CN116029470A

CN116029470A - Public transport route planning method based on crowd sensing

Info

Publication number: CN116029470A
Application number: CN202211549112.3A
Authority: CN
Inventors: 邢建川; 李万坤; 孔渝峰; 张栋; 卢胜; 陈洋; 王煜啸; 赵子航; 杜文雷
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2022-12-05
Filing date: 2022-12-05
Publication date: 2023-04-28

Abstract

The invention discloses a public transportation route planning method based on crowd sensing, and belongs to the technical field of public transportation route planning. The invention adopts a mode of wiring line by line and optimizing networking, is oriented to all bus stops, carries out random pairing between the starting point and the terminal after determining the starting point and the terminal of the bus stop, records the number of direct passengers between each starting point and each terminal, selects a pair of starting point and terminal with the largest number of direct passengers, and sets the shortest bus route along the pair of starting points and terminal. The bus route planning is further optimized by adopting modes of new construction, merging and the like for the bus stops, the next optimal route is continuously searched, and meanwhile, unreasonable routes or bus stops can be selected for deleting by combining with other methods for the established bus route network. The invention not only fully considers the convenient travel of urban residents, but also considers the benefits of public transport companies and the public transport running condition of the whole city when planning the public transport line.

Description

Public transport route planning method based on crowd sensing

Technical Field

The invention belongs to the technical field of bus route planning, and particularly relates to a bus route planning method based on crowd sensing.

Background

With the popularization and coverage of mobile intelligent devices, a crowd sensing network is getting hotter and hotter, and particularly a mobile terminal model based on crowd sensing. In this model, corresponding excitation mechanisms are configured that both compete and cooperate, which complement each other and complement each other, creating a benign excitation pattern. The visual crowd sensing utilizes the flexibility and connectivity of human beings to collect related data on a large scale, and the analysis is carried out to enhance the understanding, so that the method can make excellent contributions to public safety, environmental monitoring and other aspects of modern society.

The problem of bus route planning has existed since many years, and it mainly includes planning and construction. At present, the planning mode of the public transport line gauge mainly comprises the following steps: (1) Based on the bus route balance analysis, dictionary strategies are jointly provided, and travel distribution of people is considered. (2) In the aspect of optimizing the bus route, a double-layer planning model considering benefits of passengers and buses is provided; (3) And the genetic algorithm is applied to the optimal path searching to realize the optimization of the bus route. (4) Drawing arcs at any two adjacent stations of all bus routes for constructing a transfer route, thereby constructing a transfer network, further converting the transfer network into a familiar mathematical model, and carrying out mathematical analysis and model solving; (5) In the optimization of the bus route, the thought of 'arranging one by one and optimizing to form a network' is provided, the name is that the bus route planning of the city is that the bus network is split into one route, the method is a mode of tracing the reason from the result, the optimal alternative route of each station is analyzed, the trend is determined, and the whole bus route network is gradually constructed.

Disclosure of Invention

Aiming at the planning problem of urban bus routes, the invention acquires the passenger travel data in a crowd sensing mode, and performs optimized planning on the existing urban bus routes based on the acquired passenger travel data so as to improve planning performance.

The invention provides a public transportation route planning method based on crowd sensing, which comprises the following steps:

step 1, acquiring passenger travel data through a mobile intelligent terminal carried by a passenger based on a crowd sensing network;

step 2, inputting a bus route network of a target area, which is defined as n= (S, E), wherein S represents a bus stop set, and E represents a road section set of a bus route of the target area; defining l (E) for any road section E E to characterize the distance of the road section E;

for all bus routes of a designated starting point and destination point of a target area, a bus route set G is obtained, and X is defined for any one bus route G epsilon G in the bus route set G _g Representing the stop sequence of bus stops of a bus route g, and defining p _g Representing fare of bus route g, definition F _g The departure frequency of the bus line g is represented;

step 3, planning the stop sequence of the bus stops of each bus route in the bus route set G by adopting a double-layer planning model or a shortest path model, and obtaining and outputting a bus route planning result of the target area; wherein the planning means comprises: adding sites, deleting sites and/or merging neighboring sites;

The dual layer planning model includes: an upper planning model for a bus route and a lower planning model for passengers; in the upper planning model, the optimization objective is as follows based on the objective function of the profit construction of the bus operator: under the condition of meeting the specified influencing factors and constraint conditions, the income of the bus operator is maximized; in the lower planning model, an objective function is constructed based on the sum of travel fees of passengers, and the optimization targets are as follows: minimizing a travel cost sum of the passengers under the condition that the specified influencing factors and the constraint conditions are satisfied, wherein the travel cost sum of the passengers comprises: bus fare P and travel time fees, the travel time fees comprising: waiting for the time cost W, the time cost Y and the transfer time cost Z during taking the bus;

the shortest path model is as follows: taking a bus line network of a target area as a node network diagram of a shortest path model, for each adjacent station in the node network diagram, determining the number of travel people between any two adjacent stations on each bus line based on statistical data in a period of time, and determining the weight factors of the sides between stations based on the number of travel people; based on the initial station and the final station of each bus route, searching the shortest path from the initial station to the final station in the appointed circular area of the target area by adopting a shortest path method until the shortest path currently searched meets the appointed influence factors and constraint conditions.

Further, the influencing factors include: passenger demand, road condition, site selection of bus stops, bus conditions and management policies; the constraint conditions include: bus route length, number of compound lines and transfer times.

Further, the dual-layer planning model is specifically set as follows:

the passenger flow defining the traffic volume when passengers go out is q (i, j), and the calculation mode is as follows:

wherein ,f_k (i, j) represents a passenger flow when a transfer scheme k is adopted in a travel demand with a station i as a departure point of travel and a station j as a destination point of travel, the value of the transfer scheme number k is 1,2,., σ (i, j), σ (i, j) represents the number of transfer schemes with the departure point i reaching the destination point j, and h is defined _k (i, j) represents a specific transfer route of the transfer scheme k;

and the calculation mode of the bus fare P is set as follows: p= Σ _i，j q(i，j)∑ _g∈G p _g ；

Definition alpha represents the unit time value of the passenger, and the waiting time cost W of the passenger is calculated based on the unit time value alpha:

and the time cost Y of the passengers during taking the bus is as follows:

wherein ,v_b Representing the running speed of the bus;

calculating transfer time cost Z based on cost beta caused by excessive time cost caused by transfer:

Wherein a and b respectively represent two buses before and after passenger transfer, p _b The fare of bus b is shown;

setting an objective function of a lower planning model as follows:

wherein ,

representing the sum of the travel fees of the passengers, wherein the crowding degree function y _g (x) The input x of (a) is u _g (i ', j') having the expression: />

wherein ,u_g (i ', j ') represents the passenger flow volume of the bus in the zone where the bus route g (i ', j ') from the station i ' to the station j is located: />

n represents the total number of stations included in the current transfer route, and the coefficient mu, ρ is used for representing the comfort level (the higher the value is, the higher the comfort level is) of the passenger taking the bus, the value is a preset value, and the +.>

When indicating the section e of the bus running on the bus route g +>

The carrying capacity of a bus line g of a bus passing through a road section e is represented;

defining the unit kilometer cost of the bus as R, and calculating the cost R of the bus during running as follows:

R＝2∑ _g∈G ∑ _e∈g r·l(e)

the objective function of the upper planning model is set as follows:

wherein ,

representing the income omega of the operation of a public transport company ₁ ，ω ₂ ，ω ₃ Is a preset three weights, and omega ₁ +ω ₂ +ω ₃ ＝1。

Further, in the shortest path model, the range of values of the radius of the specified circular area of the target area is set as: and (3) half to two thirds of the diagonal length of the full graph of the node network graph, wherein the distance length of the radius is converted based on the running time of the bus.

In the invention, a mode of 'wiring line by line and optimizing networking' is adopted, firstly, all bus stops are oriented, after the starting points and the ending points of the bus stops are determined, any pairing between the starting points and the ending points is carried out, the number of direct passengers between each starting point and each ending point is recorded, a pair of starting points and ending points with the largest number of direct passengers is selected, and the shortest bus route is set along the pair of starting points and ending points. And then, further optimization of bus route planning is realized by adopting modes of new construction, merging and the like for bus stops, then the next optimal route is continuously searched, and meanwhile, unreasonable routes or bus stops can be selected for deleting by combining with other methods for the established bus route network.

The technical scheme provided by the invention has at least the following beneficial effects:

the invention ensures that passengers can travel conveniently, combines the general trend of bus lines and the passenger travel traffic data analysis, ensures that more passengers can directly reach as much as possible, reduces the transfer times, and can reduce the travel time of the passengers, thereby ensuring the convenience of the passengers in traveling. Meanwhile, the method combines the benefits of the public transport company, improves the coverage rate of the public transport line to the whole city as much as possible, reduces the white zone, and enables more city residents to enjoy the public transport conveniently and simultaneously obtains more benefits. And in combination with the integrity of the bus route, when planning the bus route, the connection between one route and the other route needs to be considered, so that the unreasonable and difficult-to-operate conditions are avoided. And considering the sustainability of the bus route, the unreasonable bus route which needs to be improved is re-planned, such as adding a new bus stop or removing an individual bus stop, and the old route which reasonably meets the traveling of passengers is selectively reserved, so that the new and old bus routes are comprehensively considered. The invention not only fully considers the convenient travel of urban residents, but also considers the benefits of public transport companies and the public transport running condition of the whole city when planning the public transport line.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in further detail below.

Aiming at the urban bus route planning problem, the invention collects data based on a crowd sensing mode and optimizes the existing urban bus route to different degrees. In the invention, based on the taking public transportation data (passenger trip data) generated by a user during trip, after comprehensive analysis and processing are carried out by combining with the trip data of the passenger, a shortest path model and a double-layer planning model between two objects of the passenger and the bus are respectively established, based on the solution of the established models (based on Sioux Falls network, two algorithms are adopted for solving), the bus route planning result is obtained, and the comparison analysis is carried out on the aspects of line coverage rate, passenger transfer times, trip cost and the like, so that the advantages and disadvantages of the two models in the bus route planning can be obtained, the relatively reasonable models are judged based on the preset inspection rules, and the bus route planning result of the area to be planned is obtained based on the model comparison result.

The crowd sensing network is a network based on the Internet of things, and benefits from the popularization of mobile intelligent terminals such as intelligent mobile phones, tablets, intelligent watches and the like. The crowd sensing network is a new sensing network based on the mobile intelligent terminals. The sensing units of the sensing network are individual mobile intelligent terminals, and users upload various data acquired (generated) by the mobile intelligent terminals through the mobile network or the WIFI and other Internet ways so as to complete the acquisition work of sensing data.

For bus route planning of different cities, the population number of the cities, the maximum travel duration of citizens and the travel mode adopted at ordinary times need to be referred to. In view of the fact that the public transportation line network has a great influence on the life of citizens, when planning urban public transportation lines, relevant data such as the scale of cities and the like and passenger travel OD (traffic volume) data are combined, the aim of ensuring that passengers can travel conveniently is achieved, the travel time of the passengers can be reduced, and the travel distance and the transfer times of the passengers can be reduced as much as possible; so as to improve the efficiency of public transportation, promote the public transportation to develop stably and orderly, improve the coverage area of the public transportation line and improve the travel experience of urban residents based on public transportation. In the invention, when the bus route planning is realized, a mode of 'wiring line by line and optimizing networking' is adopted, firstly, all bus stops are oriented, after the starting points and the ending points of the bus stops are determined, any pairing between the starting points and the ending points is carried out, the number of direct passengers between each starting point and each ending point is recorded, a pair of starting points and ending points with the largest number of direct passengers is selected, and the shortest bus route is set along the pair of starting points and ending points. And then, further optimization of bus route planning is realized by adopting modes of new construction, merging and the like for bus stops, then the next optimal route is continuously searched, and meanwhile, unreasonable routes or bus stops can be selected for deletion by combining with other modes for the established bus route network. When the existing bus line is adjusted, the bus line is not required to be changed in a large range, and the line which is updated just before can be finely adjusted in a small range; because the process of generating the bus line network is one or more of the processes of optimizing in sequence, the integrity of the bus line network is considered at last, and the bus line network is adjusted once in a global angle; the optimal solution of the finally obtained line may not be in line with the actual situation, so that some reasonable fine adjustments are needed to be performed on the public transportation line network according to the consideration of the relevant factors of the actual life and combining with the living habits of urban residents.

When planning a bus route, some influencing factors and some constraint conditions need to be considered.

Among these, influencing factors include: passenger demand, road conditions, site selection of bus stops, bus conditions, and related management policies. Firstly, the passenger demand can be set up in the place with high passenger demand under the condition of ensuring that the public transportation service level will not change; in places with small passenger demand, public transport lines are generally less opened in consideration of the income problem of bus operation companies. So as to meet the demands of most passengers on buses, not only has wide coverage, but also can reduce travel time, and can improve the direct ratio and the like. Secondly, road conditions are required to be considered when a new bus line is opened or a new bus stop is added, and whether the road conditions can meet the conditions of opening the bus line, such as road width, road surface quality and the like, is required to be considered. The method has the advantages that the method can be used for solving the problem that the number of buses is limited, and the number of buses is limited, so that the number of buses is reduced, and the cost is reduced. The conditions of the buses themselves, as for buses, physical characteristics such as passenger carrying capacity, the number of buses distributed by lines, the volume of the buses themselves and the like affect the coordination with urban roads to different degrees. Finally, there are related management policy factors, which refer to urban traffic management policies, land development policies, etc., and these policies also affect urban bus route planning.

And the constraint conditions include: the length of bus lines, the number of compound lines, the number of transfer times and the like. First, for the length of a bus route, it has a great relationship with the total area of a city and the usual riding distance of residents, but is generally between 5km and 15 km. If one line is too long, the buses are difficult to arrange, and if special conditions occur, a lot of difficulties are increased in the scheduling process, so that passenger flow is scattered, and the transportation efficiency of the buses is reduced; and one line is too short, so that the traveling experience of passengers can be influenced, and the transfer times of a plurality of passengers can be obviously increased. Secondly, the number of the compound lines is restricted, so that the distribution of the bus lines is more uniform, and the berthing performance of the bus stops can be improved. However, the number of the complex lines is not too large, otherwise, the uniformity of the bus line branches of the whole city is disturbed, and the traffic jam is easily caused at the bus stop; the number of the compound lines is too small, the influence is not too great, only the bus is available, and the number of the compound lines of a common bus line is not more than five. Finally, the passenger can increase the travel time once the passenger needs to transfer the restriction condition of the passenger transfer times, so the passenger transfer times are as few as possible, and the general transfer times cannot be more than twice.

The public transportation route planning can have great influence on residents in the city, so that the convenience trip of the residents in the city is fully considered when the public transportation route is planned, and the interests of public transportation companies and the public transportation running condition of the whole city are also considered. Thus, the following aspects need to be considered:

firstly, the passengers are guaranteed to travel conveniently, more passengers can be guaranteed to be directly in the bus line as much as possible by combining the general trend of the bus line and the OD data analysis of the passengers, the transfer times are reduced, and meanwhile, the travel time of the passengers can be reduced, so that the convenience of the passengers in traveling is guaranteed.

And secondly, considering the benefits of the public transport company, and improving the coverage rate of the public transport line to the whole city as much as possible. And then the integrity of the bus route needs to be considered, and when the bus route is planned, the connection between one route and the other route needs to be considered, so that the unreasonable and difficult-to-operate conditions are avoided. And considering the sustainability of the bus route, the unreasonable bus route which needs to be improved is re-planned, such as adding a new bus stop or removing an individual bus stop, and the old route which reasonably meets the traveling of passengers is selectively reserved, so that the new and old bus routes are comprehensively considered.

The invention provides a bus route planning method, which relates to optimization of a plurality of target combinations, including travel time, bus fare and the like, wherein in the invention, a bus route network N= (S, E) is assumed, wherein S represents a set {1,2, 3..n } of bus stops, N represents the number of stations, E represents a set of road sections of the bus route, and if the road sections E E, the distance of the road sections is l (E). Assuming that the bus route set is G, for one of the routes G ε G, X can be used _g Indicating the stop sequence of the bus stop of the line, p can be used _g Representing the fare of this bus route. Departure frequency F for buses on bus lines _g In order to simplify the problem, the departure frequency of buses on all lines can be set to be constant, so that all buses can be processed to reach the destination uniformly. And further, the bus route planning problem about the whole city can be simplified into: planning a new stop sequence X of bus stops of each bus route according to the OD data of the existing urban bus route and the traveling of passengers _g 。

In the embodiment of the invention, the adopted double-layer planning model comprises the following steps: an upper planning model for a bus and a lower planning model for passengers. And establishing a double-layer planning model between the intersection line network and passengers, and mathematically processing the bus line planning problem by using the model. The upper planning model is mainly aimed at a public transportation company, and a corresponding objective function is constructed based on the benefits of the public transportation company, so as to ensure that the benefits of the public transportation company are as maximum as possible. In the lower planning model, the main study object is a passenger, and a corresponding objective function is constructed based on the sum of all the fees of the passenger when the passenger travels, so as to minimize the sum of the travel fees of the passenger in the lower planning model.

The shortest path model is also one of mathematical models, and the specific principle is that knowledge in graph theory is utilized, and the shortest path model is a model for searching the shortest path between the points A and B in a node network diagram by combining a shortest path algorithm (such as Dijkstra algorithm, floyd algorithm and the like). The model is applicable to both the directional weighting map and the undirected weighting map, but when applied to the problem of bus route planning, additional cost and time and the problem of passenger transfer are required to be increased. Because the starting station o and the ending station d of passengers taking the bus are well determined, only the shortest path between the two points on the map needs to be found. The improvement of the model mainly considers that the number of the travel people between each pair of sites on the graph is different, so that a weight factor is required to be added, and the line planning is more targeted; meanwhile, considering the transfer of passengers, the transfer problem can be limited when an algorithm is applied, for example, the shortest path is searched in a circular area with proper size, so that the complicated and low efficiency of full-image searching can be avoided, the overlong bus route can be avoided, and the traveling experience of the passengers can be also considered.

In the embodiment of the invention, a simulated annealing algorithm is adopted for solving a double-layer planning model; and solving the shortest path model by adopting a Dijkstra algorithm.

The simulated annealing algorithm is a solution strategy based on Monte Carlo (Monte-Carlo) iteration, and is an algorithm for randomly searching an optimal solution, namely a random search algorithm. The simulated annealing algorithm derives from the physical process of annealing objects in thermodynamics. During this annealing, as the temperature decreases, the energy state of the object itself becomes low, and when the temperature becomes sufficiently low, the object starts to condense and crystallize, and the energy state is the lowest at the time of crystallization. So that the lowest energy state can be found if a slow cooling, i.e. annealing, is performed. The simulated annealing algorithm starts from a relatively high temperature, and repeatedly utilizes the probability kick characteristic of the Metropolis sampling strategy along with the continuous decline process of the temperature, and randomly searches the optimal solution of the objective function in the solution space of the multi-objective function optimization solving problem, so that the probability of finding the global optimal solution can reach hundred percent.

The simulated annealing algorithm may be divided into three parts, an objective function, an initial solution, and a neighborhood solution space, respectively.

The first step of the simulated annealing algorithm is initialization, the initial temperature regulation parameter T is large enough, and the initial solution A ₀ Is the starting point of iteration of the whole algorithm, and the initial solution is firstly regarded as the optimal solution at the beginning, namely the optimal solution A=A ₀ K iterations are completed under each temperature regulation parameter T.

The second step is an iterative process comprising the steps of: initial optimal solution A ₀ Domain N (A) ₀ ) Randomly picking out a new solution A ₁ Then the increment Δf=f (a ₁ )-f(A ₀ ). If Δf.ltoreq.0, A ₁ Is the new optimal solution, i.e. a=a ₁ The method comprises the steps of carrying out a first treatment on the surface of the Otherwise, exp (- Δf/t) is to be seen _k ) Whether > random (0, 1) is true, if so, then the optimal solution a=a ₁ 。

The third part is to repeat this step. And then readjusting the temperature parameter T, and judging whether the condition of stopping the loop can be met or not again, if so, stopping, otherwise, stopping the iteration process before the loop again until the optimal solution is finally obtained.

The Dijkstra algorithm can calculate the shortest path between any one node to other nodes on the graph. The algorithm takes the initial node as the center, searches outwards layer by layer, and finally finds the target node, thereby determining the shortest path. The idea of Dijkstra algorithm is to construct a shortest path spanning tree, which uses a starting node as a root node of the spanning tree, and paths from the root node to other nodes are shortest paths. In the embodiment of the invention, a Dijkstra algorithm is executed in the network planning of the public transport line by adopting a node marking mode.

In the invention, the planned line coverage rate, the transfer times of passengers, the travel fees of the passengers and the benefits of buses are compared. The line coverage rate refers to the percentage of planned public transportation lines to the total roads of the city. The passenger travel fee comprises two parts, one is the fare P of the bus and the other is the fee in time. The time fees include the waiting time fee W, the on-board time fee Y and the time fee Z caused by transferring the vehicle. The travel cost of the passengers is the sum of the fare and the time cost of the bus. The benefit of the public transport company is the ticket money paid by the passengers taking the public transport, but the operation cost of the public transport company is also reduced, such as maintenance cost of the public transport, fuel consumption cost, wages of staff such as public transport drivers, ticket sellers and the like. The higher the profit model of the public transport company is also the better.

There are two decision makers in the two-layer planning model: passengers and bus routes (buses). The passengers can select the travel stations by taking the decision variables of the passengers as decision makers, so that travel data are generated, the data influence the planning of the decision maker of the bus route, and meanwhile, the decision maker of the bus route can influence the travel station selection of the decision maker of the passengers after planning the route. In the lower-layer planning model, the first consideration is that the fare generated when the passenger takes the bus is an important objective function in the lower-layer planning model. Passengers can have time expenditure when waiting at buses, and have time expenditure when taking buses, which are problems to be considered in lower layer planning. In addition, passengers may not be directly in transit when taking buses, so that the problem of transfer is unavoidable, and the passengers can generate time expense for transfer when transferring buses. Finally, the problem of distributing passenger flows based on a passenger transfer scheme is resolved.

Assuming that the OD passenger flow at passenger egress is q (i, j), it can be decomposed if viewed from the perspective of the transfer scheme, namely:

wherein ,f_k (i, j) represents the passenger flow when the transfer scheme k is adopted in the process of reaching the destination point j from the starting point i, the value of the transfer scheme number k is 1,2,.. _k (i, j) represents a specific passenger transfer scheme, i.e. passenger flow f _k (i, j) from bus stop i ₁ Taking bus a to bus stop i ₂ Transfer buses b to i ₃ Eventually reaching destination j. The transfer times are generally not more than two times, otherwise, the travel time of the passengers is greatly increased, and the travel experience of the passengers is reduced.

Assuming that the passenger's fare is P, the composition of the fare should be the total number of passengers traveling times the fare of the buses on their corresponding bus routes, i.e.:

where n represents the number of stations passing by.

In addition, when passengers travel, the passengers take some time on the buses and take the buses, and transfer buses also generate some time cost. Prior to calculating the objective function of these time costs, the unit time value α of the passengers is assumed, by which the above narrow time costs of the passengers are converted into a generalized time cost. As for the value of the value alpha in unit time, the value is a fixed value, and the value can be obtained by dividing the annual average GDP of the city by the days of working days in one year and dividing by the average daily working time of urban residents.

For the cost of passengers when waiting for buses, it can be assumed that all buses arrive at the station uniformly, and then the waiting time cost W of the passengers is:

wherein ,F_g Indicating the departure frequency of the bus.

For the time cost of taking a bus between a departure station (i) and a destination station (j), it is assumed that the running speed of the bus is v _b The distance between two bus stops is l (e) and e g, then the time cost Y of the passenger during the bus taking is:

for the time cost of passengers transferring buses, the time cost is calculated on the basis of the cost of passengers and the like, and the cost beta caused by excessive time cost caused by transfer, such as the time cost caused by transfer of getting off and getting on, and the time cost caused by queuing when the passengers get on and off due to large passenger flow, then the transfer cost Z of the passengers is as follows:

wherein a and b respectively represent two buses before and after passenger transfer, p _b The fare of bus b is shown.

In addition, because buses have limited passenger carrying capacity, a function of bus congestion is also considered:

wherein ,u_g (i ', j') is the passenger flow of the bus between i, j, μ, ρThe two coefficients are related to the comfort level of the passengers taking the bus, the preset value,

when the bus runs on the section e of the bus route g, the bus is in the +.>

Is the carrying capacity of the bus route g of the bus passing through the road section e, u _g (i ', j') represents the passenger flow of the bus in the g (i ', j') section:

wherein the value range of i ', j' is the same as i, j.

The cost of these passengers, around the passenger transfer problem, can be used to build the lower planning model:

wherein, the minimum riding expense of the passengers traveling is calculated based on the lower planning model

Normally calculated riding costs for the travel of passengers +.>

In the upper layer planning model, the maximum benefit of public transport company operation is mainly considered, and the travel time of passengers is minimized, so that the method is also the aim of public transport route planning. The bus has certain cost when running, such as fuel cost, maintenance cost, wages of bus drivers and the like, and the cost can be totally integrated into the unit kilometer cost of the bus, and if the fixed value is R, the cost R of the bus when running can be expressed as:

R＝2∑ _g∈G ∑ _e∈g r·l(e)

The income of the public transport company has a direct relation with the number of passengers, and the more the passenger flow is, the more the income is obtained by the public transport company. Thus, fare revenues can be targeted for optimization in upper layer planning. Different weights are introduced into three objective functions of upper layer planning and marked as omega ₁ ，ω ₂ ，ω ₃ And has omega ₁ +ω ₂ +ω ₃ =1. The upper layer planning model can thus be constructed as:

wherein ,

representing the income of the operation of the public transportation company, and calculating the maximum income of the public transportation company based on the upper planning model>

Wherein, the minimum value of the length l of the bus line is 5km, and the maximum value is 15km; the number of the complex lines of the bus lines is not more than 5; the number of passengers' transfer does not exceed 2.

When the simulated annealing algorithm is applied to solve the double-layer planning model, the initial temperature data is ensured to be large enough firstly, and then the initial temperature data can be obtained based on statistical data analysis; a temperature-decreasing proportionality coefficient eta can be set for the temperature regulation and control parameters for regulating and controlling the temperature; when iterating at the same temperature, if the iteration is not changed after N times, the algorithm can terminate, jump out of the loop, and get the optimal solution.

In the embodiment of the invention, the approximate thought of adopting the simulated annealing algorithm when solving the double-layer planning model of the bus route planning is as follows: based on known related data, e.g. average speed v of bus _b Departure frequency F of bus line g _g Fare is p _g And the cost r of the vehicle in unit kilometer, the value alpha in unit time and the like to obtain public busesIntersection line set, and sequence X of bus stop stations _g 。

At the beginning of the algorithm, an initial temperature T can be set ₀ Temperature at termination t=100 _f =0, the scaling factor η of the temperature drop=0.8, and the initial number of iterations set at a certain temperature is k=1;

the first step is to generate an initial bus route planning network. Based on new construction, deletion and the like of bus stops, a shortest path is added to determine some basic initial lines, so that an initial line network can be constructed.

And the second step is to judge whether each bus route is reasonable. In particular, some decisions are made by means of the values of the objective functions of the upper layer model in the two-layer planning. I.e. the calculated minimum overhead is greater than a specified value;

thirdly, researching lower layer planning, and calculating a lower layer objective function value of the initially established public transportation line network

The fourth step is to make some adjustments to the line, such as some new, delete and replace operations on the bus stop, to get a new solution in the neighborhood.

Fifth step is based on fourth step, reusing lower layer plan in new neighborhood solution to calculate objective function value

And step six, comparing the two lower objective function values to obtain a new objective function solution.

And seventh, checking the iteration number, if the iteration number exceeds a certain value and the new solution is acceptable even if the new solution is worse, turning to a fourth step, finding the new solution in the neighborhood, or if the iteration number exceeds a certain upper limit, immediately adjusting the temperature parameter and turning to the fourth step.

The eighth step is to check if the result converges. If the objective function value does not change after a plurality of iterations, the result is an optimal solution, i.e. the bus route planning scheme corresponding to the objective function value is the optimal scheme. If the result is not converged, resetting the iteration number K and turning to a fourth step to find a new neighborhood solution.

The shortest path model is based on a graph consisting of vertex set V and edge set E, i.e., g= (V, E), if graph G is a directed graph, then the ordered pairs < V _i ，V _j The directed edges are represented by the unordered pairs < V if the directed edges are undirected graphs _i ，V _j And > represents an undirected edge. If the edges on the graph have weights, the graph is a weighted graph, w represents a weight function, is an edge-to-real number mapping, and uses the distance between two bus stops or the arrival time of a bus as the weight in the bus route planning problem.

For definition of the shortest path, a simple path L on the graph is defined first, simple path l= < v ₀ ，v ₁ ，...，v _k What is shown is from the starting point v ₀ To the target endpoint v _k On which no point other than the start point and the target end point can be repeated. And the weights of path L are noted:

on the basis of definition of simple paths, if a path with smaller weight between a starting point and a target end point cannot be found on the graph, namely L.ltoreq.L ', wherein L' represents any simple path between the starting point and the target end point, L is v ₀ and v_k The shortest path between them.

After the shortest path model is established, the model is further improved by combining the problem of bus route planning. In actual bus route planning, the Dijkstra algorithm cannot be solved on the whole graph, on one hand, because the algorithm complexity is low in efficiency when the algorithm is solved on the whole graph, and on the other hand, the planned route is not ideal enough, for example, the distance between the starting point and the target destination is too long, and the planned route does not meet the specification of the bus route planning.

Therefore, some improvements are needed to be made on the model, and after the starting point is determined, the Dijkstra algorithm is applied to search the shortest path in a circular area of the starting point, so that two nodes which are far apart on the graph correspond to bus stops which are far apart in reality, and the bus stops can be reached by means of passenger transfer.

Regarding Dijkstra algorithm, there are two implementation manners for solving the shortest path model, and the node or the edge is marked, and in this embodiment, the node is marked. Before the algorithm starts traversing the graph, nodes on the graph are divided into three types, namely unlabeled, temporary marked and marked nodes. Firstly, initializing the nodes on the graph, namely, all nodes are regarded as unmarked nodes, then taking the adjacent nodes of the nodes in the shortest path as temporary marked nodes, and then inquiring the nearest node from the temporary marked nodes as marked nodes every time the algorithm loops until all the nodes are marked, and ending the algorithm. All nodes on the graph can be represented by a pair of labels, i.e. (d) _j ，p _j )，d _j Represented by the shortest path length, p, from the start point s to the destination point j _j The parent node number of j in the shortest path is shown, and the main function is to facilitate the backtracking of the path by the algorithm.

The first step is initialization, S represents an initial marked node set, T represents an unmarked node set, Q represents a temporary marked node set, V represents a node complete set in the graph, u represents a current node, a starting point S is placed in the initial marked set S, namely S= { S }, u=s, other points on the graph are unmarked points at the beginning, and the unmarked node set T=V-S. After initialization, d _s ＝0，

The second step is to find all the neighbors of node u by breadth-first method, put it into set Q, and calculate the distance between u and its neighbor k, denoted as w (u, k), if d _k ＞d _u +w (u, k), then d _k ＝d _u +w (u, k), otherwise, d _k Unchanged and select one from the set Q to let d _k K, which has the smallest value, is stored in S, then at this point s=su { k }, k is also deleted from the set T, and u=k.

The third step is to find the father node of the node in the shortest path, find the k' node directly connected with the k node from the set S, take it as the father node of the k node on the shortest path from the starting point S to the node k, and record as p _k ＝k′。

The fourth step is the termination loop condition of the algorithm, namely, when the set T becomes an empty set, the algorithm is ended, otherwise, the algorithm jumps to the next node of the second step to continue to find the shortest path.

The fifth step backtracking the shortest path based on p _k The array is used for determining father nodes of all nodes on the shortest path, and the shortest path can be obtained through iteration.

When the algorithm is solved, namely when the shortest path is searched for the full graph, the shortest path search for each starting node needs to be limited in a circular area, the circle center is the starting node, the shortest path is searched for in the circular area constructed around the starting node, the radius can be determined according to the size of the full graph, and the length of the radius is preferably half to two thirds of the diagonal length of the full graph, and can be converted into time by using the average speed of the bus. The weight may be set as the travel time of the bus or the distance between bus stops, and the travel time is preferable.

The Sioux Falls network has a strong simulation reality function on objects related to an urban road traffic network, and in the Sioux Falls network, 24 nodes are shared, 38 sides (76 directed sides) are shared, 528 OD pairs are shared (the OD pairs are formed by any two points in the network, and the simulation is that passengers start and end stations when taking buses.

When the Sioux Falls network is applied to the problem of bus route planning, the embodiment of the invention can replace two directed edges between every two adjacent nodes by one straight line, and the nodes can be regarded as bus stops in a real urban road. And replacing all the directed edges in the network with undirected edges to obtain the adjusted Sioux Falls network structure. In the Sioux Falls network, the bus travel time data table 1 between each adjacent node is shown.

TABLE 1 bus travel time between adjacent nodes

In the original network, the OD data of passengers is less, which can affect the operations such as new construction and deletion of a plurality of sites, so that the application effect of the established model and the adopted algorithm can be affected, and the original data size can be amplified, so that the data analysis can be better performed.

And (3) after the simulated annealing algorithm is applied to the double-layer planning model, the double-layer planning model and the simulated annealing algorithm are realized by means of MATLAB based on the Sioux Falls network, so that a scheme of bus route planning, namely a set of all bus routes and a sequence of stations on each bus route, is obtained. In this embodiment, the station sequences on each line in the obtained bus line set are specifically:

a first line: 1. 3, 4, 5, 6, 8, 9;

the second line: 3. 12, 25, 14, 15, 19, 20;

third line: 13. 24, 21, 22, 23, 14, 11, 10;

fourth line: 13. 25, 12, 11, 10, 16, 8, 9, 5;

a fifth line: 5. 9, 10, 16, 17, 19;

sixth line: 2. 6, 8, 7, 18, 16, 17, 19, 20;

seventh line: 4. 11, 14, 23, 22, 20;

eighth line: 6. 8, 7, 18, 20, 21, 22.

Wherein the 25 th site is located intermediate the 12 th and 13 th sites. In this model, the total number of passengers is 47823, the number of direct passengers is 39945, and the direct rate is 83.53%, and the data of the number of passengers, the number of direct passengers and the direct rate are calculated through simulation. Specific principles and processes: and randomly generating travel data of the passengers on the Sioux Falls network through MATLAB, namely after the random number, analyzing the planned route on the Sioux Falls network, and judging that the passenger is a direct passenger if the starting station and the terminal station of the passenger are on the same route, otherwise, judging that the passenger is a non-direct passenger. The direct rate is the number of direct passengers divided by the number of all passengers in the model multiplied by a percentage.

Almost all passengers who cannot reach the bus can get to the target station after transferring the bus once. According to the scheme, all stations in the network are covered, after a 25-number station is added, the number of undirected edges of the network is increased to 40, and 35 stations are covered by the bus route planning scheme, so that the coverage rate of the bus route network is 87.5%, the number of complex lines passing through each station is less than 5, and the lengths of all routes also meet the conditions.

For the shortest path model, the Dijkstra algorithm is applied in a homonymous way, based on Sioux Falls network, and corresponding bus route sets can be obtained through MATLAB software:

the first line is: 1. 3, 4, 5, 9, 10;

the second line is: 3. 12, 11, 10, 16, 18;

the third line is: 2. 6, 5, 4, 11, 12;

the fourth line is: 2. 6, 8, 9, 10, 15;

the fifth line is: 13. 12, 11, 10, 9;

the sixth line is: 5. 9, 10, 15, 22;

the seventh line is: 13. 24, 23, 22, 15, 19, 17;

the eighth line is: 8. 7, 18, 20, 21, 24;

the ninth line is: 11. 14, 15, 19, 20;

the tenth line is: 5. 6, 8, 16, 17, 19;

In the scheme of the model, since the shortest path is not searched for the whole map after the model is improved, the direct passengers only occupy 69.7%, and many passengers need to make one or two transfer, but most passengers can reach the model through one transfer, and the time expense of the passengers is lower than that of the double-layer planning model based on the shortest path, but the cost of the passengers on the bus is increased due to the fact that some passengers transfer buses. For the calculation of the line coverage rate of the model, the network has 38 sides, the planned line covers 33 lines, the coverage rate reaches 86.84%, the number of the complex lines does not exceed five, and the transfer times of passengers do not exceed twice.

For example analysis based on Sioux Falls network, nodes can be added to the network, for example, 48 nodes, 96 nodes and the like are expanded, and experimental analysis can be performed on the two models again.

And (3) introducing an index of a comprehensive evaluation index U of the model for comprehensive evaluation of the model to carry out numerical comparison and analysis on the two models. The index results of the four evaluation factors (line coverage, passenger transfer times, passenger travel expense and public transport company income) are respectively marked as A, B, C and D, and weight factors lambda are respectively added ₁ ，λ ₂ ，λ ₃ ，λ ₄ And lambda is ₁ +λ ₂ +λ ₃ +λ ₄ =1. And respectively calculating comprehensive evaluation indexes U of the experimental model and the comparison model:

U＝λ ₁ ·A+λ ₂ ·B+λ ₃ ·C+λ ₄ ·D

wherein the values of A, B, C and D are the scoring results of the quality of the two models respectively. The index converts the performance of two models on four evaluation factors into a unified value, and the models are convenient to compare.

By consulting the related data of the bus route planning aspect and combining the analysis of the model in the invention, the route can be obtainedThe weight of the evaluation factors of the coverage rate and the number of times of passenger transfer is lower than the travel cost of passengers and the benefit of buses, and the importance of the two evaluation factors of the travel cost of passengers and the benefit of buses is almost the same. Thus, in the embodiment of the present invention, the weighting factors of the four evaluation factors may be respectively represented as λ ₁ ＝0.2，λ ₂ ＝0.2，λ ₃ ＝0.3，λ ₄ ＝0.3。

The overall evaluation index of the two models was calculated by scoring, assuming each module was fully divided by 10. Of the four model evaluation factors, the scoring results of the experimental model (double-layer planning model) and the comparative model (segment path model), i.e., the values of A, B, C, D of the two models, are respectively: a is that ₁＝8 and A₂ ＝7、B ₁＝7 and B₂ ＝6、C ₁＝9 and C₂ ＝8、D ₁＝8 and D₂ =7. The principle of this scoring is based on the performance of the two models on four evaluation factors, respectively.

The scores and the corresponding weight factors of the two models are integrated, the integrated evaluation index of the two models can be calculated, and the integrated evaluation index U of the experimental model ₁ Comprehensive evaluation index U of =8.1 and comparative model ₂ =7.1, so the model integrated evaluation index of the experimental model is greater than that of the comparative model, U ₁ ＞U ₂ Namely, the double-layer planning model has a better effect on bus route planning. That is, a two-layer planning model is preferred in planning.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

What has been described above is merely some embodiments of the present invention. It will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit of the invention.

Claims

1. The public transport route planning method based on crowd sensing is characterized by comprising the following steps of:

2. The method of claim 1, wherein the influencing factors comprise: passenger demand, road condition, site selection of bus stops, bus conditions and management policies; the constraint conditions include: bus route length, number of compound lines and transfer times.

3. The method according to claim 1 or 2, wherein the dual layer planning model is specifically configured to:

wherein ,f_k (i, j) represents the passenger flow when the transfer scheme k is adopted in the travel demand taking the station i as the departure point of travel and the station j as the destination point of travel, the value of the transfer scheme number k is 1,2, …, sigma (i, j) represents the number of transfer schemes from the departure point i to the destination point j, and h is defined _k (i, j) represents a specific transfer route of the transfer scheme k;

and the time cost Y of the passengers during taking the bus is as follows:

wherein ,v_b Representing the running speed of the bus;

setting an objective function of a lower planning model as follows:

wherein ,

wherein ,u_g (i ', j ') represents the passenger flow volume of the bus in the zone where the bus route g (i ', j ') from the station i ' to the station j is located: / >

When indicating the section e of the bus running on the bus route g +>

R＝2∑ _g∈G ∑ _e∈g r·l(e)

the objective function of the upper planning model is set as follows:

wherein ,

representing the income omega of the operation of a public transport company ₁ ,ω ₂ ,ω ₃ Is a preset three weights, and omega ₁ +ω ₂ +ω ₃ ＝1。

4. The method according to claim 1 or 2, wherein in the shortest path model, a range of values of the radius of the specified circular area of the target area is set as: and (3) half to two thirds of the diagonal length of the full graph of the node network graph, wherein the distance length of the radius is converted based on the running time of the bus.