CN101739655A - Method for scheduling public slow system dynamically based on rolling horizon scheduling algorithm - Google Patents

Abstract

The invention discloses a method for scheduling a public slow system dynamically based on a rolling horizon scheduling algorithm, which comprises the following steps: 1) setting a time axis for expressing an overall scheduling cycle of a working day, wherein a key point is a leasing point which is receiving a service or the leasing point to which a transport vehicle is on the way; a time window is a time range in which a public bicycle leasing point is allowed to serve; and the satisfaction degree of the leasing point i is expressed as fi(ti); 2) establishing a scheduling model; and 3) adopting a rolling horizon-based dynamic scheduling method to divide the overall scheduling cycle into a plurality of sub-scheduling cycles; considering each sub-scheduling cycle as a static scheduling problem; and solving a plurality of static scheduling problems through a variable neighborhood tabu search algorithm and realizing dynamic vehicle scheduling through a method for rolling and updating a scheduling plan discontinuously. The method can perform automation statistic, has good real time and improves operating efficiency.

Description

A kind of public system dynamics dispatching method of going slowly based on rolling time domain dispatching algorithm

Technical field

The present invention relates to a kind of public system dynamics dispatching method of going slowly.

Background technology

At present, also do not have the dispatching method of the public system of going slowly, only have the dispatching method of (as postal problem, tobacco dispensing problem, traveling salesman problem, logistics distribution problem etc.) in other the near field.The classification of vehicle dispatching problem:

According to the requirement of site to service time, vehicle dispatching problem is divided into: no time window vehicle dispatching problem and free window vehicle dispatching problem.The site did not require the time that begins service (loading or unloading) in no time window vehicle dispatching problem; Otherwise the scheduling problem that requires the site to accept to serve within a certain period of time is free window vehicle dispatching problem.Wherein according to the difference that required service time, free window vehicle dispatching problem is divided into hard time window vehicle dispatching problem and soft time window scheduling problem again, in hard time window vehicle dispatching problem, vehicle must be finished in given time window the service of site, can not shift to an earlier date and postpone; In soft time window vehicle dispatching problem, the site is not to there being very strict requirement service time, vehicle should be as far as possible arrives in given time window and serves, but allow to a certain degree in advance or postpone, and according in advance or the degree that postpones this service is punished.

Whether identical according to type of vehicle, vehicle dispatching problem is divided into: single-type vehicle dispatching problem and multi-vehicle-type vehicle dispatching problem.

According to the number in parking lot or warehouse, vehicle dispatching problem can be divided into: bicycle field vehicle dispatching problem and many depot Vehicle Scheduling Problem.

Whether known fully according to information before the scheduling, vehicle dispatching problem is divided into static vehicle dispatching problem and dynamic vehicle scheduling problem.Static vehicle dispatching problem is meant that all information all is known before formulating operation plan, comprises the service request information of information of vehicles, demand point, traffic information of road network or the like, and these information are to remain unchanged in the process of scheduling; The dynamic vehicle scheduling problem is meant that it is known having only the information of part before formulating operation plan, and some information are arranged is constantly to change in the process of scheduling, such as traffic, vehicle-state and the weather condition etc. of the position of demand point and demand, road network.

Whether must return the parking lot after finishing the work according to vehicle, vehicle dispatching problem can be divided into open scheduling problem of vehicle and vehicle closure scheduling problem.Wherein the open scheduling problem of vehicle is meant after vehicle is finished operation plan and needn't returns the parking lot, and must return the parking lot after vehicle closure scheduling problem vehicle is finished operation plan.

According to the feature of dispensing goods, vehicle dispatching problem is divided into pure deliver goods problem, purely gets the goods problem and fetch and deliver goods mixed problem.

The existing public system of going slowly, for example public bicycles need carry out complicate statistics to the number of vehicles of each site usually, and regularly carries out the scheduling of vehicle, and real-time is poor, lacks automatic dispatching system, has influenced the operational efficiency of the public system of going slowly.

Summary of the invention

Poor for the complicate statistics, the real-time that overcome the existing public system of going slowly, as to influence operational efficiency deficiency, the invention provides a kind ofly can add up automatically, real-time is good, promote the public system dynamics dispatching method of going slowly based on rolling time domain dispatching algorithm of operational efficiency.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of public system dynamics dispatching method of going slowly based on rolling time domain dispatching algorithm may further comprise the steps:

1), the setting-up time axle is represented a workaday whole dispatching cycle, on time shaft, each constantly corresponding scene, key point are meant to be accepted to serve or has haulage vehicle going to lease point on the road of this point, and the task of key point is to change;

Time window is the time range that public bicycles lease point allows service, describes a lease restriction range to service time, [WA with the Fuzzy Time window _i, WB _i] expression lease point i tolerable service time of scope, [WC _i, WD _i] scope service time of expression lease point i expectation, t _iThe time of expression vehicle point of arrival i; The satisfaction of lease point i can be expressed as:

2), set up scheduling model: parking lot P0 has T transport vehicle, and haulage vehicle k dead weight is Qk, sends some haulage vehicles to provide service to n public bicycles lease point from parking lot P0, and the capacity of each lease point is E _i(i=1,2 ... n), at t constantly, the quantity of the bicycle of each lease point is q _i(t) (i=1,2 ... n).q _i(t) and E _iRatio be H _i(t), the H of lease point i under the normal condition _iScope be [C _Min, C _Max], the services request of lease point has two kinds of situations: 1. work as H _i(t)＞C _MaxThe time, expression lease point need transport bicycle constantly at t; 2. work as H _i(t)＜C _MinThe time, expression lease point need be transported into bicycle constantly at t; Lease point i tolerable service time of scope is [WA _i, WB _i], scope service time of expectation is [WC _i, WD _i]; Be s the service time of lease point i _i, t _iThe expression haulage vehicle arrives the time that lease point i begins to serve, f _i(t _i) satisfaction of expression lease point i;

The demand of lease point is worked as H for the quantity of a lease bicycle that need be transported into or transport _i(t)＞C _MaxThe time, the lease point need transport bicycle, and establishing lease point demand at this moment is d _i(t), after the lease point is accepted service, its H _iShould be at interval [C _Min, C _Max] in, then must satisfy $C_{\min }\≤\frac{q_i\left(t\right)-d_i\left(t\right)}{E_i}\≤C_{\max },$ Be q _i(t)-C _MaxE _i≤ d _i(t)≤q _i(t)-C _MinE _i, at this moment, d _i(t) get its lower limit q _i(t)-C _MaxE _iCan meet the demands; In like manner, work as H _i(t)＜C _MinThe time, d _i(t)=C _MinE _i-q _i(t);

K: the set of haulage vehicle;

TS: time shaft (and being divided into several time periods);

T _n: n the time period on time shaft;

N (T _n): at T _nIn time period, all lease points of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

M (T _n): at T _nIn time period, all key points, the lease point of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

Q _Ki: the quantity of the bicycle that haulage vehicle K loads in lease point i service back;

Turn to the model that target is set up the public system call of going slowly with the total satisfaction maximum of lease point, objective function is as follows:

$\max Z=\underset{T_n\∈\mathrm{TS}}{\mathrm{\Σ}}\underset{i\∈N\left(T_n\right)}{\overset{t_i\∈T_n}{\mathrm{\Σ}}}\left[f_i\left(t_i\right)\right]---\left(1\right)$

Constraint condition is:

Q _ki＜Q _k $\∀i\∈M\left(t\right),k\∈K---\left(2\right)$

3), will be divided into experimental process dispatching cycle whole dispatching cycle, will regard static scheduling problem each sub-dispatching cycle as, by finding the solution several static scheduling problems and the method for upgrading operation plan of constantly rolling realizes dynamic vehicle scheduling.When finding the solution the static scheduling problem of every sub-dispatching cycle of correspondence, adopt based on the tabu search algorithm that becomes neighborhood and obtain operation plan.In the search procedure that becomes neighborhood, there is following notion:

(a) become the neighborhood search condition: when the adaptation value of separating after the continuous several times iteration in the search procedure does not improve, satisfy the neighborhood search condition that becomes, otherwise, do not satisfy the neighborhood search jumping condition that becomes;

(b) former neighborhood is separated: the neighborhood that does not satisfy when becoming the neighborhood search condition in the tabu search process is separated;

(c) becoming neighborhood separates: the neighborhood that satisfies when becoming the neighborhood search condition in the tabu search process is separated, and becomes neighborhood and separates to separate to compare with former neighborhood and have different neighbour structures.

The tabu search algorithm step that becomes neighborhood is as follows, comprising:

(3.1) produce initial solution, initiation parameter and taboo table;

(3.2) judge whether to satisfy end condition, if satisfy then introduce and export the result, otherwise, go to step (3.3);

(3.3) judge whether to satisfy the neighborhood condition that becomes, separate if satisfy the current change neighborhood of separating of generation, otherwise, generate the current former neighborhood of separating and separate;

(3.4) determine candidate solution;

(3.5) judge whether the satisfied criterion of despising, if, then replace current separating with specially pardoning to separate, go to step (3.2) and renewal and avoid showing, otherwise, then go to step (3.6);

(3.6) in candidate solution, separate current the separating of replacement, and upgrade the taboo table, go to step (3.2) with optimum non-taboo.

Further, in described step 3), change neighbour structure and can increase the probability that tabu search algorithm obtains globally optimal solution greatly, when the adaptation value of separating of continuous several times iteration acquisition does not improve, then the tabu search process might be absorbed in the local optimum trap, and jump out the quality that the local optimum point separates raising this moment is very important.Therefore, adopt following measure to improve the quality of tabu search algorithm: when not improving, then to change neighbour structure, promptly in the current new neighborhood of separating is separated, search for through the adaptation value of separating after the continuous several times iteration.

Beneficial effect of the present invention mainly shows: can add up automatically, real-time is good, promote operational efficiency.

Description of drawings

Fig. 1 is the synoptic diagram of Fuzzy Time window.

Fig. 2 is based on the process flow diagram of the public system dynamics scheduling of going slowly of rolling time domain dispatching algorithm.

Fig. 3 is the process flow diagram that becomes the taboo search method of neighborhood.

Embodiment

Below in conjunction with accompanying drawing the present invention is further described.

With reference to Fig. 1～Fig. 3, a kind of public system dynamics dispatching method of going slowly based on rolling time domain dispatching algorithm may further comprise the steps:

1), the setting-up time axle is represented a workaday whole dispatching cycle, on time shaft, each constantly corresponding scene, key point are meant to be accepted to serve or has haulage vehicle going to lease point on the road of this point, and the task of key point is to change;

Time window is the time range that public bicycles lease point allows service, describes a lease restriction range to service time, [WA with the Fuzzy Time window _i, WB _i] expression lease point i tolerable service time of scope, [WC _i, WD _i] scope service time of expression lease point i expectation, t _iThe time of expression vehicle point of arrival i; The satisfaction of lease point i can be expressed as:

2), set up scheduling model: parking lot P0 has T transport vehicle, and haulage vehicle k dead weight is Q _k, send some haulage vehicles to provide service from parking lot P0 to n public bicycles lease point, the capacity of each lease point is E _i(i=1,2 ... n), at t constantly, the quantity of the bicycle of each lease point is q _i(t) (i=1,2 ... n).q _i(t) and E _iRatio be H _i(t), the H of lease point i under the normal condition _iScope be [C _Min, C _Max], the services request of lease point has two kinds of situations: 1. work as H _i(t)＞C _MaxThe time, expression lease point need transport bicycle constantly at t; 2. work as H _i(t)＜C _MinThe time, expression lease point need be transported into bicycle constantly at t; Lease point i tolerable service time of scope is [WA _i, WB _i], scope service time of expectation is [WC _i, WD _i]; Be s the service time of lease point i _i, t _iThe expression haulage vehicle arrives the time that lease point i begins to serve, f _i(t _i) satisfaction of expression lease point i;

The demand of lease point is worked as H for the quantity of a lease bicycle that need be transported into or transport _i(t)＞C _MaxThe time, the lease point need transport bicycle, and establishing lease point demand at this moment is d _i(t), after the lease point is accepted service, its H _iShould be at interval [C _Min, C _Max] in, then must satisfy $C_{\min }\≤\frac{q_i\left(t\right)-d_i\left(t\right)}{E_i}\≤C_{\max },$ Be q _i(t)-C _MaxE _i≤ d _i(t)≤q _i(t)-C _MinE _i, at this moment, d _i(t) get its lower limit q _i(t)-C _AmxE _iCan meet the demands; In like manner, work as H _i(t)＜C _MinThe time, d _i(t)=C _MinE _i-q _i(t);

K: the set of haulage vehicle;

TS: time shaft (and being divided into several time periods);

T _n: n the time period on time shaft;

N (T _n): at T _nIn time period, all lease points of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

M (T _n): at T _nIn time period, all key points, the lease point of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

Q _Ki: the quantity of the bicycle that haulage vehicle K loads in lease point i service back;

Turn to the model that target is set up the public system call of going slowly with the total satisfaction maximum of lease point, objective function is as follows:

$\max Z=\underset{T_n\∈\mathrm{TS}}{\mathrm{\Σ}}\underset{i\∈N\left(T_n\right)}{\overset{t_i\∈T_n}{\mathrm{\Σ}}}\left[f_i\left(t_i\right)\right]---\left(1\right)$

Constraint condition is:

Q _ki＜Q _k $\∀i\∈M\left(t\right),k\∈K---\left(2\right)$ 3)、

To be divided into experimental process dispatching cycle whole dispatching cycle, will regard static scheduling problem each sub-dispatching cycle as, by finding the solution several static scheduling problems and the method for upgrading operation plan of constantly rolling realizes dynamic vehicle scheduling.When finding the solution the static scheduling problem of every sub-dispatching cycle of correspondence, adopt based on the tabu search algorithm that becomes neighborhood and obtain operation plan.In the search procedure that becomes neighborhood, there is following notion:

(a) become the neighborhood search condition: when the adaptation value of separating after the continuous several times iteration in the search procedure does not improve, satisfy the neighborhood search condition that becomes, otherwise, do not satisfy the neighborhood search jumping condition that becomes;

(b) former neighborhood is separated: the neighborhood that does not satisfy when becoming the neighborhood search condition in the tabu search process is separated;

(c) becoming neighborhood separates: the neighborhood that satisfies when becoming the neighborhood search condition in the tabu search process is separated, and becomes neighborhood and separates to separate to compare with former neighborhood and have different neighbour structures.

The tabu search algorithm step that becomes neighborhood is as follows, comprising:

(3.1) produce initial solution, initiation parameter and taboo table;

(3.2) judge whether to satisfy end condition, if satisfy then introduce and export the result, otherwise, go to step (3.3);

(3.3) judge whether to satisfy the neighborhood condition that becomes, separate if satisfy the current change neighborhood of separating of generation, otherwise, generate the current former neighborhood of separating and separate;

(3.4) determine candidate solution;

(3.5) judge whether the satisfied criterion of despising, if, then replace current separating with specially pardoning to separate, go to step (3.2) and renewal and avoid showing, otherwise, then go to step (3.6);

(3.6) in candidate solution, separate current the separating of replacement, and upgrade the taboo table, go to step (3.2) with optimum non-taboo.

Further, in described step 3), change neighbour structure and can increase the probability that tabu search algorithm obtains globally optimal solution greatly, when the adaptation value of separating of continuous several times iteration acquisition does not improve, then the tabu search process might be absorbed in the local optimum trap, and jump out the quality that the local optimum point separates raising this moment is very important.Therefore, adopt following measure to improve the quality of tabu search algorithm: when not improving, then to change neighbour structure, promptly in the current new neighborhood of separating is separated, search for through the adaptation value of separating after the continuous several times iteration.

The process of setting up of the public system call model of going slowly of present embodiment is:

Key point and time shaft: at first introduce the notion of key point and time shaft, as shown in Figure 1, time shaft is represented a workaday whole dispatching cycle, on time shaft, and each constantly corresponding scene, the t in Fig. 1 ₁Constantly corresponding scene is service object not as yet for lease point 1,3,4 and 6, and lease point 2 is to have finished service object, and lease 5 is for carrying out service object; t ₂Constantly corresponding scene is service object not as yet for lease point 1,4, and lease point the 2,5, the 6th has been finished service object, and lease point 3 is for carrying out service object, and lease point 7 is the lease point that services request is arranged that increases newly.Key point is meant to be accepted to serve or has haulage vehicle going to lease point on the road of this point, and the task of key point can not be changed, as t among Fig. 1 ₁Corresponding constantly key point is a lease point 3 and 5, t ₂Corresponding constantly key point is a lease point 1 and 3.

The satisfaction of Fuzzy Time window and public bicycles lease point: time window is the time range that public bicycles lease point allows service, describes a lease restriction range to service time with the Fuzzy Time window, as shown in the figure, and [WA _i, WB _i] expression lease point i tolerable service time of scope, [WC _i, WD _i] scope service time [9] of expression lease point i expectation, t _iThe time of expression vehicle point of arrival i.

Then the satisfaction of lease point i can be expressed as:

The foundation of the public system call model of going slowly: the public system call problem of going slowly can be described as: parking lot P0 has T transport vehicle, and haulage vehicle k dead weight (loading the quantity of bicycle at most) is Q _k, send some haulage vehicles to provide service to n public bicycles lease point from parking lot P0.The capacity (quantity of the bicycle that holds at most) of each lease point is E _i(i=1,2 ... n), at t constantly, the quantity of the bicycle of each lease point is q _i(t) (i=1,2 ... n).q _i(t) and E _iRatio be H _i(t), the H of lease point i under the normal condition _iScope be [C _Min, C _Max], the services request of lease point has two kinds of situations: (1) works as H _i(t)＞C _MaxThe time, expression lease point need transport bicycle constantly at t; (2) work as H _i(t)＜C _MinThe time, expression lease point need be transported into bicycle constantly at t.Lease point i tolerable service time of scope is [WA _i, WB _i], scope service time of expectation is [WC _i, WD _i].Be s the service time of lease point i _i, t _iThe expression haulage vehicle arrives the time that lease point i begins to serve, f _i(t _i) satisfaction of expression lease point i.Require to arrange rational route or travel by vehicle to make the satisfaction maximization of lease point.And meet the following conditions: at any time, the dead weight capacity of each transport vehicle can not surpass the dead weight of vehicle.

Determining of public bicycles lease point demand: the quantity of the bicycle that the demand of lease point need be transported into or transport for the lease point, work as H _i(t)＞C _MaxThe time, the lease point need transport bicycle, and establishing lease point demand at this moment is d _i(t), after the lease point is accepted service, its H _iShould be at interval [C _Min, C _Max] in, then must satisfy $C_{\min }\≤\frac{q_i\left(t\right)-d_i\left(t\right)}{E_i}\≤C_{\max },$ Be q _i(t)-C _MaxE _i≤ d _i(t)≤q _i(t)-C _MinE _i, at this moment, d _i(t) get its lower limit q _i(t)-C _MaxE _iCan meet the demands; In like manner, work as H _i(t)＜C _MinThe time, d _i(t)=C _MinE _i-q _i(t).

The assumed condition of model: the running time between (1) lease point is known; (4) haulage vehicle does not have the restriction of running time and distance travelled; (5) can predict each lease point solicited message in 10 minutes.

The implication of parameter is as follows in the model:

K: the set of haulage vehicle;

TS: time shaft (and being divided into several time periods);

T _n: n the time period on time shaft;

N (T _n): at T _nIn time period, all lease points of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

M (T _n): at T _nIn time period, all key points, the lease point of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

Q _Ki: the quantity of the bicycle that haulage vehicle K loads in lease point i service back;

Turn to the model that target is set up the public system call of going slowly with the total satisfaction maximum of lease point, objective function is as follows:

$\max Z=\underset{T_n\∈\mathrm{TS}}{\mathrm{\Σ}}\underset{i\∈N\left(T_n\right)}{\overset{t_i\∈T_n}{\mathrm{\Σ}}}\left[f_i\left(t_i\right)\right]---\left(1\right)$

Constraint condition is:

Q _ki＜Q _k $\∀i\∈M\left(t\right),k\∈K---\left(2\right)$

(1) formula is an objective function, the total satisfaction of expression maximization lease point; (2) formula is the dead weight capacity constraint condition of haulage vehicle, and expression at any time, the dead weight capacity of arbitrary lease point haulage vehicle all can not surpass its dead weight.

Rolling time domain dispatching algorithm: in urban transportation, public bicycles is in the state that constantly flows between each lease point, after an operation plan is determined, As time goes on, can constantly there be new site to send services request to the dispatching center, perhaps in current operation plan but also obtain of the automatic allotment of the site of service by the public bicycles traffic flow, no longer needed dispatch service, current like this operation plan just no longer is fit to new situation, must produce new operation plan immediately.Therefore adopt dynamic dispatching method,, regard each sub-dispatching cycle as static scheduling problem, realize dynamic vehicle scheduling by finding the solution several static scheduling problems with being divided into experimental process dispatching cycle whole dispatching cycle based on the rolling time domain.

At first whether have services request that public bicycles is leased the site to be divided into two classes according to the site is current:

1. current service window C is the current site set that services request is arranged;

2. non-current service window B is current not service and the site set of having served.

Working time is made as from t0 to tend in a working day, be that the dispatch service object is formulated operation plan at first with the point in the t0 C set constantly, the point that (t0+T) served in C when begin next dispatching cycle removes to set P, point with the new generation dispatch service request among P and the B joins among the C simultaneously, again to the dot generation operation plan among the C, and the like constantly roll and upgrade point among the C, arrive tend constantly until time t.

Adopt improved tabu search algorithm to find the solution the static scheduling problem: tabu search algorithm is since a given initial solution, in the current neighborhood of separating, determine some candidate solutions, if the adaptation value that optimal candidate is separated is better than the adaptation value of current optimum solution, whether then ignore it is that taboo is separated, this is separated the current optimum solution of replacement, both despised criterion, otherwise, be chosen in the non-taboo of choosing optimum in the candidate solution and separate the current optimum solution of replacement, iterate until satisfying stopping criterion.

Become the tabu search algorithm of neighborhood: when the adaptation value of separating that obtains when continuous several times iteration did not improve, then search procedure might be absorbed in the local optimum trap, and it is very important jumping out the quality that the local optimum point separates raising this moment.Therefore, adopt following measure to improve the quality of tabu search algorithm: when not improving, then to change neighbour structure, promptly in the current new neighborhood of separating is separated, search for through the adaptation value of separating after the continuous several times iteration.In the search procedure that becomes neighborhood, there is following notion:

(1) become the neighborhood search condition: when the adaptation value of separating after the continuous several times iteration in the search procedure does not improve, satisfy the neighborhood search condition that becomes, otherwise, do not satisfy the neighborhood search jumping condition that becomes;

(2) former neighborhood is separated: the neighborhood that does not satisfy when becoming the neighborhood search condition in the tabu search process is separated;

(3) becoming neighborhood separates: the neighborhood that satisfies when becoming the neighborhood search condition in the tabu search process is separated, and becomes neighborhood and separates to separate to compare with former neighborhood and have different neighbour structures.

Find the solution the process committed step of the static public system of going slowly with the tabu search algorithm that becomes neighborhood:

The expression of separating: represent to separate with the multilayer integer sequence, as " $\begin{array}{c}0-1-5-3-8-10-0\\ 0-2-4-7-9-0\end{array}$ " expression has two haulage vehicles to serve, the driving path of first transport vehicle for from the 0, parking lot successively through lease point 1,5,3,8 and lease point 10, return parking lot 0; The driving path of second transport vehicle returns parking lot 0 at last for pass through lease point 2,4,7,9 successively from the 0, parking lot.

The adaptation value function: adopting the objective function in the public system dynamics scheduling model of going slowly is the adaptation value function.

Neighbour structure:, therefore need two neighbour structures of design owing to adopted the search strategy that becomes neighborhood to improve the quality of separating:

(1) insertion: a bit being inserted on the paths generated neighborhood in another path separate.As shown below with shown in Fig. 3 .3, for two lines " $\begin{array}{c}1-2-3-4\\ 5-6-7-8\end{array}$ ", will put 2 and from article one route, move in the second road direction, the result who obtains for " $\begin{array}{c}1-3-4\\ 5-2-6-7-8\end{array}$ ”。

(2) exchange process: a pair of point of exchange on the two lines generates corresponding neighborhood and separate, and be as shown below with shown in Fig. 3 .4, for two lines " $\begin{array}{c}1-2-3-4\\ 5-6-7-8\end{array}$ ", point 3 in the exchange article one route and the point 6 in the second route, the neighborhood that obtains separate into " $\begin{array}{c}1-2-6-4\\ 5-3-7-8\end{array}$ ”。

In the iterative search procedures of the tabu search algorithm that becomes neighborhood, obtain former neighborhood with insertion and separate, obtain the change neighborhood with exchange process and separate, satisfy when becoming the neighborhood search condition, then adopt the searching method that becomes neighborhood, the rest may be inferred, finishes until algorithm.

Taboo object and taboo length: the taboo object is recorded in the taboo table, is illustrated in the search procedure operation of being avoided, and avoids the search procedure of making a circulation according to this, jumps out the local optimum state, improves the quality of separating.The adaptation value that this chapter employing is separated is as the taboo object of tabu search algorithm.Taboo length is being for to consider to despise under the situation of criterion, the maximum times avoided of taboo object, the size of taboo length generally rule of thumb or experiment determine that this chapter sets and avoids length and be Integral part, wherein n is the quantity of site.

Despise criterion: in candidate solution, the performance of certain state is better than that " " state is then ignored its taboo attribute to best so far, directly chooses it for current state.

Stop criterion: owing to the real-time of rolling Time-Domain algorithm is had relatively high expectations, so the stop criterion of tabu search algorithm is set at: when the search iteration number of times reached maximal value, algorithm stopped.

Claims (2)

1. public system dynamics dispatching method of going slowly based on rolling time domain dispatching algorithm, it is characterized in that: the described public system dynamics dispatching method of going slowly may further comprise the steps:

1), the setting-up time axle is represented a workaday whole dispatching cycle, on time shaft, each constantly corresponding scene, key point are meant to be accepted to serve or has haulage vehicle going to lease point on the road of this point, and the task of key point is to change;

Time window is the time range that public bicycles lease point allows service, describes a lease restriction range to service time, [WA with the Fuzzy Time window _i, WB _i] expression lease point i tolerable service time of scope, [WC _i, WD _i] scope service time of expression lease point i expectation, t _iThe time of expression vehicle point of arrival i; The satisfaction of lease point i is expressed as:

2), set up scheduling model: parking lot P0 has T transport vehicle, and haulage vehicle k dead weight is Q _k, send some haulage vehicles to provide service from parking lot P0 to n public bicycles lease point, the capacity of each lease point is E _i(i=1,2 ... n), at t constantly, the quantity of the bicycle of each lease point is q _i(t) (i=1,2 ... n); q _i(t) and E _iRatio be H _i(t), the H of lease point i under the normal condition _iScope be [C _Min, C _Max], the services request of lease point has two kinds of situations: 1. work as H _i(t)＞C _MaxThe time, expression lease point need transport bicycle constantly at t; 2. work as H _i(t)＜C _MinThe time, expression lease point need be transported into bicycle constantly at t; Lease point i tolerable service time of scope is [WA _i, WB _i], scope service time of expectation is [WC _i, WD _i]; Be s the service time of lease point i _i, t _iThe expression haulage vehicle arrives the time that lease point i begins to serve, f _i(t _i) satisfaction of expression lease point i;

The demand of lease point is worked as H for the quantity of a lease bicycle that need be transported into or transport _i(t)＞C _MaxThe time, the lease point need transport bicycle, and establishing lease point demand at this moment is d _i(t), after the lease point is accepted service, its H _iShould be at interval [C _Min, C _Max] in, then must satisfy $C_{\min }\≤\frac{q_i\left(t\right)-d_i\left(t\right)}{E_i}\≤C_{\max },$ Be q _i(t)-C _MaxE _i≤ d _i(t)≤q _i(t)-C _MinE _i, at this moment, d _i(t) get its lower limit q _i(t)-C _MaxE _iCan meet the demands; In like manner, work as H _i(t)＜C _MinThe time, d _i(t)=C _MinE _i-q _i(t);

N (t): the t moment, all lease points of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

M (t): the t moment, all key points, the lease point of not finishing the work, the new lease point that services request is arranged and the set in parking lot;

K: the set of haulage vehicle;

TS: time shaft;

Q _Ki: the quantity of the bicycle that haulage vehicle K loads in lease point i service back;

Turn to the model that target is set up the public system call of going slowly with the total satisfaction maximum of lease point, objective function is as follows:

$\max Z=\underset{t\∈\mathrm{TS}}{\mathrm{\Σ}}\underset{i\∈N\left(t\right)}{\mathrm{\Σ}}\left[f_i\left(t_i\right)\right]---\left(1\right)$

Constraint condition is:

Q _ki＜Q _k $\∀i\∈M\left(t\right),k\∈K---\left(2\right)$

3), will be divided into experimental process dispatching cycle whole dispatching cycle, will regard static scheduling problem each sub-dispatching cycle as, by finding the solution several static scheduling problems and the method for upgrading operation plan of constantly rolling realizes dynamic vehicle scheduling;

When finding the solution the static scheduling problem of every sub-dispatching cycle of correspondence, adopt based on the tabu search algorithm that becomes neighborhood and obtain operation plan; In the search procedure that becomes neighborhood, there is following notion:

(a) become the neighborhood search condition: when the adaptation value of separating after the continuous several times iteration in the search procedure does not improve, satisfy the neighborhood search condition that becomes, otherwise, do not satisfy the neighborhood search jumping condition that becomes;

(b) former neighborhood is separated: the neighborhood that does not satisfy when becoming the neighborhood search condition in the tabu search process is separated;

(c) becoming neighborhood separates: the neighborhood that satisfies when becoming the neighborhood search condition in the tabu search process is separated, and becomes neighborhood and separates to separate to compare with former neighborhood and have different neighbour structures.

The tabu search algorithm step that becomes neighborhood is as follows, comprising:

(3.1) produce initial solution, initiation parameter and taboo table;

(3.2) judge whether to satisfy end condition, if satisfy then introduce and export the result, otherwise, go to step (3.3);

(3.3) judge whether to satisfy the neighborhood condition that becomes, separate if satisfy the current change neighborhood of separating of generation, otherwise, generate the current former neighborhood of separating and separate;

(3.4) determine candidate solution;

(3.5) judge whether the satisfied criterion of despising, if, then replace current separating with specially pardoning to separate, go to step (3.2) and renewal and avoid showing, otherwise, then go to step (3.6);

(3.6) in candidate solution, separate current the separating of replacement, and upgrade the taboo table, go to step (3.2) with optimum non-taboo.

2. the public system dynamics dispatching method of going slowly based on rolling time domain dispatching algorithm as claimed in claim 1, it is characterized in that: in described step 3), when not improving through the adaptation value of separating after the continuous several times iteration, then change neighbour structure, promptly in the current new neighborhood of separating is separated, search for.

Images