1 and r′ i,2 .

[0122] If the site s′ x In r′ i,1 At the end of the line, flip the line, station s′ x Become the head end and connect r′ to the new tail end i,2 The stations in the form a new line r″ i,1 .

[0123] If the site s′ x In r′ i,2 At the beginning of the line, turn the line over, station s′ x Become the tail end and connect r′ at the new head end i,1 The stations in the form a new line r″ i,2 .

[0124] Compare line r′ i,1 and r″ i,2 The length of r′ i =max{r″ i,1 ,r″ i,2 }, replace s′ j From (s′1,s′2,…,s′ m ) and go to step 1.5.

[0125] Step 1.9: If the current line r′ i Not (r′1,r′1,…,r′ k ), let i=i+1 and go to step 1.7; otherwise, go to step 1.10.

[0126] Step 1.10: Traverse the line matrix x i′ If there are still sites that cannot access the current line set, then x i′ If it is an infeasible solution, discard it and start again to generate an initial feasible solution.

[0127] Step 2: Calculate X i′ =(x i′ ,p i′ ,f i′ ) The calculation steps of the final OD matrix corresponding to is as follows:

[0128] Step 2.1: Based on the given x i′ , search for feasible routes and reasonable transfer routes between each OD pair.

[0129] Step 2.2: For a given p i′ With f i′ , so that the travel time between each station is the sum of the zero flow time of each section, the number of iterations k′=0, and the passenger flow OD matrix between stations is calculated The model and calculation method are as follows:

[0130] For each OD pair (between stations i and j) in the bus network, establish a agent The cell space Ω of each traveler cell , each traveler agent individual (x, y) in this space has its own cognitive structure (decision neural network), let Represents the cellular neural network weight matrix (n layers) of individual (x, y), let represents the probability that an individual (x, y) chooses travel mode m between stations i and j. The decision evolution process of the traveler (x, y) can be expressed as:

[0131] Step 2.2.1: Each traveler initializes its own neural network weight matrix

[0132] Step 2.2.2: On day t, for each travel agent between stations i and j, calculate the net utility of each travel mode Input its own neural network, through the weight matrix The probability of choosing different travel modes is obtained through calculation processing. Traveler agent based Choose your travel method.

[0133] Among them, the net utility of different travel modes is expressed as follows:

[0134] (1) Public transport activity effectiveness

[0135] Travelers can choose to travel by bus. The generalized cost (negative utility) of the bus consists of three parts: time cost, cost of arriving early or late, congestion cost and fare. The cost of choosing bus route r between stations i and j at time τ on day t is:

[0136]

[0137] Among them, α represents the travel time cost coefficient, e represents the early arrival time cost coefficient, l represents the late arrival time cost coefficient, τ e represents the last departure time of arriving at the work place early, Λ is the departure interval, and p r is the fare of route r (if r is a transfer route, it is obtained by adding the fares of different routes), η is the congestion coefficient. Considering the impact of departure time selection on route flow, τ S-1 Not all travelers entering the road section at any given moment will be in the time period [τ S-1 ,τ S ] Leave. Define 0-1 variables When the traveler departing at time τ between stations i and j can reach section a at time τ′ otherwise Then the incoming flow of section a at time τ′ can be expressed as:

[0138] make and They represent the flow between stations i and j in the upward direction (superscript "+") and the downward direction (superscript "-") of bus line r at time τ, respectively, then: It can be seen that when the traveler departs at time τ, the total number of people on bus route r between stations i and j is When different departure times are selected, the travel time of the bus line is determined by the travel time of the line network segment at the future time τ′, so there is in, is the zero flow time between sites i and j, f bus With c bus They represent bus departure frequency and capacity respectively.

[0139] In recent years, scholars have found that the traditional choice model based on generalized cost will lead to a large deviation in the estimation of traveler behavior decisions, while the expression of activity utility is more rigorous and closer to reality. and denote the marginal utility of the traveler's home and work activities (m∈{r,c}), τ″ denotes the end time of work, then the utility function of choosing bus route r between stations i and j at time τ can be expressed as: Based on activity utility theory and the commonly used constant marginal utility form The net utility of choosing public transportation can be expressed as:

[0140]

[0141] In equilibrium, travelers cannot increase their net utility by changing their departure time, i.e. Then the net utility of choosing public transportation between stations i and j can be expressed as:

[0142]

[0143] (2) Private car activity utility

[0144] The rapid increase in the number of private cars has had a certain impact on existing public transportation. Many travelers will also choose to travel by private car. The generalized cost (negative utility) of private cars consists of three parts: time cost, cost of early or late arrival, and fare. The cost of choosing a private car between stations i and j at time τ on day t is:

[0145]

[0146] Among them, q′(τ) represents the length of the private car queue, s represents the private car capacity in units of people, and τ 1,i,j and τ′ 1,i,j Respectively represent the earliest and latest departure time of private cars, τ c,i,j is the on-time departure time for private cars, then p c Private car fees (parking fees), travel time When the OD bus cannot transfer, Among them, c car is the road capacity.

[0147] Travelers who choose to take a private car between stations i and j at time τ are in [τ 1,i,j ,τ′ 1,i,j The total utility obtained within ] is In the constant marginal utility form The net utility of the traveler is the difference between the activity utility and the travel cost. The net utility of choosing a private car is:

[0148]

[0149] In equilibrium, travelers cannot increase their net utility by changing their departure time, i.e. Therefore, we can obtain by differentiating (6) Furthermore, since the earliest and latest travelers are exempt from queuing, we have And those who choose to travel by private car can also use To express it, we can establish the equation system:

[0150]

[0151] Solve the equations to find τ 1,i,j and τ′ 1,i,j The expression of Substituting into (6), we can obtain the net utility of choosing a private car between stations i and j:

[0152]

[0153] Step 2.2.3: Each traveler agent obtains its own travel experience

[0154] Step 2.2.4: Each traveler in the cellular space searches for the travel utility (experience) value TE within the neighborhood x,y The largest source of information, denoted as (x * ,y * ), let p cr represents the interaction intensity, the relationship between the traveler (x, y) and the information source (x * ,y * ) to conduct social interaction and update the neural network weight matrix:

[0155]

[0156] Step 2.2.5: The weight matrix of the traveler's decision neural network is calculated according to the probability p mu A mutation occurs, and each traveler has a probability p mu Re-initialize the neural network weight matrix randomly.

[0157] Step 2.2.6: Check whether the learning evolution process meets the stopping condition (reaching a certain number of steps or the weight tends to be stable). If it does, stop, otherwise return to step 2.2.2.

[0158] In summary, the flow of the travel mode m (m∈{r,c}) selected at the departure time τ between stations i and j can be expressed as

[0159] Step 2.3: According to step 2.2, we can get Can calculate the traffic volume of the road section and Then update the travel time to get

[0160] Step 2.4: Calculate the utility of departure time selection as follows:

[0161] Let T e , τ * , T l They represent "early arrival time", "required on-time arrival time" and "late arrival time" respectively. represents the travel time of travel mode m at time τ on the tth day between stations i and j. Combined with people's living habits in reality, the choice of departure time is regarded as a profit-loss judgment process related to the arrival time. The utility of travelers choosing travel mode m at time τ between stations i and j is expressed as:

[0162]

[0163] Among them, λ1,λ2 represent the loss aversion coefficients in different profit and loss areas. Considering that people often pay more attention to losses than gains, λ2>λ1, α1,β1 represent the risk sensitivity in different profit and loss areas. In addition, considering that the early arrival time and late arrival time of travelers often depend on objective factors such as their place of residence, place of work, and nature of work, and are less affected by other external factors, let θ represent the utility perception coefficient, and the logit model can be used to represent the number of people who choose to depart at time τ between stations i and j: Update activity utility for different travel modes.

[0164] Step 2.5: Update the passenger flow OD matrix between stations according to step 2.2, and obtain

[0165] Step 2.6: Set the termination condition If this condition is met, the iteration stops. That is the current feasible solution X i′ The corresponding OD matrix, otherwise the number of iteration steps k′=k′+1, return to step 2.3.

[0166] Step 3: Calculate the objective function value. m (t,τ) can calculate the objective function value corresponding to individual i′ The solution F corresponding to each particle * (X i′ ) are added to the non-dominated solution set A i′ With A g , add the rule: for the non-dominated solution set A (A = A i′ ,A g ) solution A a , if there is ( Indicates a "dominant" relationship, that is, And in F1 * and There is at least one satisfying F * (X i′ )<A a ), then delete A in A a ; if F * (X i′ ) is not dominated by any solution in A, then F * (X i′ )Add A.

[0167] Step 4: Calculate constraint violation values in, <x>It means if x≤0, then <x>=0, otherwise <x>=|x|. For any two solutions F * (X i′ ) and F * (X j′ ) and its corresponding constraint violation value, a multi-objective optimization constraint processing method based on constraint violation value is used to determine F by using the principle of smaller constraint violation value dominance. * (X i′ ) and F * (X j′ )’s dominant relationship.

[0168] Step 5: Calculate solution set A i′ With A g The crowding distance of each group of solutions in is calculated as follows: g For example, let Indicates A g Objective function In descending order, the crowding distance of the i′th group of solutions is: The solution with a larger crowding distance has a greater probability of being selected.

[0169] Step 6: Select. The crowding distance corresponding to individual i′ is Select excellent individuals according to probability, refer to the "roulette wheel" method in genetic algorithm, in A i′ With A g In the equation, the probability of the solution corresponding to individual i′ being selected is In A i′ The individual selected from A is denoted as i″. g The individual selected is denoted as i″′.

[0170] Step 7: Crossover operator based on line similarity. Individual X i′ , X i″ and X i″′ To perform crossover operation, the method is designed as follows:

[0171] For individual X i′ The line matrix x in i′ :

[0172] (1) Traverse the line matrix x i′ For each line in, calculate the relationship between each line r′ and x i′ The average similarity of other routes in the statistic (average similarity: compare the number of common sites with other routes pairwise and take the average value), and find the route r′ with the greatest similarity to other routes m .

[0173] (2) Randomly select X i″ or X i″′ , with X i″ Take the line matrix x as an example. i″ For each line in, calculate the relationship between r″ and x i′ The average similarity of each route in , find the route r″ with the minimum similarity m or r″′ m .

[0174] (3) The matrix x i′ The line r′ in m Replace with r″ m or r″′ m , according to the method in step 1, check the feasibility of the matrix. If the x after replacing the line i′ If it is still an infeasible matrix, then directly use x i″ or x i″′ Replace x i′ .

[0175] It can be seen that Figure 1 As shown, for the line matrix x i′ In terms of x, the above process achieves i′ The representative paths in the excellent individuals are crossed and exchanged, and the line matrix x is achieved without destroying the rationality of the line combination. i′ Maximum update.

[0176] For individual X i′ The price column vector p in i′ and the departure frequency column vector f i′ :

[0177] Let μ p,i′ =(p i″ +p i″′ ) / 2,σ p =|p i″ -p i″′ |,μ f,i′ =(f i″ +f i″′ ) / 2,σ f =|f i″ -f i″′ |, then the intersection result of "individual optimal" and "global optimal" is:

[0178] Step 8: Mutation. Set the mutation probability to p′ mu , individual X i′ The line matrix x i′ According to the probability p′ mu Regenerate. If the newly generated line matrix has sites that cannot be connected to the current line set, then x i′ Remain unchanged. The price column vector p i′ and the departure frequency column vector f i′ According to the probability p′ mu Re-randomize the value within the constraints.

[0179] Step 9: Check whether the termination condition is met. If the condition is met, stop, otherwise return to step 2.

[0180] The main process of the method consists of Figure 2 shown.

[0181] This paper aims to reduce carbon emissions and social costs in the bus network, considers the dynamic evolution mechanism of activity utility and cluster behavior, and proposes a multi-objective optimization model for bus network structure. The conclusions are as follows:

[0182] (1) Under the existing displacement level of new energy buses, when planning bus routes, appropriately increasing the number of stops along the way and thereby increasing the route coverage length will help reduce the system displacement.

[0183] (2) Although the route design scheme oriented towards reducing emissions increases the average travel time of the line network, it can improve the balance of flow between different bus lines while achieving emission reduction. It not only increases the average cross-sectional flow of the line network, but also reduces the maximum cross-sectional flow of the line network, thereby achieving the operational effects of reducing emissions, balancing the capacity between stations, encouraging public transportation and alleviating congestion.

[0184] (3) When the traveler group has consistent and definite marginal utility information, increasing bus departure frequency and reducing fares will help reduce system displacement. However, under this condition, the route plan oriented towards reducing displacement will also increase the maximum passenger flow during the system peak period, which will have a certain impact on "peak shaving and valley filling".

[0185] (4) As members of society, the improvement of marginal utility of travelers in different activity occasions can help reduce both emission volume and social costs, among which the effect of saving social costs is more significant.

[0186] Embodiment 2

[0187] This embodiment discloses an example input and result output using this method.

[0188] 1. Example road network and parameter input

[0189] like Figure 3 Taking the internationally widely used Mandl standard test road network as an example, the optimization model proposed in the present invention is measured and calculated to find the optimal bus route plan, travel demand data between stations, zero flow time ( Figure 3 The numbers on the lines between sites are based on the literature. In addition, unless otherwise specified, the values ​​of the main parameters of the model and their basis are shown in Table 1.

[0190] Table 1 Model parameter values

[0191]

[0192] Among them, the coefficient related to the activity utility α, e, l, s, η are taken based on the literature that specifically studies the utility of travel activities. c Based on the scale of the road network in the example and the current daytime parking fee standard, the coefficients λ1, λ2, α1, and β1 related to risk attitude are taken according to the literature. and The carbon emission data per person per kilometer of new energy buses and traditional energy cars are taken from the media. According to the existing literature on the treatment of bus route scale in the Mandl road network, the shortest bus route length is set to 6 stations, the longest route is set to 9 stations, and the maximum number of routes is set to 4, 6, and 8 respectively. In addition, in order to facilitate the comparison with the bus route optimization results under the condition of no cluster behavior (travelers have consistent and deterministic utility information), the cellular neural network model is replaced by the traditional logit model for simulation again. Under the condition of ensuring the convergence of steps 1.1 to 1.6 in Section 4.2, the cluster social interaction intensity p is set to cr The value of is the same as that of θ in the logit model.

[0193] 2. Analysis of bus route optimization results

[0194] The maximum number of routes in the network is adjusted, and the bus route optimization results under different model conditions are examined.

[0195] like Figure 4 As shown in the figure, the distribution of Pareto optimal solutions under different model conditions. Each scattered point in the figure represents a feasible optimal line design scheme. Figure 4 It can be seen that increasing the number of bus routes can effectively reduce the overall displacement of the road network; in addition, under the conditions of the logit passenger flow allocation model ( Figure 4 (b)), travelers have consistent marginal utility information, and the road network as a whole produces lower displacement, while when considering cluster evolution behavior, the road network as a whole produces higher displacement. Figure 4 By selecting representative optimal solutions (distributed at both ends and the center of the Pareto front) on the Pareto front, and analyzing and comparing the optimization results based on activity utility, we can find the line design rules under different optimization goal orientations:

[0196] Case 1: When considering the cluster evolution behavior in the actual road network. The average length of the route design scheme oriented to reduce displacement is greater than that oriented to reduce social costs. This is because increasing the length of bus routes improves the coverage of buses and helps encourage bus travel. Although the route scheme oriented to reduce displacement increases the average travel time of the network, the flow balance between different bus routes is better. On the premise of improving the average cross-sectional flow of the network, the maximum cross-sectional flow of the network is effectively reduced, and the operational effect of encouraging bus travel and relieving congestion is achieved. Furthermore, when the total number of routes in the road network remains at a certain scale (4 or 6 routes), the route design scheme oriented to reduce displacement corresponds to a lower departure frequency and a higher fare (when there are fewer routes, the congestion is reduced by reducing the departure frequency and increasing the fare to improve the utility of buses). When there are a large number of lines in the road network (8 lines), the line design plan oriented towards reducing displacement corresponds to a higher departure frequency and a lower ticket price (when there are many lines, passenger flow is attracted directly by increasing the departure frequency and lowering the ticket price, forming a comprehensive policy effect to encourage public transportation travel).

[0197] The route design scheme oriented towards reducing social costs corresponds to a shorter route length, and the flow between different routes shows a phenomenon of "hot and cold" (the average cross-sectional flow of the network decreases, and the maximum cross-sectional flow increases). At the same time, the number of people who need to transfer increases, which increases the pressure on the public transportation system, and the effect of "encouraging public transportation travel" is affected. When the total number of routes in the road network remains at a certain scale (4 or 6 routes), the route scheme oriented towards reducing social costs reduces the average travel time of the network through lower fares and higher departure frequencies. When the number of routes in the road network is large (8 routes), the route scheme oriented towards reducing social costs corresponds to lower departure frequencies and higher fares. This shows that when considering the cluster evolution behavior of travelers, the reasonable departure frequency and fare corresponding to the optimized route are also affected by the number of bus routes that can be accommodated in the road network.

[0198] Case 2: When travelers have consistent marginal utility information. The average length of the route design scheme that is optimized with the reduction of displacement is greater than that of the route design scheme that is optimized with the reduction of social costs. At the same time, the route scheme oriented to reduce displacement corresponds to a higher average departure frequency and a lower fare. This shows that when the traveler group can obtain more certain marginal utility information, increasing the frequency of bus departures and reducing fares can achieve the effect of encouraging bus travel and help reduce displacement. When the total number of routes opened in the road network remains at a certain scale (4 routes or 6 routes), the route scheme oriented to reduce displacement makes the flow between different routes more balanced, which helps to relieve congestion while ensuring the operation effect.

[0199] The line design plans oriented towards reducing social costs all correspond to shorter line lengths and lower departure frequencies. When the total number of lines in the road network remains at a certain scale (4 lines or 6 lines), the flow between different lines shows an "uneven hot and cold" phenomenon (the average cross-sectional flow of the network decreases, and the maximum cross-sectional flow increases). When the number of lines in the road network is large (8 lines), the balance of flow between different lines is improved, and the average travel time and the number of people who need to transfer also decrease.

[0200] Furthermore, based on the optimal route combination scheme, Table 2 gives the standard deviation of the number of passes of each bus stop under different goal orientations and model assumptions, examining the balance of bus stops in the route combination scheme.

[0201] Table 2 Standard deviation of the number of bus stops visited on the Mandl network

[0202] Model\Maximum number of lines 4 6 8 Cluster Evolution Model Low displacement guide 0.87 1.14 1.71 Low cost orientation 1.10 1.64 1.29 Logit Model Low displacement guide 1.02 1.14 1.15 Low cost orientation 0.89 1.64 1.86

[0203] As can be seen from Table 2, in general, the low-displacement-oriented bus route optimization scheme produces the lowest standard deviation of the number of bus stop access times (0.87, 1.14, 1.15), which is conducive to the balanced allocation of capacity among various stations. In addition, when considering the complexity of the cluster behavior of the traveler group and the number of bus routes that can be accommodated in the road network is small, the low-displacement-oriented bus route optimization scheme is more balanced in the allocation of capacity resources and can avoid excessive overlap of bus routes. When travelers have consistently determined marginal utility information and there are more bus routes that can be accommodated in the road network, the low-displacement-oriented bus route optimization scheme is more balanced in the allocation of capacity resources and can avoid excessive overlap of bus routes.

[0204] like Figure 5 As shown in the figure, under the conditions of the cluster travel evolution model of the present invention, the time-sharing line flow status corresponding to the line design schemes with different optimization orientations. Among them, the numbers 1 to 4, 1 to 6 or 1 to 8 in the horizontal axis "travel mode" represent the bus lines in the road network, and the numbers 5, 7, and 9 represent private cars. It is not difficult to find that 8:20 is the peak departure time. Furthermore, compared with the optimization orientation of cost reduction, when the bus line optimization is oriented to emission reduction, it not only significantly reduces the overall displacement of the system, but also effectively reduces the maximum passenger flow during the peak period of the system, making the flow distribution between different travel modes more balanced, and achieving the effect of "peak shaving and valley filling" for the congestion phenomenon. Combined with the analysis results of the maximum cross-sectional flow in Tables 2 to 4, it can be seen that the line design scheme oriented to reduce costs has brought about the negative externality of congestion.

[0205] like Figure 5 As shown in Figure 2, under the logit model, the time-sharing line traffic status corresponding to the line design schemes with different optimization orientations. Figure 5 and Figure 6 It can be seen that when travelers have consistent marginal utility information, the maximum passenger flow during the system peak period increases significantly. In addition, Figure 6 It can be seen that when the total number of routes in the road network is 4, the route plan oriented to emission reduction increases the maximum passenger flow during the peak period of the system ( Figure 6 (a)(b)), when the total number of routes is 6 or 8, the emission reduction-oriented route plan reduces the maximum passenger flow during the system peak period. It can be seen that when travelers have consistent marginal utility information, the effect of the route plan of "shaving peaks and filling valleys" is weakened.

[0206] Embodiment 3

[0207] like Figure 7 As shown in the figure, the real road network in the East Third Ring Road and East Fourth Ring Road areas of Chaoyang District, Beijing is selected as an example. The example road network consists of 12 nodes. There is a bus stop near each node (important intersection). The OD demand between stations is obtained by desensitizing the public transportation card swiping data (collection time is from August 1 to September 30, 2020), and the travel time data between stations is obtained by map software.

[0208] like Figure 8 As shown in Figure 3, under the real road network structure and demand conditions, the distribution of Pareto optimal solutions obtained by the method of the present invention (the number of routes is the same as the actual situation) is shown. Table 3 shows the comparison of the representative optimal solutions and the calculation results of the real routes, where the real route fare is the average fare level of Beijing. As can be seen from Table 3, in the bus route design scheme with lower displacement (D3), the average route length is longer. Furthermore, compared with the real route, the feasible solution D2 reduces the social cost within the network and reduces the displacement at the cost of a slight increase in travel time; at the same time, the feasible solution D1 achieves a significant reduction in social cost and travel time at the cost of a slight increase in displacement, indicating that the model of the present invention can achieve the effect of reducing the social cost and displacement of the bus network.

[0209] Table 3 Comparison of bus route plans

[0210]

[0211]

[0212] In order to better understand the present invention, the above is described in detail in conjunction with the specific embodiments of the present invention, but it is not intended to limit the present invention. Any simple modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention. Each embodiment in this specification focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referenced to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.< / x> < / x> < / x> < / x> < / x> < / x>

Claims

1. A bus route optimization method considering the complexity of travel behavior, comprising obtaining bus station information of an existing bus line network system, characterized in that: The following steps are also included: Step 1: Construct a multi-objective optimization model for bus routes based on the traffic network objectives; Step 2: Use a multi-objective particle swarm algorithm to optimize the multi-objective optimization model of the bus route; Step 3: Input the bus stop information into the optimized bus route multi-objective optimization model and output the bus route optimization plan.

2. The bus route optimization method considering the complexity of travel behavior as claimed in claim 1, characterized in that: The bus network system G = (V, A), consisting of N bus stops, V N represents the set of stations, A represents the set of road sections between stations (a∈A), R represents the set of bus routes, each route consists of stations and road sections, D represents the OD demand matrix between bus routes and stations, D i,j is the daily average travel demand between stations i, j, i, j∈V N , between the starting and ending points i and j, the traveler can choose a private car c or a bus line r. In the daily travel activity time period [0, T], the traveler has s departure time options, expressed as [τ1=0,τ2=T / S,τ3=2T / S,…,τ S =(S-1)T / S], for Let q i,j (t,τ) represents the number of people who choose to travel at time τ on the tth day between stations i and j. represents the number of people who choose to depart at time τ on day t between stations i and j and choose travel mode m, then 3. The bus route optimization method considering the complexity of travel behavior as claimed in claim 2, characterized in that: The transportation network objectives include: (1) Minimize the social costs of the bus network; (2) Minimize carbon emissions from the bus network; The optimization variable is X = (x, p, f), where x represents the route matrix, each row of the matrix represents the stop plan of a route, x determines the basic bus route system R', p = (p1, ..., p R′ ) represents the fare column vector, f=(f1,…,f R′ ) represents the departure frequency column vector, represents the utility of choosing travel mode m on day t and departing at time τ, let represents the bus operating cost, Represents the per capita carbon emissions of new energy buses and traditional energy cars.

4. The bus route optimization method considering the complexity of travel behavior as claimed in claim 3 is characterized in that: The bus route multi-objective optimization model is: in, represents the passenger flow constraint, x∈Ψ represents the feasible constraint of the bus route, p r,min ≤p r ≤p r,ma represents the fare constraint, f r,min ≤f r ≤f r,max represents the capacity constraint, Represents the social welfare constraints of public transportation.

5. The bus route optimization method considering the complexity of travel behavior as claimed in claim 4, characterized in that: The step 2 includes setting the particle swarm size to N p , the optimization variable corresponding to particle i' is the triple X i′ =(x i′ ,p i′ ,f i′ ), where x i′ 、p i′ 、f i′ They correspond to the route matrix, fare combination, and departure frequency combination represented by particle i' respectively.

6. The method for optimizing bus routes considering the complexity of travel behavior as claimed in claim 5, characterized in that: The step 2 includes the following sub-steps: Step 21: Set the global non-dominated solution set A g , set a non-dominated solution set A for each particle i' in the population i′ , build triple X i′ =(x i′ ,p i′ ,f i′ ), initialize the feasible solution to ensure the feasibility of xi′; Step 22: Calculate X i′ =(x i′ ,p i′ ,f i′ )The final OD matrix corresponding to; Step 23: Calculate the objective function value, given by q m (t,τ) calculates the objective function value corresponding to individual i′ The solution F corresponding to each particle * (X i′ ) are added to the non-dominated solution set A i′ With A g , add the rule: for the non-dominated solution set A (A = A i′ ,A g ) solution A a , if there is F * (X i′ )<A a , then delete A in A a ; if F * (X i′ ) is not dominated by any solution in A, then F * (X i′ ) is added to A, where < indicates a dominant relationship, i.e., F1 * (X i′ )≤A a (1),F2 * (X i′ )≤A a (2) and in F1 * and There is at least one satisfying F * (X i′ )<A a ; Step 24: Calculate constraint violation values in, <x>It means if x≤0, then <x>=0, otherwise <x> ＝|x|；< / x> < / x> < / x> Step 25: Calculate solution set A i′ With A g The crowding distance of each group of solutions in Indicates A g Objective function In descending order, the crowding distance of the i′th group of solutions is: Step 26: The crowding distance corresponding to individual i′ is Select excellent individuals according to probability, in A i′ With A g In the equation, the probability of the solution corresponding to individual i′ being selected is In A i′ The individual selected from A is denoted as i″. g The individual selected is denoted as i″′; Step 27: Crossover operator based on line similarity; Step 28: Set the mutation probability to p′ mu , individual X i′ The line matrix x i′ According to the probability p′ mu Regenerate. If the newly generated line matrix has sites that cannot be connected to the current line set, then x i′ Keep unchanged, the price column vector p i′ and the departure frequency column vector f i′ According to the probability p′ mu Re-randomize the value within the constraints; Step 29: Check whether the termination condition is met. If the termination condition is met, stop. Otherwise, return to step 22. The termination condition is that the Pareto front formed by the optimal solution cannot be further advanced.

7. The method for optimizing bus routes considering the complexity of travel behavior according to claim 6, characterized in that: The method for initializing the feasible solution includes the following sub-steps: Step 2101: In x i′ When establishing the first bus route, a station in the road network is randomly selected as the starting point; Step 2102: For a bus route, traverse the stations in the road network that are connected to the i-1th station by roads and have not been used, and randomly select one of them as the i-th station; If such a site does not exist, the currently formed line (s′1, s′2, …, s′ i-1 ) is flipped to become (s′ i-1 ,…,s′2,s′1), repeat this step until at least one traversal condition is met, and the site set is stored in x in the form of a row vector. i′ ; Step 2103: For the kth line, randomly select a station from the generated lines 1 to k as the starting point, and repeat step 212; when the number of lines has reached the maximum value, if the line matrix x i′ All stations in the road network have been included, the algorithm stops, x i′ That is a feasible line matrix; Step 2104, for an infeasible route, when x i′ When it is an infeasible solution, the line matrix x i′ If all the stations in the road network are not included, then the set of stations that no bus line passes through is represented as (s′1, s′2, …, s′ m ), the current line set is represented as (r′1,r′2,…,r′ k ); Step 2105: Select a set of lines (r′1, r′2, …, r′ k ) is the current line r′ i ; Step 2106: Select a site set (s′1, s′2, …, s′ m ) is the current missing site s′ j ; Step 2107: If line r i ′ contains the same j Connected sites s′ x , go to step 2108, otherwise, go to step 2109; Step 2108: Line r′ i is divided into two parts r′ i，1 and r′ i,2 ; If the site s′ x In r′ i,1 At the end of the line, flip the line, station s′ x Become the head end and connect r′ to the new tail end i,2 The stations in the form a new line r″ i,1 ; If the site s′ x In r′ i,2 At the beginning of the line, turn the line over, station s′ x Become the tail end and connect r′ at the new head end i,1 The stations in the form a new line r″ i,2 ; Comparison line r″ i,1 and r″ i,2 The length of r i ′=max{r″ i,1 ,r″ i,2 }, replace s′ j From (s′1,s′2,…,s′ m ) is removed, and go to step 2105; Step 2109: If the current line r i ′ is not (r1′,r2′,…,r′ k ), let i=i+1, go to step 2107, otherwise, go to step 2110; Step 2110: traverse the line matrix x i′ If there are still sites that cannot access the current line set, then x i′ If it is an infeasible solution, discard it and start again to generate an initial feasible solution.

8. The method for optimizing bus routes considering the complexity of travel behavior as claimed in claim 7, characterized in that: The step 22 includes the following sub-steps: Step 221: Based on the given x i′ , search for feasible routes and reasonable transfer routes between each OD pair. Step 222: For a given p i′ With f i′ , so that the travel time between each station is the sum of the zero flow time of each section, the number of iterations k′=0, and the passenger flow OD matrix between stations is calculated Step 223: Pass Calculate the traffic volume of the road section and Then update the travel time to get Step 224: Calculate departure time utility; Step 225: Update the passenger flow OD matrix between sites to obtain Step 226: Set termination conditions If this condition is met, the iteration stops. That is the current feasible solution X i′ The corresponding OD matrix, otherwise the iteration number k′=k′+1, re-execute step 223.

9. The method for optimizing bus routes considering the complexity of travel behavior as claimed in claim 8, characterized in that: The step 122 includes establishing a system including N agen The cell space Ω of each traveler cel , each traveler agent individual (x, y) in this space has its own cognitive structure, let Represents the cellular neural network weight matrix (n layers) of individual (x, y), let represents the probability that an individual (x, y) chooses travel mode m between stations i and j. The decision evolution process of the traveler (x, y) can be expressed as: Step 2221: Each traveler initializes its own neural network weight matrix Step 2222: On day t, for each travel agent between stations i and j, the net utility of each travel mode is calculated as Input its own neural network, through the weight matrix The probability of choosing different travel modes is obtained through calculation processing. Traveler agent based Choose a travel method; Step 2223: Each traveler agent obtains its own travel experience Step 2224: Each traveler in the cellular space searches for the travel utility (experience) value TE within the neighborhood x,y The largest source of information, denoted as (x * ,y * ), let p cr represents the interaction intensity, the relationship between the traveler (x, y) and the information source (x * ,y * ) to conduct social interaction and update the neural network weight matrix: Step 2225: The decision neural network weight matrix of the traveler is calculated according to the probability p mu A mutation occurs, and each traveler has a probability p mu Re-randomly initialize the neural network weight matrix; Step 2226: Check whether the learning evolution process meets the stopping condition. If it does, stop. Otherwise, return to step 2222. The flow of travel mode m (m∈{r,c}) selected at the departure time τ between stations i and j is expressed as 10. The method for optimizing bus routes considering the complexity of travel behavior according to claim 9, characterized in that: The net utility of different travel modes includes the net utility of public transportation and the net utility of private car activities. The net utility of public transportation activities is expressed as: In equilibrium, travelers cannot increase their net utility by changing their departure time, i.e. Then the net utility of choosing public transportation between stations i and j can be expressed as: Among them, α represents the travel time cost coefficient, e represents the early arrival time cost coefficient, l represents the late arrival time cost coefficient, τ e represents the last departure time of arriving at the work place early, Λ is the departure interval, and p r is the fare of route r, and η is the congestion coefficient.

Images