CN103699982A - Logistics distribution control method with soft time windows - Google Patents

Abstract

A logistics distribution control method with soft time windows includes the following steps: (A1) a network model is built, cost resistances are assigned to roads to which network data are concentrated, and taking road nodes into consideration, toll weights are assigned to road traffic light intersections and toll stations; (A2) an optimized vehicle routing model with soft time windows (VRPTW) is built, a target function is established with lowest transportation costs, and the transportation costs are respectively composed of fixed distribution vehicle cost, transportation cost, vehicle waiting cost and delay cost; (A3) a fuzzy clustering analysis algorithm is designed, and a method based on the integration of quantitative analysis and qualitative analysis is adopted for clustering; (A4) a heuristic optimized vehicle routing algorithm is designed, the optimized vehicle routing algorithm is adopted for distribution target nodes in each class, and thereby a distribution result can be obtained. The logistics distribution control method with soft time windows adopts the distances of actual delivery road network routes between distribution nodes as a calculation basis and also takes the actual traffic capacities of roads, large network node number and transportation time needed by distribution nodes into consideration.

Description

A kind of logistics distribution control method with soft time window

Technical field

The present invention relates to Communication and Transportation Engineering, geographic information data is processed, computer application field, and operational research, graph theory and network analysis, Management Science and Engineering, in particular, logistics distribution field.

Background technology

Along with the develop rapidly of economic globalization and the network information technology, logistics distribution has caused people's common concern as a new growth engines.Dispensing is the core link of logistics system, is to be accompanied by market and a kind of inevitable market behavior of being born, and along with the fierceness day by day of market competition and improving constantly of customer requirement, dispensing will be played very important effect in following market competition.In dispensing business, the involvement aspect of Optimized scheduling of distribution vehicles problem is wider, needs the factor of consideration also a lot, and the impact of distribution enterprise being improved service quality, reduced operating cost, increase economic benefit is also very large.

Chinese scholars is put forth effort on research VRPTW (vehicle routing problem with time windows) problem, main because it is the key problem of logistics distribution and communications and transportation, only has the scheduling problem of having solved just can make to provide and deliver effective and reasonable.The objective function of VRPTW problem can be described as vehicle production and disperse the total cost of vehicle line between client's point minimum to each.Line design principle is, each client's point can only be accessed by a car, and constrained time window, if do not send to goods within time window, logistics center need to pay extra cost.The freight demand total amount that client on each line is ordered can not surpass the dead weight capacity of the vehicle on circuit.Each client's demand must meet, and can only be provided and delivered by a car.VRPTW problem practicality is stronger, especially at 3PL(thirdparty logistic) in.Lenstra and Kan (1981) prove that VRPTW problem belongs to NP difficulty combinatorial optimization problem.The solution of VRPTW problem is abundant, and more common can be divided into exact algorithm and heuritic approach.While adopting exact algorithm to solve VRPTW problem, time complexity is too high.In recent years, although some scholars have used exact algorithm in research VRPTW problem, heuritic approach can be within feasible time complexity optimization problem, Most scholars is still puted forth effort to study heuritic approach and is solved VRPTW problem.Generally speaking, VRPTW problem is more difficult than VRP problem.Therefore, adopt heuritic approach to solve VRPTW problem better.In the recent period, by heuritic approach, solved VRPTW problem and obtained good result.

But which kind of method that don't work solves logistics distribution all seldom can take 3 problems below into account: 1. the distance between each dispensing client point is to using its air line distance as basis, departed from the actual road network between dispensing client point; 2. the research of existing VRPTW problem, does not consider the driving actual conditions of road, as: the geography information factors such as through-current capability, road circuit node are taken into account; 3. mostly existing research is for hard time window situation, requires vehicle in section, to arrive at the appointed time, and less consideration vehicle can arrive outward at time window, but can increase cost.

Therefore, there is defect in existing Logistics Distribution Method, needs to improve.

Summary of the invention

For overcome between each the dispensing point in existing means of distribution, not take time window as classification foundation, do not consider road quality, the deficiency such as the geography information factors such as negotiability, net number are less, the invention provides a kind of actual road network circuit distance of receiving of take between dispensing point is basis, considers that the actual driving ability of road, net number are large, the logistics distribution control method with soft time window of dispensing point to the freight demand time simultaneously.

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

With a logistics distribution control method for soft time window, described logistics distribution control method comprises the following steps:

A1. set up network model, use ArcGIS software, set up the topological relation between Network data set and road network; According to the Shap_Length field attribute in vector data, set up the concentrated road resistance of network data; Again Network data set is carried out to network analysis, the OD matrix of the least cost that obtains providing and delivering between destination node, for optimizing and scheduling vehicle model provides Vehicle Driving Cycle expense weights, considers road circuit node, for road traffic lamp crossing, charge station give expense weight;

A2. set up and be with soft time window optimizing and scheduling vehicle model VRPTW;

Problem is described: have m production ground A to turn out a produce, its output is respectively a _i, there is n client to put B, its demand is respectively b _j, the demand of ordering according to client is provided and delivered product at the appointed time in window, if every dispensing task does not complete in the time range of appointment, punish;

Require the travel route of each vehicle to meet constraint condition: the starting point of every route of i. must be grown place, terminal is client's point, does not consider backhaul; Ii. the dead weight capacity of every route must not be greater than the maximum authorized payload of carrier vehicle; Iii. each given client must be only once serviced; Iv. each client has the restriction of its serviced official hour window, if vehicle reaches client's point in advance, pays waiting cost, if vehicle delays to reach client's point, delays in payment expense;

With trucking costs, set up objective function, trucking costs is comprised of distribution vehicle fixed expense, trucking costs, vehicle waiting cost and deferred charges respectively, wherein, distribution vehicle fixed expense is comprised of vehicle depreciation expense and maintenance cost, this part expense is only relevant with the vehicle number of dispensing, the section situation that trucking costs is exercised with vehicle is relevant, comprise road section length, the crowded state in section, also comprise in whole section the traffic lights of process, the time cost that charge station produces and charge situation, client's freight demand will be sent within the scope of the stipulated time client conventionally, otherwise logistics center need to pay extra cost, the optimizing and scheduling vehicle model is here to set up for the destination node in class after classification, concrete model is as follows:

Dispensing road net model is described as:

$\left\{\begin{array}{c}G=(A\∪B\∪V,R)\\ A=\{a_1,a_2,...,a_n\}\\ B=\{b_1,b_2,...,b_m\}\\ V=\{v_1,v_2,...v_k\}\\ R=\left\{R_s\right\}\end{array}\right.$

Wherein A is grown place set in road network, and B is client node set in road network, and V is road intersection point set in road network, and they form the summit of network, and R is the oriented section collection in road;

The minimum model of transportation cost is: with the dynamic vehicle path planning problem of soft time window, objective function is:

$\min Z=\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{c}_k+\underset{g=0}{\overset{n}{\mathrm{\Σ}}}\underset{h=0}{\overset{n}{\mathrm{\Σ}}}\underset{k=1}{\overset{k}{\mathrm{\Σ}}}{c}_{\mathrm{gh}}{x}_{\mathrm{ghk}}+p_1\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\max (S_j-t_j,0)+p_2\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\max (t_j-E_j,0)$

s.t.

①a _i＞b _j,i∈{1,2,...,m},j∈{1,2,...,n}；

2. c _kfixed cost for each car;

3.

m round numbers, is distribution vehicle number, and a is parameter, 0 < a < 1, and constraint condition is more, and cargo handler is more complicated, and a is less;

4. c _ghfor section (v _g, v _h) transportation cost, relevant with the crowded state in section with the length in this section, and the quantity of the traffic lights comprising and charge station is relevant, c _gh=c _gh' d _gh+ c _gh' ' l _ghc wherein _gh' be section (v _g, v _h) expense of unit distance, d _ghfor section (v _g, v _h) distance, c _gh' ' be section (v _g, v _h) the traffic lights of process and the unit costs of charge station, l _ghthe traffic lights comprising for this section and the quantity of charge station;

5. x _ghk={ 0,1}, vehicle k is through section (v _g, v _h), x _ghk=1, otherwise be 0; G, h ∈ 0,1,2..., n}, k ∈ 1,2 ... m};

6.

y _jkif expression client puts the task of j and is completed by vehicle k, y _jk=1, otherwise y _jk=0; Q is the dead weight of vehicle;

7. [S _j, E _j] client puts the distribution time window requirement of j;

8. t _jfor vehicle arrives the moment that client puts j;

9. p ₁, p ₂be respectively early than arriving with exceeding the punishment cost coefficient that client puts j time window;

Wherein min represents minimum, and max represents maximum, and s.t. represents constraint condition;

A3. fuzzy cluster analysis: adopt the method combining based on quantitative test and qualitative analysis to carry out cluster, first according to the time window attribute of destination node, client's point is carried out to Preliminary division, and then in conjunction with quantitative method, by Customer Location, carry out client's cluster analysis, by fuzzy cluster analysis, the problem of large-scale vehicle path planning is demoted, change into small-scale combinatorial optimization problem and solve, fuzzy cluster analysis step is as follows:

3.1) obtaining of customer order information, comprises client's geographic position and demand;

3.2), according to the object of research, the index in close relations of selection and research object, first carries out quantitative classification;

3.3) standardization to data;

3.4) set up fuzzy similarity matrix;

3.5) foundation of fuzzy equivalence relation;

3.6) carry out cluster analysis, given different confidence level, asks R _λcut battle array, the λ that finds out R shows, each sample is classified as a class, along with the reduction of λ, by thin chap gradually and class;

A4. optimizing and scheduling vehicle heuritic approach design, by A3, will be large and change littlely compared with large dispensing impact point, now again the dispensing destination node in each class is adopted to optimizing and scheduling vehicle algorithm, can obtain providing and delivering result, step is as follows:

4.1) adopt natural number coding method, according to client in class, count out and produce initial population and carry out genetic coding, structure client point is as the chromosome in distribution route optimization problem solution vector,

4.2) calculate the population's fitness function of each distribution project, for objective function, get minimized combinatorial optimization problem, its fitness function will carry out suitable variation to objective function, to be converted into the situation of maximization, and guarantee that fitness value is non-negative,

4.3) method that optimized individual preservation method and roulette are selected to combine is deleted, duplicated chromosome, finally produces new population,

4.4) employing order bracketing method is implemented interlace operation, with crossover probability p _cpopulation is carried out to interlace operation, checks and whether meet constraint condition, produce new population,

4.5) adopt and repeatedly exchange variation method, with the Probability p that makes a variation _mpopulation is carried out to mutation operation, produces at random an exchange times L, check and whether meet constraint, form new population,

4.6) judge whether to meet stopping criterion, reach maximum iteration time or reach optimum solution requirement, meet the demands and stop, otherwise proceed to 4.3),

4.7) result of calculation is decoded,

4.8) select all decoded result of calculation, and compare and choose expense reckling.

Beneficial effect of the present invention is mainly manifested in: logistics distribution of the present invention has considered that client puts the time window that receives goods, by fuzzy cluster analysis, client's point is classified, the problem of large-scale vehicle path planning is changed into and combines on a small scale optimization problem.In addition, in conjunction with GIS, also considered that the geography information factors such as negotiability, road circuit node give weight to road network, adopted heuritic approach to solve, improved the range of application of the accuracy of logistics distribution cost budgeting and the validity of decision-making and logistics distribution.

Accompanying drawing explanation

Fig. 1 is the distribution schematic diagram of dispensing point point of sale on the road network of Hangzhou.

Fig. 2 adopts the schematic diagram after fuzzy cluster analysis to whole dispensing destination nodes.

The schematic diagram of the path implement under the least cost of the not free window constraint of Fig. 3.

The schematic diagram of the operation result of the soft time window constraint condition of Fig. 4.

Fig. 5 is the process flow diagram with the logistics distribution control method of soft time window.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described.

With reference to Fig. 1～Fig. 5, a kind of logistics distribution control method with soft time window, comprises the following steps:

A1. set up network model, use ArcGIS software, set up the topological relation between Network data set and road network; According to the Shap_Length field attribute in vector data, set up the concentrated road resistance of network data; Again Network data set is carried out to network analysis, the OD matrix of the least cost that obtains providing and delivering between destination node, for optimizing and scheduling vehicle model provides Vehicle Driving Cycle expense weights, considers road circuit node, for road traffic lamp crossing, charge station give expense weight;

A2. set up and be with soft time window optimizing and scheduling vehicle model (VRPTW);

With trucking costs, set up objective function, trucking costs is comprised of distribution vehicle fixed expense, trucking costs, vehicle waiting cost and deferred charges respectively; The minimum model of transportation cost is: with the dynamic vehicle path planning problem of soft time window, objective function is:

$\min Z=\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{c}_k+\underset{g=0}{\overset{n}{\mathrm{\&Sigma;}}}\underset{h=0}{\overset{n}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{k}{\mathrm{\&Sigma;}}}{c}_{\mathrm{gh}}{x}_{\mathrm{ghk}}+p_1\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}\max (S_j-t_j,0)+p_2\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}\max (t_j-E_j,0)$

A3. Fuzzy Cluster Analysis Algorithm design, the method that employing combines based on quantitative test and qualitative analysis is carried out cluster, first according to the time window attribute of destination node, client's point is carried out to Preliminary division, and then in conjunction with quantitative method, carry out client's cluster analysis;

A4. the design of optimizing and scheduling vehicle heuritic approach adopts optimizing and scheduling vehicle algorithm to the dispensing destination node in each class, can obtain the result of providing and delivering;

Described method, wherein, in steps A 2, sets up the optimizing and scheduling vehicle model with soft time window, the schematic diagram of dispensing road network, as shown in Figure 1.

In described method, wherein, in steps A 3, as follows to the operation result of soft time window constraint condition: square frame point represents to exceed time window, as shown in Figure 3.

In described method, wherein, in steps A 4, city of Hangzhou freight transportation: 4 grown places, 30 client's points: 16 some time windows are 9:30-11:30,14 some time windows are 14:30-15:30, from 4 dispensing points, to 30 point of sale deliveries, by fuzzy cluster analysis, the allocator in each class is all the same.

The geographic coordinate that 30 clients are ordered is as shown in table 1:

The geographic coordinate of each point of sale of table 1

5 class clients put cluster centre and covering is counted as shown in table 2:

Table 2 point of sale cluster centre and covering are counted

The client of take in class count as the class of 8 be example, move 50 times, obtain best distribution project and be:

The route of the 1st car is: 0->5>0

The route of the 2nd car is: 0->7->3->2-Gre atT.GreaT.GT0

The route of the 3rd car is: 0->8->1->0

The route of the 4th car is: 0->4->6->0

In distribution project total kilometres the shortest be 29.08961km, the path implement under least cost is as shown in Figure 4.

What more than set forth is the excellent results that embodiment shows that the present invention provides, obviously the present invention is not only applicable to above-described embodiment, can do many variations to it and is implemented not departing from essence spirit of the present invention and do not exceed under the prerequisite of the related content of flesh and blood of the present invention.

Claims (1)

1. with a logistics distribution control method for soft time window, it is characterized in that: described logistics distribution control method comprises the following steps:

A1. set up network model, use ArcGIS software, set up the topological relation between Network data set and road network; According to the Shap_Length field attribute in vector data, set up the concentrated road resistance of network data; Again Network data set is carried out to network analysis, the OD matrix of the least cost that obtains providing and delivering between destination node, for optimizing and scheduling vehicle model provides Vehicle Driving Cycle expense weights, considers road circuit node, for road traffic lamp crossing, charge station give expense weight;

A2. set up and be with soft time window optimizing and scheduling vehicle model VRPTW;

Problem is described: have m production ground A to turn out a produce, its output is respectively a _i, there is n client to put B, its demand is respectively b _j, the demand of ordering according to client is provided and delivered product at the appointed time in window, if every dispensing task does not complete in the time range of appointment, punish;

Require the travel route of each vehicle to meet constraint condition: the starting point of every route of i. must be grown place, terminal is client's point, does not consider backhaul; Ii. the dead weight capacity of every route must not be greater than the maximum authorized payload of carrier vehicle; Iii. each given client must be only once serviced; Iv. each client has the restriction of its serviced official hour window, if vehicle reaches client's point in advance, pays waiting cost, if vehicle delays to reach client's point, delays in payment expense;

With trucking costs, set up objective function, trucking costs is comprised of distribution vehicle fixed expense, trucking costs, vehicle waiting cost and deferred charges respectively, wherein, distribution vehicle fixed expense is comprised of vehicle depreciation expense and maintenance cost, this part expense is only relevant with the vehicle number of dispensing, the section situation that trucking costs is exercised with vehicle is relevant, comprise road section length, the crowded state in section, also comprise in whole section the traffic lights of process, the time cost that charge station produces and charge situation, client's freight demand will be sent within the scope of the stipulated time client conventionally, otherwise logistics center need to pay extra cost, the optimizing and scheduling vehicle model is here to set up for the destination node in class after classification, concrete model is as follows:

Dispensing road net model is described as:

$\left\{\begin{array}{c}G=(A\&cup;B\&cup;V,R)\\ A=\{a_1,a_2,...,a_n\}\\ B=\{b_1,b_2,...,b_m\}\\ V=\{v_1,v_2,...v_k\}\\ R=\left\{R_s\right\}\end{array}\right.$

Wherein A is grown place set in road network, and B is client node set in road network, and V is road intersection point set in road network, and they form the summit of network, and R is the oriented section collection in road;

The minimum model of transportation cost is: with the dynamic vehicle path planning problem of soft time window, objective function is:

$\min Z=\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{c}_k+\underset{g=0}{\overset{n}{\mathrm{\&Sigma;}}}\underset{h=0}{\overset{n}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{k}{\mathrm{\&Sigma;}}}{c}_{\mathrm{gh}}{x}_{\mathrm{ghk}}+p_1\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}\max (S_j-t_j,0)+p_2\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}\max (t_j-E_j,0)$

s.t.

①a _i＞b _j,i∈{1,2,...,m},j∈{1,2,...,n}；

2. c _kfixed cost for each car;

3. m round numbers, is distribution vehicle number, and a is parameter, 0 < a < 1, and constraint condition is more, and cargo handler is more complicated, and a is less;

4. c _ghfor section (v _g, v _h) transportation cost, relevant with the crowded state in section with the length in this section, and the quantity of the traffic lights comprising and charge station is relevant, c _gh=c _gh' d _gh+ c _gh' ' l _ghc wherein _gh' be section (v _g, v _h) expense of unit distance, d _ghfor section (v _g, v _h) distance, c _gh' ' be section (v _g, v _h) the traffic lights of process and the unit costs of charge station, l _ghthe traffic lights comprising for this section and the quantity of charge station;

5. x _ghk={ 0,1}, vehicle k is through section (v _g, v _h), x _ghk=1, otherwise be 0; G, h ∈ 0,1,2..., n}, k ∈ 1,2 ... m};

6.

y _jkif expression client puts the task of j and is completed by vehicle k, y _jk=1, otherwise y _jk=0; Q is the dead weight of vehicle;

7. [S _j, E _j] client puts the distribution time window requirement of j;

8. t _jfor vehicle arrives the moment that client puts j;

9. p ₁, p ₂be respectively early than arriving with exceeding the punishment cost coefficient that client puts j time window;

Wherein min represents minimum, and max represents maximum, and s.t. represents constraint condition;

A3. fuzzy cluster analysis: adopt the method combining based on quantitative test and qualitative analysis to carry out cluster, first according to the time window attribute of destination node, client's point is carried out to Preliminary division, and then in conjunction with quantitative method, by Customer Location, carry out client's cluster analysis, by fuzzy cluster analysis, the problem of large-scale vehicle path planning is demoted, change into small-scale combinatorial optimization problem and solve, fuzzy cluster analysis step is as follows:

3.1) obtaining of customer order information, comprises client's geographic position and demand;

3.2), according to the object of research, the index in close relations of selection and research object, first carries out quantitative classification;

3.3) standardization to data;

3.4) set up fuzzy similarity matrix;

3.5) foundation of fuzzy equivalence relation;

3.6) carry out cluster analysis, given different confidence level, asks R _λcut battle array, the λ that finds out R shows, each sample is classified as a class, along with the reduction of λ, by thin chap gradually and class;

A4. optimizing and scheduling vehicle heuritic approach design, by A3, will be large and change littlely compared with large dispensing impact point, now again the dispensing destination node in each class is adopted to optimizing and scheduling vehicle algorithm, can obtain providing and delivering result, step is as follows:

4.1) adopt natural number coding method, according to client in class, count out and produce initial population and carry out genetic coding, structure client point is as the chromosome in distribution route optimization problem solution vector,

4.2) calculate the population's fitness function of each distribution project, for objective function, get minimized combinatorial optimization problem, its fitness function will carry out suitable variation to objective function, to be converted into the situation of maximization, and guarantee that fitness value is non-negative,

4.3) method that optimized individual preservation method and roulette are selected to combine is deleted, duplicated chromosome, finally produces new population,

4.4) employing order bracketing method is implemented interlace operation, with crossover probability p _cpopulation is carried out to interlace operation, checks and whether meet constraint condition, produce new population,

4.5) adopt and repeatedly exchange variation method, with the Probability p that makes a variation _mpopulation is carried out to mutation operation, produces at random an exchange times L, check and whether meet constraint, form new population,

4.6) judge whether to meet stopping criterion, reach maximum iteration time or reach optimum solution requirement, meet the demands and stop, otherwise proceed to 4.3),

4.7) result of calculation is decoded,

4.8) select all decoded result of calculation, and compare and choose expense reckling.

Images