CN106709680A - Method for optimizing optimal distribution route in dynamic logistics based on historical search information - Google Patents

Abstract

The invention provides a method for optimizing an optimal distribute route in dynamic logistics based on historical search information. The method provided by the invention has high efficiency in a dynamic logistics distribution problem and is capable of finding an optimal distribution route which is shortest and capable of meeting distribution requirements of each client between a distribution center and distribution clients. The value of paths between distribution clients in a traditional logistic problem is fixed, the modeling mode does not meet requirements of some practical problems taking the modeling mode as a model, that is, distribution nodes represented by the distribution clients and the length of distribution routes between the distribution clients can possibly be varying. Despite the variations, the optimal distribution route contains some common sub-routes at different moments, and the optimal sub-routes can be learned by a search algorithm of a follow-up moment, so that repeated searches are avoided, and thus the efficiency of searching the optimal distribution route is improved.

Description

Dynamic logistics based on historical search information most preferably dispense the optimization method of circuit

Technical field

The present invention relates to logistics distribution path optimization and the big field of evolutionary computation two, it is based on going through more particularly, to one kind The optimal dispatching line optimization method of dynamic logistics dispatching of history search information.

Background technology

Logistics distribution is a branch of combinatorial optimization problem, be the world plan strategies for educational circles it is unfailing research heat Point, is also the typical NP-hard problems for perplexing academia.The key problem of logistics distribution is needed in a fixed non-directed graph One is found in G={ U, E, P } can allow delivery vehicle (the deport)s, finally return back to home-delivery center from home-delivery center Most short dispatching circuit.The symbol definition of logistics distribution is as follows：

C={ 1,2,3,4 ..., U } represents one group of given dispatching client in above-mentioned parameter, and the client that U represents maximum compiles Number.In order to protrude effect of the historical search information in dynamic discrete particle cluster algorithm, we are by the value model of parameter [ei, li] It is [0,1440] to enclose, and its minimum particle size is minute.The value of Q is initialized as U, and the value of qi is 1 (the delivery unit of each dispatching client It is 1, the capacity of lorry is U unit), Pij is by specific data set generation.E=(i, j) | and i≤U, j≤U } represent by dispensing The set of the nonoriented edge that client is constituted, P={ p_ij| i ∈ [1, U], j ∈ [1, U] } it is that the distance between each dispatching client is constituted Path matrix.Be added to the concept of time window in logistics distribution by many scholars, that is, constitute the logistics with time window Dispatching problem.Delivery request so for client i has a time restriction TL_i=[e_i,l_i], wherein e_iRepresent client i's Earliest delivery period, l_iRepresent delivery period the latest.The value of each element in matrix is represented from a dispatching client to another One distance of dispatching client.The model of logistics distribution has successfully been expanded to path scheduling, genetic engineering, system In a series of practical problems such as optimization.What these models considered mostly is application under static scene, dispense client quantity and The distance between dispatching client is fixed.But in fact, generally can be with logistics distribution as model in real world should With often dynamically.That is, when urban transportation has highly dynamic property, such as contingency, road conditions problem, weather reason When, the quantity of client's Distribution path matrix and dispatching client in problem model is dynamic change.This results in dynamic logistics The generation of dispatching problem.Dynamic logistics dispatching problem has many application scenarios in reality, such as postman send letter problem with Dialing chauffeur problem.In other words, when some places do not have mail, postman avoids the need for these places and sends mail with charge free； If some cause specifics cause postman to send a circuit with charge free to one changed, then mention this relative to usual situation Sending with charge free may a little walk longer road.That is, being dynamic from the delivery client set of the distribution vehicle of home-delivery center Change, but its essence still will search out an optimal dispatching circuit so that distribution vehicle meets all matching somebody with somebody in delivery process Send customer requirement and operating range is most short.The popularity and actuality of dynamic logistics dispatching problem cause to ask dynamic logistics dispatching The research of topic not only great realistic meaning and also can be helped in logistic industry enterprise reduce cost of transportation.Due to dynamic logistics Dispatching question essence or combinatorial optimization problem, this allows for some evolution algorithms and can equally be well applied to dynamic logistics dispatching Optimal distribution project optimization in.At present, the evolution algorithm for being applied to the problem is broadly divided into two classes：Deterministic algorithm and evolution The class of algorithm two.Deterministic algorithm, relatively more classical has dynamic programming algorithm, branch-bound algorithm, K degree center tree algorithm and collection point Cut and col-generating arithmetic, be exactly to be easily trapped into local optimum the drawbacks of these algorithms are maximum so as to cause suboptimum distribution project.Phase Than under, evolution algorithm has wider solution space search capability, therefore than deterministic method more suitable for goer The optimization of the optimal distribution route of stream dispatching.Particle cluster algorithm is the important component of evolution algorithm, because its renewal rule is clear, It is simple and practical and there is solving precision high, strong robustness, fast convergence rate, the steady quality of solution, since the proposition just It is widely used.Current particle cluster algorithm mainly for the problem solving under continuous sky, in order to by its algorithm advantage The solution of dispersed problem is extended to, many scholars once attempted for particle cluster algorithm expanding to discrete space.A kind of comparing is successful It is the Discrete Particle Swarm Optimization Algorithm based on sets theory.This algorithm regards whole discrete search space as one entirely Collection, the corresponding feasible solution of particle is all counted as a subset of this complete or collected works, and the speed of particle is defined as being that one group of band is general The set on the side of rate.This discrete particle cluster algorithm has easily extension, steady performance and has had been extended to one In a little such as optimizations of the routing problem with time window.So, this dynamic discrete particle cluster algorithm based on set is very suitable Close the solution of optimal dispatching circuit in dynamic logistics dispatching problem.

The content of the invention

The present invention provides a kind of optimal dispatching line optimization method of dynamic logistics dispatching based on historical search information, the party Method can shorten the time of the dispatching circuit for searching more excellent.

In order to reach above-mentioned technique effect, technical scheme is as follows：

A kind of optimal dispatching line optimization method of dynamic logistics dispatching based on historical search information, comprises the following steps：

S1：Build dynamic logistics dispatching model；

S2：Dynamic logistics dispatching model to building is evolved using particle cluster algorithm under discrete space；

S3：Introduce population history optimal solution the dynamic logistics dispatching model after evolution is learnt to obtain excellent dispatching again Circuit.

Further, the detailed process of the step S1 is as follows：

When the path value changes between dispatching client, t₀Initial time is represented, path values are from moment t_kIt is converted to the moment t_k+1Mode be：

Wherein, For home-delivery center needs to dispense the quantity of client under initial time,It is t_kThe change weight factor in moment client i and client j paths, for the change size of metrology path value, k It is dispatching environment numbering, lp and up difference delegated paths change the maximum and minimum value of the factor,WithRepresent respectively In moment t_kWith in moment t_k+1In from dispatching client i to dispatching j path values size,It is from seing a visitor out in initial time Family i to the size of dispatching client's j path values, dispatching client i that its value and initial data are concentrated and the path between dispensing client j Value is equal, and pro is the parameter between [0,1], and in order to simulate path values dynamic change situation in non-directed graph G, model is every road Footpath is worthThe random number num for producing_ijIf, path valuesHave occurred and that change and num_ijNo more than parameter pro, then This paths value will be arranged to initial value, and otherwise its path values can become big；

When home-delivery center needs the customer quantity of dispatching to change, from moment t_kIt is converted to t_k+1Mode is：

Wherein,Represent in t_kThe client set that moment need not temporarily dispense,WithIt is illustrated respectively in t_kWhen Carve and t_k+1Moment needs the client set of delivery,For one can just can negative integer,rd_τRepresent random It is selecting and belong to setDispatching customer number.So ensure t at any time_kThe client for dispensing goods is needed to move The amplitude that state changes and changes has good controllability, so as to the reality accurately simulated and test optimal with line sending The performance of road algorithm.

Further, the detailed process of the step S3 is as follows：

S31：The speed and position and algorithm partial parameters of population are initialized, includingμ, number, whereinIt is whole Shape parameter, before expression to be introducedThe history of secondary dispatching environment intermediate ion group most preferably dispenses circuit, and μ represents that dispatching environment occurs The number of times of change, number represents the size of population, and algebraically counter iteration finally is set into 0；

S32：If dispatching environment does not change plus 1 by iteration, speed, position, the history of each particle are updated The global optimum of optimal dispatching circuit pbest and population dispenses circuit gbest；If the value of iteration is equal to Model can change dispatching environment according to the step of claim S1, then perform S33；

S33：If dispatching environment changes, judge whether the number of times of change is equal to μ.If equal to μ algorithms terminate, it is defeated Go out the path values of global optimum's distribution route of population；Otherwise by the momentAll particles The optimal dispatching circuit of history is stored in a specific collection, is designated as pset, wherein, k (k ∈ [1, μ]) is used for identifying dispatching ring Number in border.Pset set sizes be designated as π andThen population is reinitialized；

S34：Calculate each dispatching circuit pset in set pset_iDispatching client number θ_i, wherein i ∈ [1, π]；

S35：IfThen need the client of dispatching to reduce, at this moment need to will appear in pset_iWithout appearing inClient temporarily delete, keep pset_iIn other clients dispatching order it is constant；

S36：IfThen needing the client of dispatching increases, and at this moment needs that pset will not be appeared in_iAnd appear inClient be temporarily inserted into pset_i, to make current path values most short all the time in the process of insertion；

S37：IfThen illustrate in t_kMoment needs the customer quantity of dispatching not change, but dispatching client Between path values change, then pset_iThe optimal dispatching circuit for representing is not adjusted；

S38：Before savingThe history of all particles of population is optimal in secondary environment dispenses circuit and makes the appropriate adjustments Afterwards, dispatching circuit therein is screened using K-means clustering algorithms.In order to ensure dispensing circuit in set pset Diversity, randomly selects from π dispatching circuitCentered on individual path dispatching circuit, the path values for calculating each center are

S39：Calculate the path values L of each dispatching circuit_i(i ∈ [1, π]) arrives center λ_ψDifference and it is grouped into difference most Class belonging to small center, categorization results ξ_iψRepresent；

S310：The center of each class for having obtained is recalculated, method is as follows：

If ξ_iψIt is equal with ψ, then { ξ_iψ=ψ } operation result be 1, be otherwise 0；

S311：When the center newly calculated and equal original center, the distribution project cluster in pset is terminated, S39-S310 steps are otherwise repeated until former center is identical with the value at new center；

S312：After cluster terminates, the diversity value of the diversity value of each class, class is asked to be defined as follows：

The diversity value of class has reacted such diversity size to a certain extent, and diversity value is more big, matching somebody with somebody in class Send the otherness of route bigger and the bigger possibility that optimal solution is found with regard to representing of class diversity is higher；

S313：ChooseThe dynamic discrete that circuit replaces new initialization is dispensed in individual cluster in the maximum class of diversity value The optimal dispatching circuit of the history of some particles, then performs S32 in population.

Compared with prior art, the beneficial effect of technical solution of the present invention is：

The inventive method logistics distribution need to be found between home-delivery center and dispatching client one it is most short and full The optimal dispatching circuit that foot each client dispatching is required, the Distribution path between dispatching client and client in traditional logistics problem Value is all fixed, the need for this modeling pattern does not meet some practical problems as model, that is to say, that with seing a visitor out The path length between node and dispatching client representated by family is all probably change.Despite change, but in difference Moment, optimal dispatching circuit includes some public sub-line roads, and these optimal sub-line roads can be by the searching algorithm of following instant Practise, it is to avoid repeat search, so as to improve the search efficiency of optimal dispatching circuit.

Brief description of the drawings

Fig. 1-3 is the real example of dynamic logistics dispatching problem in the embodiment of the present invention；

Fig. 4 is the positional representation mode of particle in the embodiment of the present invention；

Fig. 5 is the speed representation of particle in the embodiment of the present invention；

Fig. 6 is the overall flow figure of the inventive method.

Specific embodiment

Accompanying drawing being for illustration only property explanation, it is impossible to be interpreted as the limitation to this patent；

In order to more preferably illustrate the present embodiment, accompanying drawing some parts have omission, zoom in or out, and do not represent actual product Size；

To those skilled in the art, it can be to understand that some known features and its explanation may be omitted in accompanying drawing 's.

Technical scheme is described further with reference to the accompanying drawings and examples.

Embodiment 1

As shown in fig. 6, a kind of optimal dispatching line optimization method of dynamic logistics dispatching based on historical search information, including Following processing procedure：

1st, the dynamic model of logistics distribution：

Dynamic logistics dispense modelling from initial time t₀Start.In t₀Under state, the dispatching client that will be accessed and Ren Weights and the medium ratio of test data set between two dispatching clients of meaning, but value reduced.After this, these test datas Collection can change in each following instant, and these continuous moment for changing can be expressed as：

T=t₀→t₁→t₂→…t_k…→t_μ

Wherein, t_kThe dispatching environment of population is inscribed when representing k-th, including is dispensed the distance of client and is dispensed client's Number.μ represents the total degree that the dispatching environment of home-delivery center changes.

In local map as shown in Figure 1, there are home-delivery center's (being represented with deport) and 10 dispatching clients, logistics The goal in research of dispatching problem is to find out from the dispensing vehicle of home-delivery center to service each under the requirement for meeting each client Dispatching client, finally returns back to the optimal dispatching circuit of home-delivery center.Wherein, what Fig. 1 was represented is traditional static dispatching problem, Its optimal distribution project is deport → c1 → c2 → c3 → c4 → c5 → c6 → c7 → c8 → c9 → c0 → deport.But Two kinds of situations generally occur in actual conditions：I) due to some path problems, such as peak period on and off duty or contingency is caused Blocking or traffic control, originally from c1 to c2 and the shortest path (path shown in dotted line) of c3 to c4 can become solid line institute's generation The path (shown in Fig. 2) of table；Ii) because c2 and c7 dispensing stations do not have the demand of goods delivery, so original dispatching circuit meeting From deport → c1 → c2 → c3 → c4 → c5 → c6 → c7 → c8 → c9 → c0 → deport become deport → c1 → c3 → C4 → c5 → c6 → c8 → c9 → c0 → deport (as shown in Figure 3).Two kinds of situations based on Fig. 2 and Fig. 3, we build them Mould into the distance between the dispatching client in dynamic logistics dispatching problem dynamic change and home-delivery center need delivery with seing a visitor out The dynamic change at family.For the situation shown in Fig. 2, model can be path values between each two dispatching client generate one [0,1] the several num between_ijIf, num_ijThe parameter pro and the path values of this edge being more than do not change, then this two Path values between individual dispatching client can become greatly so as to simulate the path problem being likely to occur；If num_ijNo more than pro and Path values between two dispatching clients have occurred and that change, then its path values can be reset as initial value (and initial time Path values it is equal).In the situation of the dispatching client's dynamic change shown in Fig. 3, in the same time, not dispensing if desired The quantity of client is reduced, then some dispatching clients temporarily remove from dispatching list, in otherwise never dispensing client set Choose some increase to dispatching client set in.So allow for not needing to match somebody with somebody in the same time from the dispensing vehicle of home-delivery center The client of delivery thing will dynamically change.Equally, the optimal dispatching circuit of dispensing vehicle is not also what dynamic changed in the same time.

2nd, the evolution strategy of dynamic discrete particle cluster algorithm：

During the optimal dispatching circuit of search, each particle is with a kind of feasible dispatching circuit of table, and it is fast for home-delivery center Degree and location updating formula are：

Wherein, i (i ∈ [1, number]) represents the numbering of particle, and number represents the size of population,Represent i-th The position of the d dimensions of individual particle,I-th speed of the d dimensions of particle is represented,The history for representing i-th particle is optimal The position of the d dimensions of position, ω is inertia weight, and c is accelerator coefficient, f_iD () represents i-th particle right in the study that d is tieed up As r^dIt is two from 0 to 1 equally distributed random number.In discrete space, the position of particle is made up of some discrete sides The loop that home-delivery center is finally returned to from home-delivery center.The speed of particle is some discrete side collection with probability Constitute.The speed that we set each particle is some discrete sides, and the position of particle is an end to end Hamiltonian cycle. In discrete environment, the whole search space of population is expressed as complete or collected worksAnother expression of E Mode isThe position of each particle is represented by (as shown in Figure 4).The speed of particle can be expressed as(as shown in Figure 5).When the speed of particle After renewal, we are screened opposite side:Wherein α pre-sets Screening parameter, σ representative edges, pr (σ) represents the corresponding probability in the side, and i is particle numbering, and d represents dimension.Only when side is corresponding Probability can just be retained more than α, can so abandon the less side of some probability.With the side compared with small probability be difficult to be selected for Distribution route is set up, being eliminated speed collection can accelerate the search speed of population.After the speed of particle updates, grain Son builds a complete goods delivery route in the following manner：

(1) first fromSets of speeds in choose appropriate side；

(2) ifIn side collection can not constitute complete dispatching circuit, particle just chooses suitable from current location When side；

(3) if the Bian Renran in (1st) step and (2nd) step can not meet condition, particle is selected in will being chosen at complete or collected works E Suitable side is taken to build complete dispatching circuit.

3rd, path values dynamically shift gears between dispatching client：

In the case of weights change, each specific state has a unique path matrix corresponding.In t_k Weight matrix under state be designated as P andRepresent from dispatching client i to dispatching client j path values.In static mould In type, p_ijPath values between represented client i to client j are constant all the time, so all the time can be with symbol p_ijRepresent it.But It is that the path values in the same time between client i to client j are not dynamic changes in the model that weights dynamically change.For area Point, dynamic model is usedRepresent in moment t_kThe path values of (k ∈ [1, μ]) between them.In t₀Moment its value and static mould The value of type is identical, but it is change in following instant.From moment t_kTo moment t_k+1Transformation can be expressed as follows：

Wherein, For home-delivery center needs to dispense the quantity of client under initial time,It is t_kThe change weight factor in moment client i and client j paths, lp and up difference delegated path change because The maximum and minimum value of son,WithRepresent respectively in moment t_kWith in moment t_k+1In, from dispatching client i to dispatching j Path values size,It is from dispatching client i to the size of dispatching client's j path values, num in initial time_ijBe for The random number for producing.That is, in the case of the dynamically change of customer path value is dispensed, goods is needed in all moment The customer quantity of dispatching is constant, i.e.,

4th, dispatching client dynamically shifts gears：

In the case where dispatching client dynamically changes, the client set for dispensing goods is needed to be expressed as in initial timeDispatching client's sum is designated asIn static logistics distribution, because dispatching client is constant all the time, it is possible to use Same symbol U dispenses the quantity of client to represent, but in the dynamic logistics dispatching problem represented by Fig. 2, not in the same time Dispatching customer quantity is different, for the purposes of distinguishing it, Wo MenyongTo represent that moment k needs the quantity of the client of dispatching. WhereinRepresent initial time dispatching client's number andFor example, in eil76 data sets, dispensing customer quantity It is 76, in static logistics distribution, U=76, but in dynamic logistics dispatching, the dispatching customer quantity of initial timeSo the client set of initial time is a complete or collected works, follow-up any one moment t_kThe middle visitor for needing to dispense goods Gather at familySubset and dispatching customer quantity on meett_kThe client of dispatching goods is needed in moment Set is designated asSize is designated ast_kThe set that moment need not temporarily dispense the client of goods is designated asSo collect CloseSize beFrom moment t_kTo t_k+1Transformation in, some dispatching clients can by temporarily from dispatching gather Middle addition or deletion.One kind of middle dispatching client set is expressed as

WhereinRepresent in moment t_kHome-delivery center needs the maximum customer number of dispatching.In transformationIt Before, model can at random generate one can just can negative integerMeet following two conditions：

Wherein Sc (Sc ∈ [0,1]) and Sd (Sd ∈ [0,1]) are called controlling elements, and Sc is used for controlling each moment to need to match somebody with somebody The quantitative range of the client for sending, Sd is used for controlling to dispense two differences of adjacent moment dispatching client during client dynamically shifts gears Size.That is, t at any time_k,WithTransform mode be expressed as follows

Wherein,Represent from setIn pick outThe individual dispatching client for meeting condition, this A little dispatching clients for meeting condition use rd_τRepresent.When the customer quantity for needing dispatching is reduced, once this customer number occurs In, the client is just designated as rd_τAnd by its fromIn temporarily delete.When needing the quantity of client of dispatching to increase, Model will be randomly selectedIt is individualIn but do not existClient be added toIn.This assures in office Meaning moment t_kThe amplitude for needing the client for dispensing goods dynamically to change and change has good controllability.

5th, the implementation method of the optimal dispatching circuit of population history is introduced：

In dynamic model, when the number for dispensing client is n, it is n* to constitute the size that side integrates by these dispatching clients (n-1)/2, the size of solution space isWith the increase of dispatching scale of consumer, Home-delivery center will be in geometric growth in the search complexity of solution space.In dynamic logistics dispatching problem, although some dispatchings visitor Path values between family can change.But these optimal dispatching circuits are all more or less deleted, add or recombinated Dispatching client, and the sequence constituted between other local dispatching clients is still constant, for example in figure 3, although c2 and C7 is without accessing, but the local subpaths of c3 → c4 → c5 → c6 and c8 → c9 → c0 this two do not change.Change sentence Talk about, these local subpaths can be learnt and be optimized by the population under new environment.This learning strategy again can keep away Exempt from repeat search and accelerate convergence of algorithm speed, it is more excellent so as to allow home-delivery center that an overall situation is found within the shorter time Dispatching circuit.Because when different, part dispatching client can temporarily be added into dispatching list or from dispatching list Delete, when dispensing client and changing, we by all particle search of previous section moment population to optimal match somebody with somebody Line sending road saves and is expressed as set pset.Before the optimum line optimization that home-delivery center carries out subsequent time, these The pset being conserved will replace the history optimum line of population some particles, and that is inscribed when so new is optimal with line sending Some discrete sides of the local optimum subpath on road can be learned to and be added to the speed concentration of current particle, so Repeat search is just avoided well, so as to improve the efficiency of algorithm.

In order to the optimal dispatching circuit of history is incorporated into t_kWe will newly in the population of initialization at momentThe history of all particles of population at moment most preferably dispenses circuit and is stored in pset.Introducing Before the history preserved in pset most preferably dispenses circuit, if the quantity that client is dispensed in subsequent time changes, need Will to pset in optimal dispatching circuit be modified, step is as follows：

(1) each dispatching circuit pset in pset is calculated_iDispatching client number θ_i, wherein i ∈ [1, π]；

(2) ifThen need the client of dispatching to reduce, at this moment need to will appear in pset_iWithout appearing inClient temporarily delete, keep pset_iIn other clients dispatching order it is constant；

(3) ifThen needing the client of dispatching increases, and at this moment needs that pset will not be appeared in_iAnd appear inClient be temporarily inserted into pset_i.To make current shortest path all the time in the process of insertion.

(4) ifThen illustrate in t_kMoment need dispatching customer quantity do not change, but with seing a visitor out between Path values change, then pset_iIn optimal dispatching circuit can be introduced directly into.

From moment t_kTo moment t_k+1During transformation, the dispatching circuit of the optimal classification in pset will be selected as new particle group The optimal dispatching circuit of history of some particles, so may insure that the local optimum circuit of the optimal dispatching circuit in the new moment can Arrived with by study as early as possible.In follow-up search, the optimal dispatching circuit of history that these are introduced in population can be substituted To prevent whole population to be excessively absorbed in local optimum by the attraction of these circuits in optimization process.Based on historical search The flow chart of the dynamic discrete particle cluster algorithm optimization dynamic logistics dispatching problem of information is as shown in Figure 6.In dynamic logistics dispatching In the test data of problem, the maximum duration at each moment isSecondary iteration, in order in simulating reality as far as possible Different degrees of dispatching environment changes, and the relative parameters setting of dynamic model is as follows：

Parameter	Value
		lp	1.0
up	{1.5,2.0}
		{Sc,Sd}	{0.8,0.05},{0.7,0.1},{0.6,0.15}
pro	{0.75,0.65,0.55}
		μ	30

The relative parameters setting of dynamic discrete particle cluster algorithm is as follows：

In order to verify above-mentioned strategy validity, we test in 15 groups of test datas altogether.Result shows, dynamic The speed and location Update Strategy energy of dynamic discrete particle cluster algorithm in the search problem of the optimal dispatching circuit of state logistics distribution Enough keep good diversity.Additionally, the optimal dispatching circuit by population at the preamble moment is saved as next The strategy of the optimal dispatching circuit of history of the population at moment can allow the part of these optimal dispatching circuits optimal sub with line sending Road is arrived by fully study, unnecessary repeat search is reduced, so that it is guaranteed that the dispensing vehicle of home-delivery center is in the shorter time Inside search more excellent dispatching circuit.On the whole, the algorithm search time is shortened %5-%12 by the method.

The same or analogous part of same or analogous label correspondence；

Position relationship for the explanation of being for illustration only property described in accompanying drawing, it is impossible to be interpreted as the limitation to this patent；

Obviously, the above embodiment of the present invention is only intended to clearly illustrate example of the present invention, and is not right The restriction of embodiments of the present invention.For those of ordinary skill in the field, may be used also on the basis of the above description To make other changes in different forms.There is no need and unable to be exhaustive to all of implementation method.It is all this Any modification, equivalent and improvement made within the spirit and principle of invention etc., should be included in the claims in the present invention Protection domain within.

Claims (3)

1. a kind of dynamic logistics based on historical search information most preferably dispense the optimization method of circuit, it is characterised in that including with Lower step：

S1：Build dynamic logistics dispatching model；

S2：Dynamic logistics dispatching model to building is evolved using particle cluster algorithm under discrete space；

S3：Introduce population history optimal solution the dynamic logistics dispatching model after evolution is learnt to obtain excellent with line sending again Road.

2. the dynamic logistics dispatching based on historical search information according to claim 1 is optimal dispenses line optimization method, Characterized in that, the detailed process of the step S1 is as follows：

When the path value changes between dispatching client, t₀Initial time is represented, path values are from moment t_kIt is converted to moment t_k+1's Mode is：

$p_{{\mathrm{ijt}}_0}*lp\≤p_{{\mathrm{ijt}}_{k+1}}\≤p_{{\mathrm{ijt}}_0}*up$

Wherein, For home-delivery center needs to dispense the quantity of client under initial time,For t_kThe change weight factor in moment client i and client j paths, for the change size of metrology path value, k is compiled for dispatching environment Number, lp and up difference delegated paths change the maximum and minimum value of the factor,WithRepresent respectively in moment t_kWith when Carve t_k+1In from dispatching client i to dispatching j path values size,It is from dispatching client i to dispatching client j in initial time The size of path values, its value and initial data concentrate the path values between dispatching client i and dispatching client j equal, pro for [0, 1] parameter between, in order to simulate the situation of free routing value dynamic change in non-directed graph G, model is every paths valueProduce A raw random number num_ijIf, path valuesHave occurred and that change and num_ijNo more than parameter pro, then this paths value Initial value will be arranged to, otherwise its path values can become big；

When home-delivery center needs the customer quantity of dispatching to change, from moment t_kIt is converted to t_k+1Mode is：

$C_{1t_{k+1}}=\left\{\begin{array}{cc}{C}_{1t_k}\&cup;\left\{{\mathrm{rd}}_{\mathrm{\&tau;}}\right\}|{\mathrm{rd}}_{\mathrm{\&tau;}}\&Element;C_{2t_k},& if{\mathrm{Rn}}_{t_{k+1}}>0\\ C_{1t_k}-\left\{{\mathrm{rd}}_{\mathrm{\&tau;}}\right\}|{\mathrm{rd}}_{\mathrm{\&tau;}}\&Element;C_{1t_k},& if{\mathrm{Rn}}_{t_{k+1}}<0\end{array}\right.$

Wherein,Represent in t_kThe client set that moment need not temporarily dispense,WithIt is illustrated respectively in t_kMoment and t_k+1Moment needs the client set of delivery,For one can just can negative integer,rd_τExpression is selected at random And belong to setDispatching customer number, be so able to ensure that t at any time_kNeed client's dynamic of dispatching goods The amplitude for changing and changing has good controllability, so as to the reality accurately simulated and test optimal dispatching circuit The performance of algorithm.

3. the dynamic logistics dispatching based on historical search information according to claim 1 is optimal dispenses line optimization method, Characterized in that, the detailed process of the step S3 is as follows：

S31：The speed and position and algorithm partial parameters of population are initialized, includingμ, number, whereinIt is shaping ginseng Number, before expression to be introducedThe history of secondary dispatching environment intermediate ion group most preferably dispenses circuit, and μ represents that dispatching environment changes Number of times, number represents the size of population, algebraically counter iteration finally is set into 0；

S32：If dispatching environment not change and plus 1 by iteration, the speed of each particle, position, history are updated optimal The global optimum of dispatching circuit pbest and population dispenses circuit gbest；If the value of iteration is equal toModel Dispatching environment can be changed according to the step of claim S1, S33 is then performed；

S33：If dispatching environment changes, judge whether the number of times of change is equal to μ.If equal to μ algorithms terminate, grain is exported The path values of global optimum's distribution route of subgroup；Otherwise by the momentAll particles history most Excellent dispatching circuit is stored in a specific collection, is designated as pset, wherein, k (k ∈ [1, μ]) is used for identifying dispatching environment numbering, Pset set sizes be designated as π andThen population is reinitialized；

S34：Calculate each dispatching circuit pset in set pset_iDispatching client number θ_i, wherein i ∈ [1, π]；

S35：IfThen need the client of dispatching to reduce, at this moment need to will appear in pset_iWithout appearing in Client temporarily delete, keep pset_iIn other clients dispatching order it is constant；

S36：IfThen needing the client of dispatching increases, and at this moment needs that pset will not be appeared in_iAnd appear in's Client is temporarily inserted into pset_i, to make current path values most short all the time in the process of insertion；

S37：IfThen illustrate in t_kMoment needs the customer quantity of dispatching not change, but between dispatching client Path values change, then pset_iThe optimal dispatching circuit for representing is not adjusted；

S38：Before savingThe optimal dispatching circuit of the history of all particles of population and after making the appropriate adjustments in secondary environment, profit Dispatching circuit therein is screened with K-means clustering algorithms, in order to ensure dispensing the diversity of circuit in set pset, Randomly selected from π dispatching circuitCentered on individual path dispatching circuit, the path values for calculating each center are

S39：Calculate the path values L of each dispatching circuit_i(i ∈ [1, π]) arrives center λ_ψDifference and that it is grouped into difference is minimum Class belonging to center, categorization results ξ_iψRepresent；

S310：The center of each class for having obtained is recalculated, method is as follows：

${\mathrm{\&lambda;}}_{\mathrm{\&psi;}}=\frac{\underset{i=1}{\overset{\mathrm{\&pi;}}{\&Sigma;}}{L}_i*\{{\mathrm{\&xi;}}_{i\mathrm{\&psi;}}=\mathrm{\&psi;}\}}{\underset{i=1}{\overset{\mathrm{\&pi;}}{\&Sigma;}}1*\{{\mathrm{\&xi;}}_{i\mathrm{\&psi;}}=\mathrm{\&psi;}\}}$

If ξ_iψIt is equal with ψ, then { ξ_iψ=ψ } operation result be 1, be otherwise 0；

S311：When the center newly calculated and equal original center, the distribution project cluster in pset is terminated, otherwise S39-S310 steps are repeated until former center is identical with the value at new center；

S312：After cluster terminates, the diversity value of the diversity value of each class, class is asked to be defined as follows：

${\mathrm{dl}}_{\mathrm{\&psi;}}=\underset{i=1}{\overset{\mathrm{\&pi;}}{\&Sigma;}}|L_i-{\mathrm{\&lambda;}}_{\mathrm{\&psi;}}{|}^2*\{{\mathrm{\&xi;}}_{i\mathrm{\&psi;}}=\mathrm{\&psi;}\}$

The diversity value of class has reacted such diversity size, the dispatching road in the more big then class of diversity value to a certain extent The otherness of line is bigger and diversity of class it is bigger just represent find optimal solution possibility it is higher；

S313：ChooseThe dynamic discrete particle that circuit replaces new initialization is dispensed in individual cluster in the maximum class of diversity value The optimal dispatching circuit of history of some particles, then performs S32 in group.