CN117575116A

CN117575116A - Method and system for customizing public transport service applicable to large-scale urban road network

Info

Publication number: CN117575116A
Application number: CN202311303528.1A
Authority: CN
Inventors: 王旭; 周童; 戴荣健; 孙浩文; 杨维浩; 李彦震
Original assignee: Shandong University
Current assignee: Shandong University
Priority date: 2023-10-09
Filing date: 2023-10-09
Publication date: 2024-02-20

Abstract

The invention relates to the technical field of customized public transportation service, and provides a method and a system for customizing public transportation service applicable to a large-scale urban road network, wherein the method comprises the following steps: based on departure places and destinations of all passengers, obtaining theoretical stations through clustering, and converting the theoretical stations to nearest bus stations to obtain customized bus stations; based on the customized bus stop, departure time and arrival time, aiming at minimizing operation cost and waiting time of passengers, adding time constraint, and solving the customized bus stop and service starting time of each bus route; the time constraint includes: the time each passenger takes the bus is a time window centered on the departure time. The passenger satisfaction is improved, the bus passenger flow is increased, and the operation cost is reduced.

Description

Method and system for customizing public transport service applicable to large-scale urban road network

Technical Field

The invention belongs to the technical field of customized public transportation service, and particularly relates to a method and a system for customizing public transportation service suitable for a large-scale urban road network.

Background

The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.

With the ever-increasing traffic demand, existing road infrastructure is overwhelming and cities are facing unprecedented congestion challenges. In order to maximize the utilization of road resources, it is particularly important to guide the private car traveler to take the bus during the Gao Fengtong duty period. However, on the one hand, the conventional buses are inconvenient for urban and rural commuters, and although a good road network in the urban center can provide high accessibility, suburban public transportation systems are imperfect and passengers have long waiting time. On the other hand, conventional bus services are not capable of providing direct traffic to long distance or cross-regional commuters, who are often forced to transfer multiple times, experiencing lengthy ride times in metropolitan or metropolitan areas. This is in contrast to the timely and comfortable riding experience that people expect, resulting in a reduction in the total mass transit passenger traffic. Therefore, searching for new public transportation means, improving convenience of commuters and sustainability of cities has become a major focus of researchers. The customized bus service is used as a supplement to the traditional bus, so that the selection of routes and station configuration is enlarged, and the commute requirements can be well met.

The customized public transportation service is a differentiated, intensive and high-quality urban public transportation service mode for providing a reserved line or train number for passengers by integrating travel demands with similar travel origin-destination, travel time and the like. The specific characteristics are still public transportation, and a fixed station and a completely unfixed line are adopted to provide reservation service for travelers in a sheet area, and a novel public travel service of planning lines is generated by system aggregation matching. The novel service characteristic between private cars and traditional buses can meet diversified travel requirements of passengers, particularly high-quality travel requirements, and part of private car travel can be converted into bus travel.

However, the existing customized bus trip service system (Customized bus service systems, CBTS) is not allowed for arrival delay, and the passenger waiting time and bus service time are long, which not only results in higher operation cost, but also lower passenger satisfaction. In addition, the study of customizing the bus service scope is relatively few, the service level is not quantified aiming at various site scale cases, and the rationality is lacking.

Disclosure of Invention

In order to solve the technical problems in the prior art, the invention provides a method and a system for customizing public transportation service suitable for a large-scale urban road network, which are used for determining the position of a bus stop based on the spatial distribution characteristics of historical commute demands, and introducing the waiting time of passengers to measure late arrival punishment of buses later than the expected receiving time of the passengers, and prescribing that the vehicles must arrive in the receiving stop time window of each passenger, so that the passenger satisfaction degree is improved, the bus passenger flow is increased, and the operation cost is reduced.

In order to achieve the above purpose, the present invention adopts the following technical scheme:

a first aspect of the present invention provides a method of customizing bus services for a large-scale urban road network, comprising:

acquiring departure place, destination, departure time and arrival time of each passenger;

based on departure places and destinations of all passengers, obtaining theoretical stations through clustering, and converting the theoretical stations to nearest bus stations to obtain customized bus stations;

based on the customized bus stop, departure time and arrival time, aiming at minimizing operation cost and waiting time of passengers, adding time constraint, and solving the customized bus stop and service starting time of each bus route; the time constraint includes: the time each passenger takes the bus is a time window centered on the departure time.

Further, adding vehicle allocation constraint in the solving process;

the vehicle allocation constraint includes: each demand is serviced by only one bus and passengers cannot transfer to other buses.

Further, the time constraint further includes: and the sum of the service start time of the bus at the start station, the service time of the bus at the start station and the running time of the bus between the start station and the end station is less than or equal to the service start time of the bus at the end station.

Further, the time constraint further includes: assuming that the bus sequentially serves a first customized bus stop and a second customized bus stop, the sum of the service start time of the bus at the first customized bus stop, the service time of the bus at the first customized bus stop and the running time of the bus between the first customized bus stop and the second customized bus stop is less than or equal to the service start time of the bus at the second customized bus stop.

Further, the solving process adopts a self-adaptive large-scale neighborhood searching algorithm, and the initial solution is obtained by adopting an interpolation method.

Further, the self-adaptive large-scale neighborhood search algorithm adopts three destruction strategies of random elimination, worst elimination and related elimination in the destruction stage.

Further, the adaptive large-scale neighborhood search algorithm adopts three insertion operators, namely random insertion, greedy insertion and regrettably insertion, in the repair stage.

A second aspect of the present invention provides a system for customizing bus services for a large-scale urban road network, comprising:

a data acquisition module configured to: acquiring departure place, destination, departure time and arrival time of each passenger;

A bus stop customization module configured to: based on departure places and destinations of all passengers, obtaining theoretical stations through clustering, and converting the theoretical stations to nearest bus stations to obtain customized bus stations;

a bus service customization module configured to: based on the customized bus stop, departure time and arrival time, aiming at minimizing operation cost and waiting time of passengers, adding time constraint, and solving the customized bus stop and service starting time of each bus route; the time constraint includes: the time each passenger takes the bus is a time window centered on the departure time.

A third aspect of the invention provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of a method of customizing bus services for a mass transit network as described above.

A fourth aspect of the invention provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in a method of customizing bus services for a mass transit network as described above when the program is executed.

Compared with the prior art, the invention has the beneficial effects that:

the present invention optimizes CBTS systems from a site, passenger and operator perspective; in the aspect of the bus stop, the distribution characteristics of the commute demands of passengers in the rush hour are researched, the problem of the acceptance of the passengers in the early stage of system construction is considered, and the bus stop is optimized and customized on the basis of the conventional bus stop; for passengers, the waiting time of the passengers is introduced to measure late arrival punishment of buses after the expected receiving time of the passengers, the service time needs to meet the receiving time window of each passenger, the passenger satisfaction is improved, meanwhile, the passenger flow of the buses is increased, and the operation cost is reduced; in the aspect of public transport operators, under the constraints of vehicle capacity, vehicle speed and the like, the fixed operation cost and the fuel consumption of the vehicles are considered, so that the operation cost is minimized.

According to the invention, an improved large-scale neighborhood searching algorithm is adopted to solve a customized bus route optimization model, so that the model solving efficiency is improved; meanwhile, a plurality of operators and strategies are adopted to improve the candidate path set, and a better solution is obtained.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.

FIG. 1 is a flow chart of a method of customizing bus services for a large-scale urban road network according to a first embodiment of the present invention;

FIG. 2 (a) is a graph showing the clustering result when the K value is 60 according to the first embodiment of the present invention;

FIG. 2 (b) is a graph of clustering results when the K value is 80 according to the first embodiment of the present invention;

FIG. 2 (c) is a graph of clustering results when the K value is 100 according to the first embodiment of the present invention;

FIG. 2 (d) is a graph of clustering results when K is 120 according to the first embodiment of the present invention;

FIG. 2 (e) is a graph of clustering results when the K value is 140 according to the first embodiment of the present invention;

FIG. 2 (f) is a graph of clustering results when the K value is 160 according to the first embodiment of the present invention;

FIG. 3 is a schematic diagram of average service radius and number of important sites according to a first embodiment of the present invention;

FIG. 4 is a graph showing average solution time for Gurobi and ALNS according to one embodiment of the invention;

FIG. 5 (a) is a graph of comparative results with respect to cost for the first embodiment of the present invention;

fig. 5 (b) is a graph of comparison results with respect to average travel time according to the first embodiment of the present invention;

FIG. 6 (a) is a schematic diagram of the evaluation result of the AWT according to the first embodiment of the present invention;

fig. 6 (b) is a schematic diagram of the evaluation result of ATT according to the first embodiment of the present invention.

Detailed Description

The invention will be further described with reference to the drawings and examples.

It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

Example 1

The embodiment provides a method for customizing bus service applicable to large-scale urban road network.

The method for customizing bus service applicable to the large-scale urban road network provided by the embodiment develops a customized bus stop selection and path optimization model aiming at large-scale commute requirements. On the one hand, the operation cost of operators and the waiting time of passengers are considered, a customized bus route optimization model is established, and a self-adaptive large-scale neighborhood search algorithm is adopted for solving. On the other hand, through the combined configuration of the large-scale service area and the bus line, the performance of the urban mass customized bus service is optimized. The result shows that for large-scale cases, the proposed customized bus service mode is approximately the same as a taxi in terms of average travel time of passengers, but the operation cost is saved by more than 30%. And the method is also superior to the existing solver Gurobi in terms of calculation efficiency. The method can effectively improve the satisfaction of passengers and increase the passenger flow of buses, thereby relieving traffic jam.

The method for customizing bus service applicable to large-scale urban road network provided by the embodiment, as shown in fig. 1, comprises the following steps: acquiring departure place, destination, departure time and arrival time of each passenger; based on departure places and destinations of all passengers, obtaining theoretical stations through clustering, and converting the theoretical stations to nearest bus stations to obtain customized bus stations; based on the customized bus stops, departure time and arrival time, the method aims at minimizing operation cost and waiting time of passengers, and adds constraint, and solves the customized bus stops and service starting time of each bus route.

The method for customizing the public transportation service suitable for the large-scale urban road network aims at providing low-carbon, environment-friendly and highly personalized travel service for urban commuters. Passengers must enter their travel requests before traveling; and comprehensively considering all travel requirements according to the constructed optimization framework, and configuring bus stops and routes. It is expected that the built framework will provide a timely and optimal solution to meet travel requirements. The application of the proposed optimization framework is based on the following assumptions:

(a) To a directed graph g= (N, a), N being a set of nodes comprising a start station O and an end station D, a being an arc connecting each pair of nodes (a= { (i, j): i e N, j e N, i noteqj }); in addition, each node i e N contains a service time s _i And each arc (i, j) is associated with a travel distance d (i, j);

(b) The capacity of the customized bus is fixed, denoted as C;

(c) Enough buses are available, and all buses running are not overloaded;

(d) For each passenger, his travel demand Q is desirably associated with departure Q.o, destination Q.d, estimated departure time q.ot, and estimated arrival time q.dt. I.e. each passenger has a departure time window q.ot-q.wp, q.ot+q.wp and an arrival time window q.dt-q.wd, q.dt+q.wd, where q.wp and q.wd represent buffering times to allow the operator to flexibly arrange the service plan;

(e) In order to ensure that each passenger can be served according to the travel plan, the bus is set to arrive before (q.ot-q.wp) and then waits for the corresponding passenger to be picked up;

(f) On the basis of analyzing the space distribution of the historical travel demands, the customized bus stop is determined. Passengers need to get on or off the bus at selected stations;

(g) When a bus is driven from one stop to another, the shortest route is taken.

The symbols used in this study are listed in table 1.

Table 1, symbols and definitions

When the CBTS system receives a travel requirement of a certain passenger, if the adjacent customized bus stop is within the acceptable walking range of the passenger, the passenger can be considered to obtain the customized bus service or can be covered by the customized bus stop network. The number and the positions of the customized bus stops are the key of selecting the addresses, and how to determine the proper number of the stops and find the positions of the stops meeting the requirements of passengers and meeting the actual road conditions has important significance for the subsequent customized bus route planning and design.

Compared with the common bus stop position, the traffic cell centroid is adopted as the site selection, and the customized bus is mainly considered in consideration of factors possibly influencing the passenger to select the customized bus trip. And establishing a two-dimensional coordinate system, and marking out the OD points (trip requirement points) of each trip requirement. By M _m ＝(x _m ,y _m ) The relative position of travel demand points m is represented, and all travel demand points are represented by U: = [ (x) _m ，y _m )] _{m＝1，2，…，u} And (3) representing. In addition, the present embodiment aims to minimize the total walking distance of all passengers, and the walking distance from the travel demand point to the station is represented by the euclidean distance, as shown in formula (1).

Wherein, (x) _n ,y _n ) Indicating the relative position of the station.

Further, a K-means clustering algorithm is adopted to solve and obtain the customized bus stop. The method comprises the following specific steps:

input: a demand set U; the number of clusters f is classified.

Step 1, setting an iteration count t=0, and initializing a central point of each cluster

Step 2, calculating the walking distance between the nth clustering center and the mth demand point

Step 3, distributing each demand point to a cluster center nearest to the demand point;

step 4, each cluster centerUpdating the average value of the n-th cluster demand points;

step 5, calculating a disfigurement Z to describe the sum of the distances in all clusters, wherein the smaller the distance is, the better the distance is;

step 6, if the clustering is converged, returning to a final clustering centerOtherwise, t=t+1 is set and step 3 is returned.

And (3) outputting: clustering the center point coordinates, namely theoretical sites.

After all theoretical stations are determined, the theoretical stations are combined with the existing bus stations according to the nearby principle, so that the construction cost of the CBTS system is reduced, and the optimization from the theoretical stations to the actual stations is realized. That is, according to the principle of nearby, the theoretical station is converted to the bus station, that is, the theoretical station is converted to the bus station closest to the theoretical station, so as to obtain the customized bus station.

Generally, for most traffic system planning problems, including CBTS systems, two aspects need to be considered: reachability and operating costs. However, the two are mutually conflicting, and when the service quality is improved by customizing the public transport operators, the operation cost is increased. Thus, the present embodiment builds a mathematical model that considers the operating costs and the passenger waiting time to balance the passenger interests and the operator costs.

The operating costs of CBTS system operators are related to a number of factors such as the total number of vehicles required and the cost of routes (fuel consumption, vehicle depreciation and staffing). As mentioned in the description of the problem, the entire CBTS system is intended to generate a set of vehicle trips, determining the number of vehicles and service routes to serve the passengers. Reducing the total number of service vehicles can significantly reduce operator costs. Likewise, the cost of operating routes formulated for service passenger demand should also be considered as operating costs.

In some vehicle path planning problems, the vehicle is allowed to arrive in advance or delayed, but a penalty function is added. However, in the latest CBTS system study, arrival delays are not allowed. In this embodiment, in order to improve the quality of service, this embodiment provides that the vehicle must arrive within the pickup station time window of each passenger as much as possible before the passenger's desired point in time.

Thus, the objective function combining the operation cost and the passenger waiting time is shown in formula (2).

Where K is the total number of vehicles, Q.t is the time desired to arrive at the station,the service start time for vehicle k to arrive at station n.

Model constraints are as follows:

(1) Vehicle allocation constraints:

each demand q can only be serviced by one bus:

furthermore, during service, passengers cannot transfer other buses:

if bus k passes arc (j, n + i),otherwise->n meterShowing the number of stops the vehicle k needs to travel for service demand q.

(2) Time constraint:

the boarding time of each passenger is within a specified time, namely, the boarding time of each passenger on the bus is a time window taking departure time as a center:

in addition, the bus should arrive at the boarding station before arriving at the corresponding alighting station, that is, the sum of the service start time of the bus at the starting station, the service time of the bus at the starting station and the travel time of the bus between the starting station and the destination station is less than or equal to the service start time of the bus at the destination station:

wherein,the service start time of site d is indicated.

Likewise, when bus k serves stations i and j successively, the arrival time at the stations should satisfy the following constraints;

wherein, And->The service start times for site i and site j are shown, respectively.

That is, assuming that the bus sequentially serves the first customized bus stop and the second customized bus stop, the sum of the service start time of the bus at the first customized bus stop, the service time of the bus at the first customized bus stop, and the travel time of the bus between the first customized bus stop and the second customized bus stop is less than or equal to the service start time of the bus at the second customized bus stop.

(3) Bus capacity constraint:

the custom bus cannot be overloaded due to safety requirements. Thus, at any given time, the number of passengers on the vehicle does not exceed the passenger capacity C of the bus:

L _k (i)+q _j ≤C,i,j∈N (9)

(4) Flow balance constraint:

and the continuity of the bus service chain is ensured. I.e. buses arriving at station i start from the same station.

The generation of all feasible solutions of the CBTS system model is an NP-hard problem, and the problem of overlong running time can occur when a precise algorithm is adopted to solve a large-scale example. Thus, the present embodiment proposes an adaptive large-scale neighborhood search Algorithm (ALNS) to solve this problem. ALNS is a meta-heuristic algorithm (MetaHeuristic Algorithm) that weights destruction operators and repair operators based on large neighborhood searches And->The algorithm can automatically select the operator improvement solution with better performance, so that the better solution can be obtained in reasonable running time. The algorithm selects a damage operator and a repair operator according to the weight of each operator by using a roulette principle, wherein the selection probability of the r damage operator is as follows:

the weights of the operators are dynamically adjusted according to the performance score change in the iteration process. When a new solution generated after operator search is a new global optimal solution, omega is given ₁ Dividing; when the new solution is better than the current solution, ω is given ₂ Dividing; when the new solution is worse than the current solution but accepted, ω is given ₃ Score, otherwise, omega is given ₄ Score, omega ₁ ≥ω ₂ ≥ω ₃ ≥ω ₄ . Equation (12) and equation (13) apply to the updating of the destruction operator and repair operator:

wherein lambda is E [0,1 ]]The attenuation parameter is used for controlling the sensitivity degree of the performance change of the destruction operator and the repair operator, and the unused operator weight in iteration is kept unchanged; psi phi type _r Is the fractional value of the operator.

The specific flow of the self-adaptive large-scale neighborhood search algorithm comprises the following steps:

input: feasible solution, x; operator weights, ω ^- ,ω ⁺ The method comprises the steps of carrying out a first treatment on the surface of the Removing probability, rate ^- The method comprises the steps of carrying out a first treatment on the surface of the A destruction operator and a repair operator, d _op ,r _op The method comprises the steps of carrying out a first treatment on the surface of the Score, ψ, ω of each operator ₁ ,ω ₂ ,ω ₃ ,ω ₄ ；

And (3) outputting: x is x _best ；

(1) Let x _best ＝x _current ＝x；

(2) Let ρ ^- ＝(1,…,1)；ω ⁺ ＝(1,…,1)；rate ^- = (0.1, …, 0.1); the iteration number m=0;

(3) Based on weight omega ^- Selecting a destruction operator; and based on the weight omega ⁺ Selecting a repair operator;

(4) If the operator d is destroyed _op Unchanged, let m=m+1；

(5) If m is>Threshold value, then randomly select the destruction operator d _op ；

(6)x _new ＝r _op (d _op (x,rate ^- ) And d), wherein _op For destruction, r _op To repair, get new solution x _new ；

(7) If c (x) _new )<c(x _best ) Where c is cost, then x _best ＝x _current ＝x _new ；Ψ＝ω ₁ ；

(8) If c (x) _new )<c(x _current ) X is then _current ＝x _new ；Ψ＝ω ₂ ；

(9) If accept (x) _new ) I.e. accept new solution (when new solution is worse than current solution but the difference is less than the threshold value, then new solution is accepted), x _current ＝x _new ；Ψ＝ω ₃ The method comprises the steps of carrying out a first treatment on the surface of the Updating the destruction operator d _op Probability rate ^- ；

(10) If none of the conditions of steps (7) - (9) is satisfied, ψ=ω ₄ ；

(11) Judging whether a termination condition is met, if not, returning to the step (2); otherwise, output x _best 。

The process of constructing the initial CB route is mainly divided into two steps, namely, forming a plurality of demand pools according to the starting and ending position information of the passenger demand. Secondly, a path which is as long as possible is formed in the demand pool.

In the path planning problem, adjacent nodes in the road network tend to form paths with shorter total paths more easily. Therefore, before forming the demand pool, it is necessary to arrange all demands in descending order of road network distance between the starting and ending points (custom bus stops), and execute the following processes in sequence: 1) If the current demand does not belong to any demand pool, the current demand is taken as the main demand of the new demand pool. There is one and only one master demand per demand pool. 2) After the new demand pool appears, traversing the ordered demand list, and adding the demands meeting the constraint conditions into the demand pool. The constraint conditions are as follows:

The earliest departure time of the passenger demand is not earlier than the earliest departure time of the main demand;

the latest arrival time of the passenger demand is no later than the latest arrival time of the main demand;

road network distance between the Pickup point of the passenger demand and the Pickup point/Delivery point of the main demand and TWdist smaller than the main demand;

the road network distance between the Delivery point of the passenger demand and the Pickup point/Delivery point of the main demand and TWdist smaller than the main demand;

non-primary demands may belong to multiple demand pools at the same time.

Wherein TWdist is the longest distance that the vehicle can travel in the earliest departure time and latest arrival time ranges.

And further utilizing the distributed demand pool to form a long path in the demand pool. There are also a number of possible solutions for the paths that each demand pool can form due to the different number of passenger demands in the demand pool. But as a result the initial solution path that is ultimately obtained is the path that contains the greatest demands of the passengers. Thus, in order to include as much demand as possible into the path, the present example uses an interpolation method to obtain an initial solution. The specific flow is as follows:

input: the distributed demand pool, two-dimensional array and P; start-end point demand, p ^o ,p ^d The method comprises the steps of carrying out a first treatment on the surface of the Optimal path Routebest _poor ；

And (3) outputting: initial feasible solution, two-dimensional array, x _init ；

Step (1), let i=0;

step (2), if pi]With unassigned pairs of orders, let j=0 to pi]Length, perform: if demand P [ i ]][j]The order pair is not allocated, p ^o ,p ^d ←P[i][j]；

Step (3), if p ^o ,p ^d Cannot be allocated to intermediate transition path Route _pool Then determine if it is the nth failed attempt (i.e., nth cannot be allocated) and if so, then follow the path Route _pool A pair of requirements (i.e. from Route) is randomly removed _poor remove a demand at random), and let j≡j-n; if p is ^o ,p ^d Can be allocated to a Route _pool Then p is ^o ,p ^d Distribution tasks assigned to path Route _pool . Wherein the attempt means to allocate an incoming path Route _pool In the trial and error method adopted in the embodiment, multiple insertion attempts are needed to determine whether the path can be really allocated; therefore, the trial-and-error number n needs to be set, and if the nth trial-and-error number is not reached, the trial is continued; after the nth time, randomly removing; and subtracting the trial-and-error times to allocate the next order requirement. Wherein, the vehicle can still serve p if it can serve all other requirements during the middle transition path ^o ,p ^d And meets the time window and vehicle load requirements, then is considered to be allocable; if p is inserted ^o ,p ^d The rear vehicles are considered to be unallocated if they cannot meet the time window and vehicle load requirements of other demands.

Step (4), if Route _poor .length>Routebest _poor Lengt (i.e. Route _poor The number of the customer points is larger than Routebest _poor Customer points on) then Route _poor ＝Routebest _poor ；

Step (5), for k from 0 to x _init Length, route _temp ←x _init [k]If Route _temp Can be combined with Routebest _poor Merging, then x _init [k]←Route _temp Merging Routebest _poor Otherwise, x _init .append(Routebest _poor ) (i.e., routebest) _poor As x _init Is a new path of (a); wherein, ζ represents replacement;

step (6), judging whether i is smaller than P.length, if yes, adding 1 to i, returning to step (2), otherwise, returning to x _init 。

The large-scale neighborhood search algorithm continuously generates new neighborhood solutions through a destroy-reconstruction strategy, so that better solutions are obtained, but a single delete or insert strategy easily causes the algorithm to fall into a local optimal solution. In the destruction phase, to obtain a diversified neighborhood solution, three destruction strategies are available to remove customer points on the path.

(1) And (5) randomly eliminating.

A random removal algorithm is used to select phi requirements from the original solution and remove the selected phi requirements from the corresponding route or trip.

(2) Worst rejection.

The operator aims to reduce the total driving range of the route. Selecting a demand and deleting the demand from the paths, and calculating path saving values before and after deleting the demand; then repeating the process for all demands and ordering the resulting values from big to small; after the first phi requirements are selected and deleted, the obtained path is the output path after the operator is executed.

(3) And (5) removing the correlation.

The operator is mainly used for removing demand points with higher correlation with selected demands, and in solving a vehicle path planning problem, pisinger and Ropke select the demands to be deleted by calculating the correlation among the demands by using the distance between the demands. Equation (14) describes the correlation between the two requirements of A and B.

Wherein o is the start of demand; d is the end point of demand; m represents the number of non-zero entries, if all nodes are different from one site, then m=4.

In the repair phase, three types of insertion operators are used:

(1) Randomly inserting.

And randomly inserting the removed phi requirements into any line according to the initial solution generation mode.

(2) Greedy insertion.

And (3) putting the removed phi requirements in the original solution in isolation, searching the position with the least cost increment of all the nodes to be inserted and the cost increment thereof, and finally selecting the node with the least cost increment from all the required nodes to be inserted to the optimal position by comparing the increment sizes, and continuously executing the processes until all the nodes are inserted into the original travel solution.

(3) Unfortunately, insertion.

When the path is onlyWhen there are few nodes that can move, greedy insertion operators often need to iterate to a later stage to meet the conditional insertion. To solve this problem, the reglet's criterion is used by the insertion operator. Let Deltac _i The cost value saved after searching the optimal insertion position of the node i is represented, and the calculation formula is as follows:

wherein Γ represents the node value to be inserted;representing a cost saving value corresponding to the first optimal insertion point; />And representing a cost saving value corresponding to the second optimal insertion point. And selecting the optimal insertion position according to the formula (15) during each iteration until all client points in Γ are inserted into the path.

In this embodiment, the CBTS system is designed in the central urban area of tokyo. Compared with the conventional bus, the customized bus has advantages in comfort, rapidity, accessibility and the like, and the taxi has similar characteristics with the customized bus in the aspects. Therefore, taxis (including network bus) contain a large number of potential customized bus travel demands in the passenger group in the commute period, and can be used as the basis for selecting customized bus stops. In the embodiment, 7-9 pieces of taxi data (including net taxi) of 5 working day early peaks from 4 months, 6 days, 4 months and 10 days in 2016 in Nanjing city are taken as study samples, and 39463 pieces of effective data are obtained. The processed effective field comprises the boarding time and the boarding requirement longitude and latitude coordinates, and a taxi requirement point distribution map is obtained. Based on the method, a customized bus stop is arranged.

In the experimental section, a set of CB networks with different site numbers was used to verify the validity of the proposed model and solution. Based on the OD station pair data with different starting and ending points generated randomly by the customized bus station, 10,20,30,40,60,80,110 station data cases with different scales are formed, and the demand is generated randomly. Networks are divided into three scales: small (10, 20,30,40 sites) and large (60, 80, 112 sites). Meanwhile, comparing the optimal solution obtained by the algorithm and the Gurobi solver based on a small-scale case, and verifying the effectiveness of the algorithm; all experiments were performed on a computer configured as Intel (R) Core (TM) i5-8300H CPU@2.30GHz with 16G memory, with all algorithms and test programs programmed to JetBrains PyCharm 2019.

The central urban area of Nanjing and the surrounding 600 square kilometers are taken as a research area, the demand coordinates in a CBTS system service area are screened out, and the demand of passengers is clustered by using a site selection algorithm. The clustering centers were marked by taking K values of 60,80, 100, 120, 140, and 160, respectively, to obtain cluster scatter plots 2 (a), 2 (b), 2 (c), 2 (d), 2 (e), and 2 (f). The average radius of service for sites at each K value, the number of sites servicing over 100 persons, the profile factor (Silhouette Coefficient), and the davison fort Ding Zhishu (Davies-Bouldin index) were also counted, see Table 2 and FIG. 3.

Table 2, quantization index

As can be seen from table 2, as the K value increases, the profile coefficient increases linearly, and davison burg Ding Zhishu decreases linearly, indicating that the larger the K value, the higher the clustering effectiveness. Meanwhile, the average service radius of the station is reduced, so that the walking time of the passengers to the station is shortened, and the satisfaction of the passengers is improved. However, the number of sites available to service more than 100 people decreases as the K value increases. For customized public transport operators, the increase of the number of stations can disperse the demands of passengers in the area, increase the bus detour distance and cause the waste of bus resources.

The optimal service radius of the customized bus is 700m, in three conditions of 120, 140 and 160 of K values meeting the requirements, the number of stations for serving more than 100 persons is the largest under the condition of K=120, and the average service radius is 759m, so that the passenger travel satisfaction degree and the bus operation cost can be considered. Therefore, the number of stations is set to 120, and the initially selected customized bus station is obtained. Because the coordinates of the stations solved by the model are all theoretical values, the coordinates may deviate from the actual road, and the positions of the stations need to be optimized. In order to reduce the construction cost of the CBTS system, the acceptance of the passenger to the customized bus stop in the early operation stage is considered, the customized bus stop is arranged at the conventional bus stop according to the nearby principle, and the stops with the distance less than 100m are combined.

In order to solve the line optimization model, parameters in the model are calibrated by combining the actual running condition of the customized buses. According to the demand characteristics of customized buses, medium buses with maximum passenger capacity of 40 persons and vehicle speed of 40km/h are adopted, and the number of the buses is not limited. The fuel consumption cost per kilometer is 1.2 yuan/kilometer, the fixed operation cost per vehicle is 100 yuan/kilometer, and the outdoor waiting time cost is 0.2 yuan/min according to the optimized customized buses, road network information and travel demand information of each station. Experiments were performed through small-scale problems to verify the effectiveness of the proposed model and solution. The small-scale cases contained a total of 4 experiments: 10 nodes, 20 nodes, 30 nodes and 40 nodes, each experiment was run in 10 simulations to evaluate the randomness of the ALNS results.

Taking the 20-site case as an example, table 3 compares the performance of the ALNS algorithm with that of a commercial slowness Gurobi. It can be seen that in both solution algorithms, 4 buses are required to meet the trip requirement of 20 stops. The ALNS algorithm achieves a total travel distance and travel time that is similar to the Gurobi algorithm, but the overall cost of the system is somewhat higher.

TABLE 2ALNS and Gurobi comparison results

/>

Experiments further tested the applicable effects of ALNS and Gurobi at 10, 30 and 40 site scales. The average calculation time of Gurobi and ALNS in small-scale case testing is shown in fig. 4. Fig. 4 shows that Gurobi is able to optimize a small-scale CB network. However, as the number of nodes increases, the computation time increases exponentially, and 31 hours are used when solving 40 site cases, so that the solution cannot be found within a limited time. The ALNS solves the small-scale cases with the time within 30s, and the operation speed is obviously superior to that of Gurobi.

Table 4 summarizes the statistics that as the number of nodes increases, the minimum system cost of the Gurobi and ALNS solutions increases, while the resultant randomness of ALNS increases (e.g., SD and CV). In terms of result accuracy, the result obtained by ALNS in the 10-site case is the same as the standard result; 20. the 30 and 40 site case results differ from the standard results by about 10%.

TABLE 4 results of small-scale case analysis

Comprehensively considering the calculation time and accuracy, the embodiment provides that the system cost obtained by model solving is approximate to the optimal result, the model solving efficiency is obviously superior to Gurobi, and the algorithm calculation performance of the embodiment is effective and has accuracy. While the Gurobi is 31 hours when solving the 40-site case, in the subsequent large-scale case, the Gurobi is difficult to directly solve, so that the algorithm proposed by the embodiment can be used for substitution.

To evaluate the effectiveness and applicability of the developed models and algorithms to large-scale cases, 60, 80, and 112 site cases were designed. Two evaluation indexes were used: the average waiting time AWT and the average travel time ATT, and the algorithm effect are evaluated as shown in formulas (18) and (19).

Where AWT represents the time Q.t of all passengers serviced by vehicle k to make a request and the time T of arrival at the station of vehicle k _k (j) Average value of time difference, if the vehicle arrives before Q.t, the time difference is 0; ATT represents the average of the vehicle hours of all passengers serviced by vehicle k; q _k The number of passengers servicing vehicle k.

Table 5 lists examples of 60 stops, including the on-board stop ID, the off-board stop ID, the on-board and off-board expected time, and the stop passenger demand. The default buffer time (buffer time) is 5min.

Table 5, data example

Index	Boarding station point ID	Get-off station ID	Desiring to get on the vehicle time	Time when getting off is expected	Demand number
						1	24	4	7:05	7:35	13
2	8	21	7:23	7:40	15
						3	39	74	7:10	7:35	13
4	47	27	7:20	7:50	12
						5	15	40	7:10	7:35	8
…	…	…	…	…	…
						30	72	85	7:05	7:45	17

TABLE 6 results of large-scale case analysis

Table 6 shows the large-scale case statistics. As the case size increases, the standard deviation SD and coefficient of variation CV of the model results increase. The reason is that the number of vehicles is not limited, and part of the results tend to send out 1-2 vehicles in the optimization process so as to reduce the waiting time of passengers, and the system cost is improved slightly. The average waiting time AWT of the passengers is 2.11min and less than 5min (buffer time) in the 112-station case, which indicates that the model can meet the travel demands of the passengers in the large-scale case optimization. 60. The average travel time ATT of the 80-station case is 20.35min and 19.875min respectively, which are slightly higher than the average travel time expected by passengers by 20.08min and 19.28min. The average travel time ATT of the 112 sites is 19.775, compared with the expected average travel time of the passengers of 23.57min, 19.19% is saved, and the model effect is good.

Experiments further verify the difference between CBTS and taxis in the same demand for service. The fuel consumption of the taxies is 0.6 yuan/km, and the fixed operation cost of each vehicle is 20 yuan.

From the results shown in fig. 5 (a) and 5 (b), CB service may travel a longer distance than taxi service, but save 30% of system cost. Notably, in the 110-station case, the average journey time of passengers is approximately the same, but the cost of the CBTS system is saved by 34.3% compared with that of taxi service, so that the operation cost can be effectively reduced.

Taking 60, 80, 110 station scales as an example, the buffer time of the passenger demand is changed to be 4, 6, 8 respectively, and the optimization results under different time window widths are shown in table 7.

TABLE 7 results of different buffering times

As can be seen from table 7, the cost value decreases with increasing time window width, and the decrease is gentle. In addition, the relaxed time window allows the vehicle to pick up more passengers in the path, and the wider the time window, the fewer vehicles are needed for different number of stops.

Fig. 6 (a) and 6 (b) are results of the average waiting time AWT and the average trip time ATT of passengers at different buffering times in each case. As the width of the time window increases, the waiting time and travel time of the passenger at the station correspondingly increases. In the case of 110 stations, when the buffer time is 5min, the AWT and ATT are greatly amplified, and the travel cost of passengers is increased; similarly, in the case of 60 and 80 sites, the two indexes are greatly amplified when the buffer time is 6 min.

In summary, the width of the time window affects the path planning result, the longer the buffering time is, the more beneficial to the formulation of the ride-through scheme, and the system cost is reduced, but the travel cost of passengers is correspondingly increased.

As an emerging public transportation service, customized buses can better meet the travel demands of passengers with similar travel preferences than conventional buses, and have proven to be an economical and efficient commuting solution. Custom bus passengers are more focused on timeliness and comfort than passengers served by conventional buses. Thus, the present embodiment combines the concepts of the PDPTW problem to build an optimization model and optimize the CBTS system from a site, passenger, and operator perspective. In the aspect of the bus stop, the distribution characteristics of the commute demands of passengers in the rush hour are studied, the problem of the acceptance of passengers in the early stage of system construction is considered, and the custom bus stop is optimized and customized on the basis of the conventional bus stop. For passengers, the waiting time of the passengers is introduced to measure late arrival punishment of buses after the expected receiving time of the passengers, the service time needs to meet the receiving time window of each passenger, the passenger satisfaction is improved, meanwhile, the passenger flow of the buses is increased, and the operation cost is reduced. In the aspect of public transport operators, under the constraints of vehicle capacity, vehicle speed and the like, the fixed operation cost and the fuel consumption of the vehicles are considered, so that the operation cost is minimized. Further, a self-adaptive large-scale neighborhood searching algorithm is adopted to solve a customized bus route optimization model, so that model solving efficiency is improved. Meanwhile, a plurality of operators and strategies are adopted to improve the candidate path set, and a better solution is obtained.

Taking the actual road network of the central urban area of Nanjing as an example, the effectiveness of the method in the aspect of reducing the operation cost of the system is verified. In addition, the embodiment compares the operation cost of customized bus service and taxi service, and performs sensitivity analysis on the buffering time to obtain the following conclusion: (1) Compared with taxi service, the customized bus service saves more than 30% of operation cost, and the time loss is only 14%. (2) The larger the buffering time is, the lower the operation cost is, which is more beneficial to the formulation of the ride-through scheme, but the travel cost of passengers is correspondingly increased. Research results show that customizing public transportation service is an economical and effective choice for urban commuters.

Future research will be directed to further revising the path optimization model. On the one hand, in the actual operation process, heterogeneous vehicles based on different requirements can reduce the operation cost of operators, and the problem can be optimized into a heterogeneous vehicle scheduling problem. On the other hand, the present study assumes that traffic conditions are in ideal traffic conditions, and does not take traffic congestion, intersection delays, and the like into consideration. In future research, road traffic conditions may be introduced into the model to accurately calculate the travel time of the customized bus service.

In order to alleviate the increasing traffic congestion, public transportation systems are required to not only provide flexible and reliable on-demand transportation, but also make full use of existing road infrastructure. The embodiment provides customized public transportation service suitable for a large-scale urban road network, and the public transportation station and the scheduling strategy are jointly optimized. Firstly, determining the position of a bus station by adopting a K-means clustering algorithm based on the spatial distribution characteristics of historical commute demands; the bus route optimization problem is then expressed as a Mixed Integer Linear Programming (MILP) model that takes into account system costs and passenger waiting times. In order to effectively apply the model to urban environments, an adaptive large-scale neighborhood search (ALNS) algorithm is proposed to solve the MILP model. And finally, verifying the effectiveness of the proposed customized public transportation service method by utilizing the data of small and large scenes of Nanjing in China, and performing sensitivity analysis on the buffering time of passengers. The result shows that the customized bus service mode is superior to the traditional taxi service in the aspects of average waiting time of passengers and average travel time of buses, the operation cost is saved by more than 30%, and the calculation efficiency is obviously superior to that of a Gurobi solver. The method can improve the satisfaction of passengers, increase the traffic flow and relieve traffic jams.

Example two

The embodiment provides a system for customizing bus service applicable to a large-scale urban road network, which specifically comprises the following steps:

It should be noted that, each module in the embodiment corresponds to each step in the first embodiment one to one, and the implementation process is the same, which is not described here.

Example III

The present embodiment provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps in a method for customizing bus services for a large-scale urban road network as described in the above embodiment.

Example IV

The present embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor executes the program to implement the steps in a method for customizing bus services applicable to a large-scale urban road network according to the first embodiment.

The data acquired by the method, the device (system) and the computer program product are only used for legal scenes, wherein the data comprises the departure place, the destination, the departure time, the arrival time and the like of passengers.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

Claims

1. A method of customizing bus services for a large-scale urban road network, comprising:

2. A method of customizing a bus service for a mass transit network as claimed in claim 1, wherein vehicle allocation constraints are added during said solving;

3. A method of customizing a bus service for a mass transit network as claimed in claim 1, wherein said time constraints further comprise: and the sum of the service start time of the bus at the start station, the service time of the bus at the start station and the running time of the bus between the start station and the end station is less than or equal to the service start time of the bus at the end station.

4. A method of customizing a bus service for a mass transit network as claimed in claim 1, wherein said time constraints further comprise: assuming that the bus sequentially serves a first customized bus stop and a second customized bus stop, the sum of the service start time of the bus at the first customized bus stop, the service time of the bus at the first customized bus stop and the running time of the bus between the first customized bus stop and the second customized bus stop is less than or equal to the service start time of the bus at the second customized bus stop.

5. The method for customizing bus service for a large-scale urban road network according to claim 1, wherein the solving process adopts a self-adaptive large-scale neighborhood searching algorithm, and the initial solution is obtained by adopting an interpolation method.

6. The method for customizing a bus service for a large-scale urban road network according to claim 5, wherein the self-adaptive large-scale neighborhood search algorithm adopts three destruction strategies of random elimination, worst elimination and related elimination in a destruction stage.

7. The method for customizing a bus service for a large-scale urban road network according to claim 5, wherein the adaptive large-scale neighborhood search algorithm adopts three insert operators of random insert, greedy insert and regrette insert in a repair stage.

8. A system for customizing bus services for a large-scale urban road network, comprising:

9. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps in a method of customizing bus services for a mass transit network according to any one of claims 1-7.

10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of a method of customizing bus services for a mass transit network as claimed in any one of claims 1 to 7.