CN110458456A - The dispatching method and system of demand response formula public transit system based on artificial intelligence - Google Patents
The dispatching method and system of demand response formula public transit system based on artificial intelligence Download PDFInfo
- Publication number
- CN110458456A CN110458456A CN201910741446.2A CN201910741446A CN110458456A CN 110458456 A CN110458456 A CN 110458456A CN 201910741446 A CN201910741446 A CN 201910741446A CN 110458456 A CN110458456 A CN 110458456A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- demand
- max
- artificial intelligence
- overall cost
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000004044 response Effects 0.000 title claims abstract description 50
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000011156 evaluation Methods 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000005457 optimization Methods 0.000 claims abstract description 23
- 238000004088 simulation Methods 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 230000037213 diet Effects 0.000 claims description 4
- 235000005911 diet Nutrition 0.000 claims description 4
- 230000008676 import Effects 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 8
- 230000001419 dependent effect Effects 0.000 abstract description 4
- 230000032258 transport Effects 0.000 description 26
- 230000006870 function Effects 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 9
- 238000011160 research Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 230000003068 static effect Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 239000000523 sample Substances 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000002045 lasting effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000013439 planning Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000003915 air pollution Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012384 transportation and delivery Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Educational Administration (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The dispatching method and system for the demand response formula public transit system based on artificial intelligence that the invention discloses a kind of, method include: that S1 generates starting point, the requirements set of the point of arrival and demand moment;S2 gathers building initial plant scheduling scheme according to demand;S3 calculates all evaluations of picking for ordering vehicle demand according to initial plant scheduling scheme and divides;S4 will order vehicle demand properties, vehicle attribute, running environment attribute and pick evaluation and point be put into artificial intelligence model and be trained;Wherein, ordering vehicle demand properties, vehicle attribute and running environment attribute is independent variable, picks evaluation and is divided into dependent variable;S5 picks evaluation point according to what training prediction obtained, and selection is picked the optimal vehicle of evaluation point and picked;Compared with prior art, present invention introduces artificial intelligence technology, the scheduling model optimization and the vehicle scheduling of lower overall cost for learning by oneself response type are realized.
Description
Technical field
The present invention relates to the scheduling field of public transit system, in particular to a kind of demand response formula public transport based on artificial intelligence
The dispatching method and system of system.
Background technique
Demand response formula public transit system is a kind of special bus operation mode, is with passenger demand for guiding, not by solid
The limitation of alignment road, provide that more people multiply altogether through or picks public transit system.Since its website is stopped completely according to the trip of passenger
Depending on, therefore for relatively conventional public transport, response type public transport is not limited by route website, can left home on nearest any website
Lower visitor can provide servicing without transfer close to " door-to-door ", and can also match the service mode of the following unmanned bus.
Traditional response type public transit system is also known as mobile type bus service (Flexible Transit Services) in state
30 years Experience in Development have been had more than outside, concept most proposes (Daganzo, 1984) earlier than the eighties in last century, subsequent nearly 30
Research and practice over year about the field is in lasting expansion.According to the definition of U.S.'s Public Transport Cooperation research project survey report
(Koffman, 2004), be further subdivided into 6 kinds of operation modes: variable line public transport (Route deviation) can be changed
Website public transport (Point deviation), website demand response formula public transport (Request stops), demand response formula pick public affairs
(Demand-responsive connector), part route is handed over to can be changed public transport (Flexible-route segments), area
Domain route public transport (Zone route).Passenger can be by terminus of going on a journey of reserving by phone, by public transport company according to all demands
Place and arrangement of time public transit vehicle are picked.There is spirit in the area Sweden Plustrafik and the Hibernian area Bealach
The typical case of type bus service living, it is that system is picked in a kind of public transport for serving primarily in low-density area.
In above-mentioned six kinds of mobile types bus service mode, " demand response formula picks public transport " is the highest mould of flexibility ratio
Formula, substantially process are as follows: vehicle provides demand response type bus service in a specific region, and start, end are operation model
Interior urban railway station or regular public traffic hinge are enclosed, settable fixed station, can also basis for passenger getting on/off in the area
Passenger demand is flexibly got on or off the bus, but the route of vehicle each run is all adjusted flexibly according to passenger demand.
With the development of internet, the novel traffics service mode such as net about taxi, shared bicycle starts to traditional friendship
Logical means of transportation produces certain competition and impact.And with the continuous breakthrough of unmanned technology, also will further it expedite the emergence of
The following efficient and convenient and fast trip mode.Under such historical background, the concept of response type public transit system is redefined
A kind of digitlization scheduling variable line bus operation mode meeting middle-high density demand.
Novel response type public transit system is a kind of combination of net about vehicle and auxiliary type public transport, and operation logical AND is traditional
Response type public transport is similar, is equally that passenger proposes demand, then is met by vehicle scheduling.Unlike, novel response
Formula public transit system is passenger in mobile phone mobile terminal proposition real-time requirement, and vehicle need to be dispatched in selection path in complicated demand network
Vehicle is picked.Therefore, support the key of novel response type public transport just in the response system of its efficient quick, and with minimum
Vehicle, shortest deadhead operation distance, complete period of reservation of number it is most short, arrive at the destination most fast vehicle scheduling.
In development process, response type public transit system is mainly the trip requirements for meeting passenger and changing to intensive public transport, drop
The both ends time of low tradition transfer manner, the cost of last one kilometer is reduced, competitive way is mainly taxi, net about vehicle etc..
In view of the peak valley feature in resident trip division of day and night, response type public transit system can also provide light-duty logistics service at flat peak,
Such as express delivery or shared bicycle carry scheduling, to cut operating costs.Into unmanned period, response type public transport passes through it
The advantages such as efficient scheduling, the comfortable service of a people one, cheap price, it would be possible to become resident trip mode, become private
The main competition object of family's vehicle.
As technology develops, the coverage of response type public transport can also be each website in region beyond the range picked
Between trip bus service is provided, picking can be used as the higher service of priority wherein and treats, in this way, entire response type public affairs
The available raising of the service efficiency of friendship.
In order to fill up the unserviceable area of city rail and regular public traffic, " last one kilometer " problem of resident trip is solved,
It is come into being with the community bus that " being sent to door " is service aim.From the whole nation first community bus --- Shang Haipu in 2006
From eastern 84 road transports battalion, various forms of community bus continue to bring out in each city.
Accurate analysis of the community bus of early stage due to shortage to passenger demand, causes actual operating state not ideal enough,
There is the situation that vehicle no-load ratio is high, operation is unable to make ends meet.After opening such as Nanjing Line 2 Metro in 2010,4 have been opened
The community bus of subway station is picked, but cooperates with consideration with passenger demand since route, website layout lack, operation 2 days occurs
Fall into the awkward situation of stoppage in transit.With Information Technology Development in recent years and the continuous depth to " last one kilometer " properties study
Enter, the operation optimization of community bus is in lasting progress.If 2011 Shanghai Pudong buses take " big carriage type mixed operation ",
The peak period peace peak period takes the vehicle of 10 meters and 8 meters respectively, preferably reduces operation cost;Wenzhou Long Wan in 2016
5 community bus carried out after indefinite station stops, shortens a series of adjustment such as route, day the volume of the flow of passengers be promoted to 2989 people/day,
Increase by 32% on a year-on-year basis;Quanzhou in 2017 has carried out complete upgrading to community bus, using pure electric vehicle sightseeing battery truck, with indefinite
Line, unfixed point, HOP ON-HOP OFF mode operation, by getting on the bus, either order is sent to, and is developed cell phone application and carried out network about vehicle.
All in all, " last one kilometer " that community bus mainly undertakes all kinds of Public Transport Junctions to community picks function,
Optimization direction is from traditional " vehicle, is determined at alignment in fixed station " to service personalization shift in demand.The current country is for picking bus
It research and practices still in probe phase, how more accurate response passenger demand and to ensure that bus operation main body benefit becomes
The problem of urgently furtheing investigate.
Theoretical research both at home and abroad about " demand response formula picks public transport " at present is more, but practices less, and research is main
Concentrate on optimization object function building, system core parameter setting, three broad aspect of line optimization dispatching algorithm.
Di Di company and September in 2016 22 days belong to demand response formula public transit system, lead in the online advantage bus in Shenzhen
Excessive intelligent data analysis and line optimization realize more complicated route distribution and operation management on some scale.Day
3000 person-times are serviced, 2700 kilometers of daily operating mileage has been more than American-European normal response formula public transport scale.But it is in operation
There is fixed vehicle traffic direction, is not complete dynamic allocation vehicle, customization public transport is more closely similar in operation mode, visually
For the customization public transport of demand response.
Based on state algorithm, a small amount of dynamic dispatching algorithm research is also based on response type bus dispatching algorithm at this stage
The optimization that static scheduling carries out, it is difficult to adapt to high-frequency dynamic trip requirements, and the dynamic needs such as net about vehicle, taxi
Trip mode be not present rideshare scheduling problem, Dispatch by appointment algorithm is for response type public transport and is not suitable for.
Summary of the invention
It is a kind of based on artificial the technical problem to be solved by the present invention is in order to overcome the above-mentioned defects in the prior art, provide
The dispatching method and system of the demand response formula public transit system of intelligence.
The present invention is to solve above-mentioned technical problem by following technical proposals:
In a first aspect, the present invention provides a kind of dispatching method of demand response formula public transit system based on artificial intelligence, packet
Include following steps:
S1, starting point, the point of arrival, the requirements set at demand moment are generated;
S2, initial plant scheduling scheme is constructed according to the requirements set;
S3, according to the initial plant scheduling scheme calculate it is all order vehicle demand pick evaluation point;
S4, vehicle demand properties, vehicle attribute, running environment attribute will be ordered and pick evaluation and point be put into artificial intelligence model
In be trained;Wherein, vehicle demand properties, vehicle attribute and the running environment attribute ordered is independent variable, picks evaluation
It is divided into dependent variable;
S5, evaluation point is picked according to what training prediction obtained, selection is picked the optimal vehicle of evaluation point and picked.It is optional
Ground specifically includes in the step S2:
In the case where meeting default constraint condition, the vehicle for selecting vehicle distances demand point nearest is picked.
Optionally, the constraint condition includes at least:Pi<Pmax、Ti≤Tmax、Si≤Smax;
Wherein, LxyIt is stroke xy around row distance, lxyFor the shortest distance of stroke xy, γmaxFor the maximum system of detouring of constraint
Number;PiFor vehicle carrying number, PmaxFor vehicle maximum carrying number;TiFor Waiting time, TmaxTo constrain maximum Waiting time;SiFor
Stops, SmaxTo constrain most stop frequencies.
Optionally, it is specifically included in the step S3:
All overall cost UC for ordering vehicle demand are calculated according to initial plant scheduling schemeIt is comprehensive:
Wherein, Q is response type public transit vehicle sum in coverage;N is user demand total amount in region;T is the time;D
For distance;k1~k4For weight term;UCIt is comprehensiveUnit be value of utility, i.e., each index item is converted into the value after Costco Wholesale;θ
For other unforeseen costs;
All evaluations of picking for ordering vehicle demand, which are calculated, according to the overall cost divides δ=UCmin/UC;
Wherein, UC indicates every calculating cost for ordering vehicle demand, UCminIndicate that this orders the shortest path N-free diet method of vehicle demand
The objective cost of time,
UCi=k2×tIt waits+k3×tIt is interior+k4×DTraveling+k5×β;
Wherein, k5β is the residual error for picking value of utility every time.
Optionally, it is specifically included in the step S4: ordering vehicle demand for every, select Pi<PmaxAnd Di≤DmaxVehicle
It calculates in the artificial intelligence model and picks evaluation point;Wherein, PiFor vehicle carrying number, PmaxFor vehicle maximum carrying number,
DiIndicate the distance of vehicle i distance requirement point, DmaxExpression system maximum picks tolerance.
Optionally, the method also includes: if meeting Pi<PmaxAnd Di≤DmaxVehicle number be less than preset threshold, then hold
Poor range increases, and range of tolerable variance is not more than the coverage of demand response formula public transit system, until meeting the vehicle of above-mentioned condition
Number be greater than minimum standard number.
Optionally, it is specifically included in the step S4: placing into the artificial intelligence after being pre-processed to independent variable
It is trained in model.
Optionally, further comprising the steps of after the step S5:
If S6, system operation time are greater than optimization cycle of training, actual operation data are put with evaluation point is actually picked
Enter and is trained in the artificial intelligence model;
S7, training after the completion of by actual demand data import simulation system in, using Optimized model as pick evaluation divide in terms of
Foundation is calculated, operation simulation, and the overall cost of calculating simulation scheduling are carried out to actual demand;
S8, the overall cost by the overall cost of the operation simulation in actual schedule compare, if in the simulation
Occur being higher than the Arbitrary Term of the overall cost of the actual schedule in the overall cost of scheduling, then in overall cost calculation formula
Constant term be modified.
Optionally, optimize the calculation formula of cycle of training in the step S6 are as follows:
Wherein, m is minimum optimization cycle of training;UCIt is practical comprehensiveFor the overall cost of actual schedule;UCI predictionFor vehicle i simulation
The overall cost of scheduling;α is constant term;N is user demand total amount in region.
Second aspect, the present invention provide a kind of scheduling system based on artificial intelligence demand response formula public transit system, comprising:
Memory, for storing computer program;
And processor, for executing the computer program, to realize method as described in relation to the first aspect.
The third aspect, the present invention provide a kind of computer storage medium, and being stored in the computer storage medium can hold
Line program, to realize method as described in relation to the first aspect when the executable program is performed.
On the basis of common knowledge of the art, above-mentioned each optimum condition, can any combination to get each preferable reality of the present invention
Example.
Compared with prior art, present invention introduces artificial intelligence technologys, according to the data building for meeting true trip characteristics
Passenger demand is dispatched using simulation software simulating vehicle;Integrated objective function is constructed, determines to go out every time according to routine dispactching algorithm
Capable picks evaluation point, predicts in dynamic context the evaluation point of picking of different vehicle scheduling, and press optimum prediction value
It dispatches buses, generates actual operation data.By the actual operating data accumulation in one section of period, by artificial intelligence model
It practises, relationship of the evaluation point between route choosing, and constantly study optimization are picked in building, finally realize efficient self study intelligence
Energy scheduling model, to realize intelligentized dynamic line planning.
Detailed description of the invention
Fig. 1 is a kind of dispatching method of the demand response formula public transit system based on artificial intelligence provided in an embodiment of the present invention
Flow chart.
Fig. 2 is the dispatching party of another demand response formula public transit system based on artificial intelligence provided in an embodiment of the present invention
The flow chart of method.
Fig. 3 is the dispatching party of another demand response formula public transit system based on artificial intelligence provided in an embodiment of the present invention
The flow chart of method.
Specific embodiment
The present invention is further illustrated below by the mode of embodiment, but does not therefore limit the present invention to the reality
It applies among a range.
The dispatching method for the demand response formula public transit system based on artificial intelligence that the present embodiment provides a kind of, as shown in Figure 1,
The following steps are included:
Step 101 generates starting point, the point of arrival and the requirements set at demand moment: { (Ox;Dx;Tx)|x∈N,1≤x
≤ N }, wherein OxFor starting point, DxFor the point of arrival, TxFor the demand moment, N is aggregate demand quantity in network.
In optionally a kind of embodiment of step 101, the first modeling for specific region orders vehicle firstly the need of prediction
Demand, wherein it includes being not limited to that the basic data that vehicle demand is ordered in prediction, which is collected: passenger demand questionnaire survey, net about vehicle are hired out
Vehicle operation data, mobile phone signaling data etc..By the cleaning to basic data, processing, expansion sample, starting point, the point of arrival, need are obtained
Seek the forecast demand set at moment.
Step 102 constructs initial plant scheduling scheme according to the requirements set.
In step 102, in order to meet model training requirement, multiple and different initial demand set need to be constructed, is needed between set
There are notable differences.According to obtained requirements set, initial plant scheduling scheme is constructed, the purpose is to provide artificial intelligence mould
The data of type training.
In optionally a kind of embodiment, initial schedule is static scheduling model, and dispatching method is there are many selection
The dynamic dispatching model of cooperation artificial intelligence driving, initial static scheduling use maximum compatibility model, i.e., meet it is default
Under constraint condition, the vehicle for selecting next destination distance requirement point of vehicle nearest is picked, i.e. selection Min { LO1,
LO2,LO3…LOnVehicle picked.LOnTo meet the next destination of n vehicle of constraint condition apart from the demand most
Short path.
In optionally a kind of embodiment, the constraint condition is included at least:Pi<Pmax、Ti≤
Tmax、Si≤Smax;
Wherein, LxyIt is stroke xy around row distance, lxyFor the shortest distance of stroke xy, γmaxFor the maximum system of detouring of constraint
Number;PiFor vehicle carrying number, PmaxFor vehicle maximum carrying number;TiFor Waiting time, TmaxTo constrain maximum Waiting time;SiFor
Stops, SmaxTo constrain most stop frequencies.
Step 103, according to the initial plant scheduling scheme calculate it is all order vehicle demand pick evaluation point.
Construct the analog simulation operating system based on road network, analogue system can input area road network and website net coordinate, root
According to dynamic need matrix allocation OD amount.When output valve includes that the OD of each demand waits, response time, response distance, user
Between, user's riding time, Multi-feed-eye ring;Operating mileage, total passenger, dead mileage of each car etc..
In optionally a kind of embodiment, after each virtual stroke, analogue system records the output of the trip
Attribute, and be stored in background data base.Simulation system is run multiple times, enough learning samples is accumulated and (is greater than 50000
Item).
Reasonable objective function is constructed, since in determining demand program simulation, all demands all finally obtain satisfaction,
Therefore, in the sample training of artificial intelligence model, the value of utility of evaluation response type public transport should be its overall cost, be a knot
The operation costs such as vehicle expense, driver's expense, energy consumption, running cost and period of reservation of number, running time are purchased in conjunction
The integrated value of equal user costs.
In optionally a kind of embodiment, target is with least vehicle, shortest unloaded distance, passenger waiting time
It is most short, fastest to the vehicle scheduling for reaching destination, specifically include in step 103:
All overall cost UC for ordering vehicle demand are calculated according to initial plant scheduling schemeIt is comprehensive:
Wherein, Q is response type public transit vehicle sum in coverage;N is user demand total amount in region;T is the time;D
For distance;k1~k4For weight term;UCIt is comprehensiveUnit be value of utility, i.e., each index item is converted into the value after Costco Wholesale;θ
For other unforeseen costs;
All evaluations of picking for ordering vehicle demand, which are calculated, according to the overall cost divides δ=UCmin/UC;
Wherein, UC indicates every calculating cost for ordering vehicle demand, UCminIndicate that this orders the shortest path N-free diet method of vehicle demand
The objective cost of time;
In present embodiment, model overall goal is UCIt is comprehensiveMinimum, can be by UCIt is comprehensiveIt divides to each demand in picking,
It is to be understood that
UCi=k2×tIt waits+k3×tIt is interior+k4×DTraveling+k5×β;
Wherein, k5β is the residual error for picking value of utility every time,The calculating of residual error includes
Vehicle purchases hiring cost, depreciable cost, personnel cost, for single car, be substantially it is a kind of with odd-numbered day traveling away from
Numerical value from the variation of, linearly, therefore can convert to (k3·tTraveling+k4·DTraveling) in, remainder residual error can be understood as
Influence of remaining external attribute to target effectiveness is picked.
In step 103, according to the initial demand of simulation generation and parameter is picked, with reasonable objective function (i.e. above-mentioned public affairs
Formula one) calculate the target value of utility after each demand is satisfied.The value of utility can be understood as the production of cost caused by picking every time
Out, directly using cost as optimization aim and unreasonable but since optimal pick can still generate cost, cope with value of utility
Carry out function conversion.Since the cost value of running time and waiting time are continuous variations with demand trip distance on vehicle
And change, therefore by UCmin/ UC is as optimization target values, UCminIndicate the target of the shortest path N-free diet method time of the demand at
This, by the optimization target values be defined as it is each pick evaluation point, it is higher to pick evaluation point, represents the efficiency that this time is picked and gets over
It is high.
Step 104 will order vehicle demand properties, vehicle attribute, running environment attribute and pick evaluation and point be put into artificial intelligence
It can be trained in model;Wherein, vehicle demand properties, vehicle attribute and the running environment attribute ordered is independent variable, is connect
Evaluation is sent to be divided into dependent variable.
It in optionally a kind of embodiment, is specifically included in step 104: ordering vehicle demand for every, select Pi<Pmax
And Di≤DmaxVehicle calculated in the artificial intelligence model pick evaluation point;Wherein, PiFor vehicle carrying number, PmaxFor
Vehicle maximum carrying number, DiIndicate the distance of the next destination distance requirement point of vehicle i, DmaxExpression system maximum picks appearance
Difference.
In optionally a kind of embodiment, if meeting Pi<PmaxAnd Di≤DmaxVehicle number be less than preset threshold, then
Range of tolerable variance increases a certain range, until the vehicle number for meeting above-mentioned condition is greater than minimum standard number.In specific implementation, it presets
Threshold value is 1~5, such as 3.
In optionally a kind of embodiment, ordering vehicle demand properties mainly has the terminus for ordering vehicle and orders the vehicle time, will
The terminus position data of demand are converted into the classified variable based on grid cell, order vehicle time decomposition be moment and date,
Wherein, the date is converted to the classified variables such as working day, day off, festivals or holidays, and the moment is converted to digital continuous variable.
In optionally a kind of embodiment, vehicle when vehicle attribute includes type of vehicle (BusType), is connected to demand
Position coordinates (BusPosition), vehicle are according to next destination distance (Dis2NextStation), the next purpose of vehicle
The stroke distances that ground is not completed according to having passenger on demand point distance (DisOfNext2OStation), vehicle
(RemainDistance), have that passenger do not complete on vehicle around row distance (OldBypassDistance) if, carry out it is current
This is picked the increased ridership (PassengersOnBus) in row distance (IncreasedDistance), vehicle, expects
Number of stops (PredictNumOfStops).
In a kind of optional embodiment, running environment attribute include be not limited to weather condition, congestion in road situation,
The external factor such as air pollution situation.
In optionally a kind of embodiment, specifically includes in step 104: being placed into after being pre-processed to independent variable
It is trained in the artificial intelligence model.In the present embodiment, independent variable attribute is pre-processed, continuous variable is advised
One changes, and by distribution conversion to -1~1, and according to certain priori knowledge, carries out Regularization to part attribute, reduces model mistake
The probability of fitting.
In optionally a kind of embodiment, the artificial intelligence model is instructed using neural network in step 104
Practice.Pretreated independent variable attribute is picked to evaluate to divide to be put into artificial intelligence model with dependent variable target and is trained, by
It is one in its essence and learns by oneself the regression problem responded, therefore neural network is selected to be trained it.Neural network model choosing
It selects multilayer perceptron (MLP) to be calculated, since it is with nonlinear perception layer, more complicated functional relation can be fitted,
The hiding number of plies is defined as two layers, and neuronal quantity is calculated automatically by algorithm.In order to preferably carry out forecast analysis to target variable,
With Boosting Fusion Model, the neural network model of 10 different weights is constructed in the algorithm, and is merged, and is trained
Journey is sequential build, the parameter assignment when confidence level of the predicted target values of last round of neural network is as next round training, most
The model merged eventually has stronger predictive ability.
Step 105 picks evaluation point according to what training prediction obtained, and selection is picked the optimal vehicle of evaluation point and connect
It send;Selection is picked the maximum vehicle of evaluation point and is picked in a specific example, i.e. Max { δ1,δ2,δ3…δn}。
In optionally a kind of embodiment, as shown in Fig. 2, further comprising the steps of after step 105:
If step 106, system operation time are greater than optimization cycle of training, evaluation is picked actual operation data and actually
Divide to be put into the artificial intelligence model and be trained.
After meeting and picking demand every time, system is by actual operation data (the practical Waiting time, reality of this demand
Journey time, traveled distance distance, it is practical around row distance, practical stops, actually pick evaluation point) save to data storage
End.
Since training uses simulation demand and STATIC SIMULATION to pick model to model for the first time, it is difficult to accurately to practical feelings
Condition is simulated, thus needs to continue to optimize training to scheduling model.The present embodiment uses artificial intelligence technology and carries out target
Function prediction, therefore can realize the adaptive adjustment of scheduling optimization model in actual operations, system can in operation cumulative number
According to continuous iterative learning carries out more accurate prediction to objective function, realizes efficient intelligent dispatch.
Optimization training is divided into two steps, and system can optimize cycle of training according to the efficiency evaluation of actual operation first, specifically,
Optimize the calculation formula of cycle of training in step 106 are as follows:
Wherein, m is minimum optimization cycle of training;UCIt is practical comprehensiveFor the overall cost of actual schedule;UCI predictionFor vehicle i simulation
The overall cost of scheduling;α is constant term;N is user demand total amount in region.
Step 107, training after the completion of by actual demand data import simulation system in, using Optimized model as pick evaluate
Divide calculation basis, operation simulation, and the overall cost of calculating simulation scheduling are carried out to actual demand.
Step 108, the overall cost by the overall cost of the operation simulation in actual schedule compare, if described
Occur being higher than the Arbitrary Term of the overall cost of the actual schedule in the overall cost of operation simulation, then overall cost is calculated public
Weight term in formula is modified.
It further include picking evaluation point to artificial intelligence model re -training with revised, passing through gradient after step 108
Descent method is iterated calculating, until function convergence.Hereby it is achieved that scheduling model is to the trip requirements generated after actual operation
Adaptive fine tuning.For overall model since training is optimized with real data, model automatic Fitting constantly approaches optimization mesh
Mark corrects the trip data self-teaching of each direction, point, is finally completed the intelligent high-efficiency scheduling for tending to optimal solution.
In a specific embodiment, the dispatching method of demand response formula public transit system is as shown in figure 3, according to being initial
Change or actual motion executes different steps, if initialization, then needs to set weight term and vehicle demand and fortune are ordered in emulation
Row environment;If actual motion, then read and actually order vehicle demand and running environment, and according to whether reach learning cycle, i.e., on
The optimization cycle of training in step 106 is stated, the optimization of weight term and other parameters is carried out, to realize the excellent of artificial intelligence model
Change, the final intelligent high-efficiency scheduling for realizing demand response formula public transit system.
The embodiment of the present invention introduces artificial intelligence technology, constructs passenger demand according to the data for meeting true trip characteristics,
It is dispatched using simulation software simulating vehicle;Integrated objective function is constructed, the evaluation gone on a journey every time is determined according to routine dispactching algorithm
Point, the value of utility of different vehicle scheduling is predicted in dynamic context, and dispatches buses by optimum prediction value, is generated practical
Operation data.By the data accumulation in one section of period, pass through artificial intelligence technology, pass of the building evaluation point between route choosing
System, and constantly study optimization, finally realize efficient self study intelligent scheduling model, to realize intelligentized dynamic line
Planning.
The embodiment of the present invention also provides a kind of scheduling system of demand response formula public transit system based on artificial intelligence, packet
It includes:
Memory, for storing computer program;
And processor, for executing the computer program, to realize the method such as above-described embodiment.
Although specific embodiments of the present invention have been described above, it will be appreciated by those of skill in the art that these
It is merely illustrative of, protection scope of the present invention is defined by the appended claims.Those skilled in the art is not carrying on the back
Under the premise of from the principle and substance of the present invention, many changes and modifications may be made, but these are changed
Protection scope of the present invention is each fallen with modification.
Claims (10)
1. a kind of dispatching method of the demand response formula public transit system based on artificial intelligence, which comprises the following steps:
S1, starting point, the point of arrival and the requirements set at demand moment are generated;
S2, initial plant scheduling scheme is constructed according to the requirements set;
S3, according to initial plant scheduling scheme calculate it is all order vehicle demand pick evaluation point;
S4, will order vehicle demand properties, vehicle attribute, running environment attribute and pick evaluation point be put into artificial intelligence model into
Row training;Wherein, ordering vehicle demand properties, vehicle attribute and running environment attribute is independent variable, picks evaluation and is divided into because becoming
Amount;
S5, evaluation point is picked according to what training prediction obtained, selection is picked the optimal vehicle of evaluation point and picked.
2. the method as described in claim 1, which is characterized in that specifically included in the step S2:
In the case where meeting default constraint condition, the vehicle for selecting distance requirement point nearest is picked.
3. method according to claim 2, which is characterized in that the constraint condition includes at least:Pi<
Pmax、Ti≤Tmax、Si≤Smax;
Wherein, LxyIt is stroke xy around row distance, lxyFor the shortest distance of stroke xy, γmaxTo constrain maximum Multi-feed-eye ring;Pi
For vehicle carrying number, PmaxFor vehicle maximum carrying number;TiFor Waiting time, TmaxTo constrain maximum Waiting time;SiIt is secondary to stop
Number, SmaxTo constrain most stop frequencies.
4. the method as described in claim 1, which is characterized in that specifically included in the step S3:
All overall cost UC for ordering vehicle demand are calculated according to initial plant scheduling schemeIt is comprehensive:
Wherein, Q is response type public transit vehicle sum in coverage;N is user demand total amount in region;T is the time;D be away from
From;k1~k4For weight term;UCIt is comprehensiveUnit be value of utility, i.e., each index item is converted into the value after Costco Wholesale;θ is not
Other costs of prediction;
All evaluations of picking for ordering vehicle demand, which are calculated, according to the overall cost divides δ=UCmin/UC;
Wherein, UC indicates every calculating cost for ordering vehicle demand, UCminIndicate that this orders the shortest path N-free diet method time of vehicle demand
Objective cost;
UCi=k2×tIt waits+k3×tIt is interior+k4×DTraveling+k5×β;
Wherein, k5β is the residual error for picking value of utility every time.
5. the method as described in claim 1, which is characterized in that specifically included in the step S4: vehicle demand is ordered for every,
Select Pi<PmaxAnd Di≤DmaxVehicle calculated in the artificial intelligence model pick evaluation point;Wherein, PiFor vehicle load
Objective number, PmaxFor vehicle maximum carrying number, DiIndicate the distance of vehicle i distance requirement point, DmaxExpression system maximum picks tolerance.
6. method as claimed in claim 5, which is characterized in that the method also includes: if meeting Pi<PmaxAnd Di≤Dmax
Vehicle number be less than preset threshold, then range of tolerable variance increase, and range of tolerable variance be not more than demand response formula public transit system service
Region, until the vehicle number for meeting above-mentioned condition is greater than minimum standard number.
7. the method as described in claim 1, which is characterized in that specifically included in the step S4: being located in advance to independent variable
It places into the artificial intelligence model and is trained after reason.
8. the method as described in claim 1, which is characterized in that further comprising the steps of after the step S5:
If S6, system operation time are greater than optimization cycle of training, actual operation data are put into institute with evaluation point is actually picked
It states in artificial intelligence model and is trained;
S7, training after the completion of by actual demand data import simulation system in, using Optimized model as pick evaluation point calculating according to
According to, to actual demand carry out operation simulation, and calculating simulation scheduling overall cost;
S8, the overall cost by the overall cost of the operation simulation in actual schedule compare, if in the operation simulation
Overall cost in occur be higher than the actual schedule overall cost Arbitrary Term, then to the power in overall cost calculation formula
Weight item is modified.
9. method according to claim 8, which is characterized in that optimize the calculation formula of cycle of training in the step S6 are as follows:
Wherein, m is minimum optimization cycle of training;UCIt is practical comprehensiveFor the overall cost of actual schedule;UCI predictionFor vehicle i operation simulation
Overall cost;α is constant term;N is user demand total amount in region.
10. a kind of scheduling system of the demand response formula public transit system based on artificial intelligence characterized by comprising
Memory, for storing the data of computer program and operation generation;
And processor, for executing the computer program, to realize method as claimed in any one of claims 1-9 wherein.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910741446.2A CN110458456B (en) | 2019-08-12 | 2019-08-12 | Demand response type public transportation system scheduling method and system based on artificial intelligence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910741446.2A CN110458456B (en) | 2019-08-12 | 2019-08-12 | Demand response type public transportation system scheduling method and system based on artificial intelligence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110458456A true CN110458456A (en) | 2019-11-15 |
CN110458456B CN110458456B (en) | 2023-01-31 |
Family
ID=68486048
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910741446.2A Active CN110458456B (en) | 2019-08-12 | 2019-08-12 | Demand response type public transportation system scheduling method and system based on artificial intelligence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110458456B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111105141A (en) * | 2019-11-23 | 2020-05-05 | 同济大学 | Demand response type bus dispatching method |
CN111507502A (en) * | 2020-02-21 | 2020-08-07 | 海南太美航空股份有限公司 | Virtual flight number optimization method and system |
CN112862262A (en) * | 2021-01-18 | 2021-05-28 | 厦门金龙联合汽车工业有限公司 | Park unmanned vehicle scheduling method, terminal equipment and storage medium |
CN113160602A (en) * | 2021-04-27 | 2021-07-23 | 华录智达科技股份有限公司 | Intelligent bus dispatching simulation system |
CN113406951A (en) * | 2020-03-16 | 2021-09-17 | 株式会社东芝 | Travel control device, travel control method, and computer program |
CN113486136A (en) * | 2021-08-04 | 2021-10-08 | 泰瑞数创科技(北京)有限公司 | Method and system for assembling geographic entity service on demand |
CN115203982A (en) * | 2022-09-14 | 2022-10-18 | 环球数科集团有限公司 | Parallel computing method and simulation system for intelligent operation of public transport vehicle |
CN116929781A (en) * | 2023-06-12 | 2023-10-24 | 广州汽车集团股份有限公司 | Vehicle evaluation method, cloud platform, vehicle and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5168451A (en) * | 1987-10-21 | 1992-12-01 | Bolger John G | User responsive transit system |
US5799263A (en) * | 1996-04-15 | 1998-08-25 | Bct Systems | Public transit system and apparatus and method for dispatching public transit vehicles |
JP2014032531A (en) * | 2012-08-03 | 2014-02-20 | Toyota Motor Corp | Operation plan system of car sharing system, and operation plan method |
CN105070044A (en) * | 2015-08-17 | 2015-11-18 | 南通大学 | Dynamic scheduling method for customized buses and car pooling based on passenger appointments |
CN109308574A (en) * | 2018-09-07 | 2019-02-05 | 华南理工大学 | A kind of flexible bus dispatching method in internet of real-time response half |
US20190108468A1 (en) * | 2017-10-10 | 2019-04-11 | Khanh Vinh Nguyen | Method and apparatus to operate smart mass transit systems with on-demand rides, dynamic routes and coordinated transfers |
-
2019
- 2019-08-12 CN CN201910741446.2A patent/CN110458456B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5168451A (en) * | 1987-10-21 | 1992-12-01 | Bolger John G | User responsive transit system |
US5799263A (en) * | 1996-04-15 | 1998-08-25 | Bct Systems | Public transit system and apparatus and method for dispatching public transit vehicles |
JP2014032531A (en) * | 2012-08-03 | 2014-02-20 | Toyota Motor Corp | Operation plan system of car sharing system, and operation plan method |
CN105070044A (en) * | 2015-08-17 | 2015-11-18 | 南通大学 | Dynamic scheduling method for customized buses and car pooling based on passenger appointments |
US20190108468A1 (en) * | 2017-10-10 | 2019-04-11 | Khanh Vinh Nguyen | Method and apparatus to operate smart mass transit systems with on-demand rides, dynamic routes and coordinated transfers |
CN109308574A (en) * | 2018-09-07 | 2019-02-05 | 华南理工大学 | A kind of flexible bus dispatching method in internet of real-time response half |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111105141A (en) * | 2019-11-23 | 2020-05-05 | 同济大学 | Demand response type bus dispatching method |
CN111507502A (en) * | 2020-02-21 | 2020-08-07 | 海南太美航空股份有限公司 | Virtual flight number optimization method and system |
CN113406951A (en) * | 2020-03-16 | 2021-09-17 | 株式会社东芝 | Travel control device, travel control method, and computer program |
CN112862262A (en) * | 2021-01-18 | 2021-05-28 | 厦门金龙联合汽车工业有限公司 | Park unmanned vehicle scheduling method, terminal equipment and storage medium |
CN112862262B (en) * | 2021-01-18 | 2022-09-30 | 厦门金龙联合汽车工业有限公司 | Park unmanned vehicle scheduling method, terminal equipment and storage medium |
CN113160602A (en) * | 2021-04-27 | 2021-07-23 | 华录智达科技股份有限公司 | Intelligent bus dispatching simulation system |
CN113160602B (en) * | 2021-04-27 | 2022-05-10 | 华录智达科技股份有限公司 | Intelligent bus dispatching simulation system |
CN113486136A (en) * | 2021-08-04 | 2021-10-08 | 泰瑞数创科技(北京)有限公司 | Method and system for assembling geographic entity service on demand |
CN115203982A (en) * | 2022-09-14 | 2022-10-18 | 环球数科集团有限公司 | Parallel computing method and simulation system for intelligent operation of public transport vehicle |
CN116929781A (en) * | 2023-06-12 | 2023-10-24 | 广州汽车集团股份有限公司 | Vehicle evaluation method, cloud platform, vehicle and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110458456B (en) | 2023-01-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110458456A (en) | The dispatching method and system of demand response formula public transit system based on artificial intelligence | |
WO2021248607A1 (en) | Deep reinforcement learning-based taxi dispatching method and system | |
CN110472764B (en) | Method and system for coordinating multiparty services using semi-collaborative Nash balancing based on reinforcement learning | |
Chen et al. | Evolutionary computation for intelligent transportation in smart cities: A survey | |
Li et al. | Feeder transit services: Choosing between fixed and demand responsive policy | |
CN109034566B (en) | A kind of intelligent dispatching method and device based on passenger flow above and below bus station | |
CN101950479B (en) | Passenger travel-oriented intelligent urban public transport system and implementation method thereof | |
Zhang et al. | The sightseeing bus schedule optimization under park and ride system in tourist attractions | |
CN110135755A (en) | A kind of method of complex optimum section urban public transit timetable establishment and vehicle scheduling | |
CN103177575A (en) | System and method for dynamically optimizing online dispatching of urban taxies | |
CN107330716A (en) | A kind of customization public transport pricing method for considering system optimal | |
CN112085249B (en) | Customized bus route planning method based on reinforcement learning | |
CN113642758B (en) | Supply and demand prediction and scheduling method and device for shared electric automobile | |
CN110598971B (en) | Responsive bus service planning method based on ant colony algorithm | |
CN116324838A (en) | System and method for scheduling shared rides through a taxi calling platform | |
CN112700029A (en) | Customized bus planning method based on simulation optimization framework | |
CN104477217A (en) | Operation and circulation of trains on city rail with single end station | |
CN113947287A (en) | Electric bus regional dispatching method and system | |
Tang et al. | Dynamic operations of an integrated mobility service system of fixed-route transits and flexible electric buses | |
Gonzales et al. | Optimizing ADA Paratransit Operation with Taxis and Ride Share Programs | |
Robèrt | Backcasting and econometrics for sustainable planning: information technology and individual preferences of travel | |
Wei et al. | Optimal routing design of feeder transit with stop selection using aggregated cell phone data and open source GIS tool | |
Han et al. | Two-stage optimization of airport ferry service delay considering flight uncertainty | |
Pratelli et al. | Comparing route deviation bus operation with respect to dial-a-ride service for a low-demand residential area | |
CN115936330A (en) | Demand response type BRT vehicle scheduling algorithm based on Lagrangian principle |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |