CN105976030A - Multi-agent-based platform scheduling intelligent sorting model structure - Google Patents

Abstract

The invention discloses a multi-agent-based platform scheduling intelligent sorting model structure. Steel material logistics distribution belongs to heavy-duty transportation in transportation and needs special handling equipment and large-scale transportation vehicles. In a distribution and loading process, the loading efficiency and sequence (platform scheduling) of vehicles are directly related to distribution aging. Platform scheduling is a multi-constraint combination optimization complex system. Reasonable ordering and optimization are essential. According to the invention, a multi-agent technology is used to research real-time vehicle scheduling decisions; a combination optimization complex system using a platform-based scheduling ordering system and multi-agent algorithm design as multiple constraints is constructed; a multi-agent technology provides a new solution for the research of real-time vehicle scheduling decisions to improve the efficiency of freight operations; the optimal solution is acquired; an auxiliary vehicle identification technology is used to carry out whole-process monitoring on distribution vehicles; efficient vehicle scheduling and precise handling are realized; and precise steel material distribution is realized.

Description

Structure based on multiple agent railway platform scheduling intelligent sequencing model

Technical field

The present invention relates to the structure of a kind of multiple agent railway platform dispatching patcher model, specifically a kind of according to supply chain and The model that Agent constructs.

Background technology

Before the present invention proposes, in supply chain, the factor that the transport of goods relates to is many, the level under various factors Index also can exist inconsistent, such as:

1) delivery period

Owing to customer order quantity is different with the cycle, when DPS system receives order, by automatic comparison on-hand inventory and goods Data, identify the down times of material simultaneously, and prompt service person pays close attention, and according to the adjustment of stock coefficient, intelligence is carried out Product distribution and production schedule arrangement.And early stage determines the delivery period of optimum.

2) vehicle fleet size

The vehicle number participating in transport operation affects the operational paradigm of whole railway platform equally, and different vehicles participates in handling to be needed Different optimizing scheduling strategies is provided.Many transport vehicle can be simultaneously in different warehouses in the case of other conditions permits Door loads and unloads, it is also possible to wait before same warehouse gate on the premise of more time-consuming.

3) type of vehicle

In view of loading in mixture, it is simply that if a chassis is the biggest, two client's things may be gathered together.Type of vehicle Different meaning dressing fortune kind just now is different from important, the most to a certain degree affects lay day and the logistics cost of goods, From scheduling, enter storehouse, the motility upstream of assembling is improved significantly, oversize vehicle loading capacity is big, and cargo handling is the longest, right The management of railway platform and vehicle scheduling bring certain impact.

4) emergency response

System is provided with warning function in the key link of each transport instruction, when reality performs to confirm instruction not In planned time internal feedback to system, early warning information can be displayed immediately on system billboard, the process initiative of abnormal conditions By force, fast response time.Such as when the required time of client is the most urgent, system needs to carry out corresponding vehicle loading time Advanced processing.

5) equipment failure rate

In the long-term operation of company, equipment inevitably breaks down, when the driving participating in loading and unloading occurs During fault, the handling task of opposite depot can be had a strong impact on, thus cause the scheduling of whole railway platform to adjust, waiting facilities event Barrier is repaired.

Summary of the invention

It is an object of the invention to, overcome drawback present in above-mentioned various factors, it is provided that a kind of multi-agent system (Multi-Agent System, MAS) is theoretical, constructs scheduling system based on railway platform excellent as the combination of multi-constraint condition Change complication system, it is achieved real-time vehicle scheduling decision-making is studied and provided new solution by multi-agent Technology, with Improve freight operation efficiency.

Realization the technical scheme is that, this structure based on multiple agent railway platform scheduling intelligent sequencing model, its It is characterised by: multi-agent system (Multi-Agent System, MAS) is the multiple intelligence mutual in an environment by The calculating system of energy body composition；Multi-agent system also can be used in the intelligent body solving to separate and single-layer system and be difficult to solve Problem certainly；Intelligent body passes through certain methods, and function, process, searching algorithm realizes, and the communication of the most each intelligent body is to pass through The communication shelving behavior realization each other coordinated with each other, is based on communication technology；

The present invention includes ordering system based on railway platform scheduling and multi-agent algorithm two parts of design: wherein, described Based on railway platform scheduling ordering system also include dis-tribution model, Agent modeling, vehicle scheduling；

Described dis-tribution model is logistics distribution model based on railway platform scheduling, and home-delivery center obtains the source of goods from upstream, passes through Flow processs such as receiving, store, assemble and dispatch buses, finally delivers to client by goods；The logistic distribution vehicle scheduling of any situation Problem can be expressed as object function and the Mathematical Planning mould of constraints two parts composition according to the method for mathematical modeling Type；The vehicle scheduling demand scheme that the present invention dispatches based on railway platform, its mathematical model can be expressed as: min or max z=f X (), according to the demand of actual items, can be analyzed and draw constraints: constraint, haulage vehicle that delivery period is relevant are correlated with Constraint that the relevant constraint of constraint, standard work force, the constraint that emergency response is relevant, human resources are correlated with, handling facilities are relevant Constraint；

Described Agent modeling refers to: process the clear superiority of multi-constraint condition objective optimisation problems based on MAS, this Bright being applied to by MAS in railway platform dispatching patcher, the main body (influence factor) of different levels can be come by the Agent of different levels It is described and expresses；The Agent of different levels connects each other, interacting has collectively constituted an actual dispatching patcher；

Described vehicle scheduling is the most key module designed for the railway platform scheduling target of intelligent sequencing, accordingly Influence factor relates generally to warehouse and two aspects of vehicle, in the environment in many warehouses, by as ready vehicle scheduling to suitable storehouse Ku Men improves the efficiency of loading and unloading of whole plant area and relates to more influence factor, such as busy extent, the goods in warehouse in this warehouse Storage level etc.；The factor of comprehensive each side, in conjunction with each warehouse gate of on-site, vehicle and corresponding warehouse Intelligent Matching；Right In the different sink door in a warehouse, store different products, under normal circumstances, each storehouse door correspondence one product (arrowband, plate Deng), owing to different vehicle needs the practical situation that loads in mixture, it is understood that there may be this warehouse gate is deposited and a small amount of is originally not belonging to this storehouse door Product, after customer order completes, arrange vehicle according to customer order to correspondence warehouse Ku Men go to take corresponding product；Root According to busy extent or the impact of handling horizontal factor of each warehouse gate, the transportation route that vehicle may be different arrives corresponding Warehouse gate completes entrucking work, the running of convenient vehicle below with the shortest time, improves the operational efficiency that system is overall；

Being constructed as follows of present invention railway platform scheduling mathematical model based on multiple agent:

$\begin{array}{c}{\mathrm{minZ}}_1=\underset{i=1}{\overset{G}{\Σ}}\underset{j=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{c}_{ij}{X}_{ijk}+\underset{i=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{P}_i{Y}_{ik}+f_1\underset{i=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{Y}_{ik}\max (a_i-t_i,0)+\\ f_2\underset{i=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{Y}_{ik}\max (t_i-b_i,0)\end{array}\mathrm{...}\left(3.1\right)$

$\begin{array}{c}s.t.\\ E_i\≤t_i\≤L_i\end{array}\mathrm{...}\left(3.2\right)$

$\underset{i=0}{\overset{G}{\Σ}}{n}_i{Y}_{ik}\≤q,k\∈\{1,2,\mathrm{...}V\}\mathrm{...}\left(3.3\right)$

$\underset{i=0}{\overset{G}{\Σ}}{Y}_{ik}\≤G,k\∈\{1,2,\mathrm{...}V\}\mathrm{...}\left(3.4\right)$

$\underset{i=0}{\overset{G}{\Σ}}{Y}_{ik}^{*}+\underset{i=0}{\overset{G}{\Σ}}{Y}_{ik}=G,K\∈\{1,2,\mathrm{...}V\}\mathrm{...}\left(3.5\right)$

The parameter related in above-mentioned model is explained as follows:

The set of G: warehouse gate

V: participate in the set of distribution vehicle

t_i: vehicle arrives the used time of warehouse i

L_i: the time (plant area's work in 24 hours, negligible) that warehouse i services the latest

E_i: the time (plant area's work in 24 hours, negligible) that warehouse i services the earliest

a_i: warehouse can provide the time (shortest time that vehicle needs to wait for when arriving warehouse) that entrucking services the earliest

b_i: warehouse can provide the time (maximum duration that vehicle needs to wait for when arriving warehouse) that entrucking services the latest

f₁: vehicle arrives warehouse and provides the penalty coefficient of service time early than warehouse

f₂: vehicle arrives warehouse and is later than the penalty coefficient of warehouse offer service time

c_ij: the vehicle time-consuming (scope is 5-10min) between warehouse i and j

P_i: warehouse gate i installs the time-consuming of a car required product

The capacity of q: vehicle

n_i: warehouse gate i is supplied to the density of cargo of car

The algorithm of present invention railway platform based on multiple agent scheduling model uses two kinds of wide heuristic calculations sent out of application Method, the hybrid algorithm that genetic algorithm combines with tabu search algorithm realizes, and genetic algorithm (GA) is according to Darwinian nature Select and theory of heredity, by during biological evolution survival of the fittest rule with same a group chromosome with enter information exchange combine Intelligent algorithm；The performance of genetic algorithm is largely dependent upon intersects and the operation of variation, and this depends on solving concentration How sample drawn solution；Tabu search algorithm (TA) most important thought is some of the corresponding locally optimal solution searched for of labelling Object, and in further iterative search, avoid these objects rather than absolute prohibition circulation as far as possible, thus ensure difference The exploration having efficient search approach；

After mixing, the main policies of algorithm is exactly: first passes through genetic algorithm and carries out global search, uses natural number to institute Having warehouse and adjustable vehicle to encode, the path that with the carrying capacity of vehicle, the availability in each warehouse is carried out the overall situation is excellent Change；Then TABU search is used with certain probability, the individuality in population to be carried out Local Search, namely for same car All warehouses are carried out local transportation route optimization；First the present invention arranges initial population, then simulates biological evolution, initially Between population, generation selects, makes a variation, intersects as a new generation population, a new generation population is cooked only chess game optimization, stays Individuality, through the heredity in too much generation, eventually forms the individuality that fitness is best；

Hybrid algorithm solution procedure of the present invention is as follows:

(1) warehouse gate directly arranges natural number coding

First can design the different natural number arrangement not sticking with paste repetition of multiple 1-G, the arrangement of this natural number just constitutes one Individual.Successively warehouse gate can be inserted in travel route according to constraints, such as, call two cars and arrive the warehouse of 4 Point, it is assumed that the driving path of car is 1234, traversal is numbered 1 the most successively, and the warehouse point of 2,3,4, first by first warehouse point Being inserted in travel route, if meeting above-mentioned institute Prescribed Properties, inserting second warehouse point, proceeding if meeting, when During beyond the carrying capacity of vehicle, call second car.

(2) initial population is set

It is randomly generated this G of 1-G mutual unduplicated natural number arrangement, i.e. generates body one by one.Assume initial population Number is N, then produce the most different N number of individuality.

(3) fitness evaluation standard determines

Because optimization aim is to minimize, and the fitness of genetic algorithm represents the individuality that adaptation ability is the strongest, therefore can Fitness is represented with the inverse of object function.

F=1/Z₁

(4) operation is replicated

The design completes the operation to the population at individual survival of the fittest by retaining optimized individual and gambling dish strategy.

(5) intersection operation

Exchange the individual Partial Fragment of two parents by certain probability and complete intersection operation, common crossover operator There are partial intersection operator, ordered crossover operator, recycling cross operator and class OX operator etc..

The design uses ordered crossover operator, such as one car completes to need through 1 according to shipping work, 2,3,4,5,6, 7 these seven warehouse gates load corresponding product, existing two kinds of different route or travel by vehicle: R₁=1234567, R₂= 3425167, R₁Represent that vehicle sequentially passes through No. 1 door, No. 2 door ..., No. 7 doors, R₂Represent vehicle sequentially pass through No. 3 doors, No. 4 Door ..., No. 7 doors, therefrom select a matching section,

According to the mapping relations of matching section, position corresponding outside matching section region is labeled as A, it may be assumed that

Shifted matching section is to original position again, and the reserved position identical with matching section space later, is labeled as A, it may be assumed that

Finally the matching section of two sequences is exchanged with each other, obtains two new offsprings, it may be assumed that

(6) mutation operation

Mutation operation embodies the thought of nature gene mutation, and common mutation operator has reverse variation, exchange mutation Make a variation with inserting.

The design uses reverse variation, randomly chooses two in a sequence and clicks on line position exchange, by 2 interior words Symbol inverted sequence is inserted in former sequence.The such as travel route for a car is R₁=1234567, by second position and the 5th Individual position carries out reversing variation, and the sequence obtained is R₁ ^*=1543267.

(7) utilize TS algorithm that current solution is improved

Tabu search algorithm uses

Current solution is evaluated

Wherein T (i) represents that vehicle arrives the time that warehouse point needs, E (i) represent vehicle warehouse point handling goods and etc. The time treated, W (i) represents the capacity of carriage of vehicle, and p represents penalty coefficient.

The execution step of tabu search algorithm is as follows:

Step one: selected initial solution (having genetic algorithm to obtain) x^now, make taboo list

Step 2: if meeting stop criterion, go to step four；Otherwise, at x^nowField N (x^nowSelect in) to meet and avoid wanting Candidate Set can_N (the x asked^now), perform step 3.

Step 3: at can_N (x^now) select one group of evaluation of estimate optimal solution x^best, make x^now=x^best, update taboo list, turn Step 2.

Step 4: output operation result.

(8) stop criterion

Because the dynamic factor affecting vehicle dispatch system stability is more, ageing to decision system wants height, originally sets Meter is intended using the stop criterion specifying algebraically step number to terminate；

The algorithm design of present invention railway platform based on multiple agent scheduling model is as follows:

Input parameter:

Population scale N, represents the running route of different initial vehicle

Evolutionary generation T, represents the algebraically that population is to be multiplied

Crossover probability P_c

Mutation probability P_m

Penalty coefficient p

Output result:

Vehicle scheduling route and optimization target values

Algorithm main body:

Multiple different initial population P (0) are produced according to vehicle park coupling matrix,

Current algebraically is t=0；

Calculate the fitness of initial population

Specifically, described Agent modeling includes: client Agent, warehouse Agent, vehicle Agent, road network Agent, work Power Agent, order processing Agent, emergency response Agent, vehicle scheduling Agent, handling scheduling Agent and total activation Agent, Wherein:

1) client Agent comprises customer name, client codes, shipment schedule, requires delivery time, the class of required bale packing The information such as type and quantity；

2) warehouse Agent comprises the information such as bale packing type, bale packing number, stock coefficient；

3) vehicle Agent with driver bind, comprise vehicle location state (vehicle wait, just in entrucking, install vehicle), car Useful load, car plate, driver, car such as criticize at the information；

4) road network Agent comprise the different Ku Men of warehouse point, warehouse compartment and they between switch time-consuming information；

5) labour force Agent comprises single packaging car standard work force, stevedore's information (total number of persons, is distributed number, treated point Join number) etc. information；

6) order processing Agent can be existing according to the order requirements entry evaluation of the goods storage present situation in warehouse and client Can car loading meet the demand of client, and final assessment report informs the total activation Agent of the superiors, total activation Agent determines to take generate loading and transporting plan list (plan odd numbers is as order ID) or arrange complementary goods goods according to assessment report The production schedule of thing；

7) emergency response Agent belongs to the intelligent body that one direction relies on, for the emergency in processing system；Act on In the key link of each transport instruction of system, when reality perform confirm instruction not in planned time internal feedback to total activation Agent, early warning information can be displayed immediately on system billboard, and the process initiative of abnormal conditions is strong；

Total activation, according to the loading and transporting plan generated, calculates shipment shipment line and transport mileage (is used according to theory when way Between calculate on the time of making the product)；Loading and transporting plan is supplied to vehicle scheduling Agent, vehicle scheduling Agent and completes vehicle scheduling；? March into the arena the time according to the result of calculation of the vehicle scheduling Agent vehicle that falls back afterwards；

8) vehicle scheduling Agent is the key of whole dispatching patcher, when customer order generation and system are not anticipated In the case of Wai, total activation Agent extracts the human resource information loading scheduling Agent, and it is handed down in the lump vehicle with order Agent, vehicle scheduling Agent are according to the resource such as order and vehicle in scheduling, to entrucking order rational management, are system whole efficiency Reach the highest；

9) handling scheduling Agent is responsible for the transfer of labour force, reasonably arranges worker-hours and place, allows workman exist The correct time occurs in correct doorway, warehouse handling goods, calculates the entrucking used time, is returned to higher level Agent with regard to result of calculation, Make vehicle minimum in the warehouse time of staying；

Total activation Agent is the brain of whole system, is responsible for synchronization and the management of whole system, coordinates each Agent, make Cooperate between them, it is ensured that the operation that whole system is orderly.

Specifically, in described vehicle scheduling module, D represents warehouse,Represent the weight of different affecting factors, Δ_ijRepresent the I warehouse evaluation situation to jth influence factor, warehouse handling capacity represents the entrucking level of Current warehouse, at real-time shape The used time of unit product is loaded under state；Vehicle time-consumingly represent to warehouse empty-car (not loading goods) dispatch a car in field a little to The used time of opposite depot door；Articles from the storeroom reserves represent the maximum of the corresponding goods that Current warehouse door can provide, and decide Whether meet the demand of vehicle；Vehicle rate of actual loading represents that the loading condition of Current vehicle, residue are available for the space of product loading also Have how many, affect vehicle and select suitable warehouse gate.Weight represents each factor significance level to vehicle scheduling.

The best match algorithm of vehicle initial rest position is (with warehouse gate D₀₁As a example by):

Step one: the index of standardization coupling matrix, becomes to can be used to the same amount compared by different unit conversions；

Δ₁₀₁=P_{Operation number}*0.6+P_{Driving is away from Ku Men position}*0.2+P_{Waiting time}*0.2

Δ₂₀₁=P_{Arrive the warehouse gate time}

Δ₃₀₁=P_{Corresponding product amount}*0.9+P_{Other products}*0.1

Δ₄₀₁=P_{Vehicle is real to be carried}/P_{Vehicular load}

To Δ₁₀₁、Δ₂₀₁、Δ₃₀₁、Δ₄₀₁、Δ₅₀₁、Δ₆₀₁Carry out unit-normalization

Step 2: the weight of each influence factor of normalized；

WeighValue, and make its meet:

${\∂}_1+{\∂}_2+{\∂}_3+{\∂}_4+{\∂}_5+{\∂}_6=1$

Step 3: calculate the comprehensive evaluation value that warehouse is stopped；

The computing formula of comprehensive evaluation value is:

$S_{01}={\mathrm{\Δ}}_{101}*{\∂}_1+{\mathrm{\Δ}}_{201}*{\∂}_2+{\mathrm{\Δ}}_{301}*{\∂}_3+{\mathrm{\Δ}}_{401}*{\∂}_4+{\mathrm{\Δ}}_{501}*{\∂}_5+{\mathrm{\Δ}}_{601}*{\∂}_6$

Step 4: when there being one or more vehicle to simultaneously participate in transport, considers the warehouse conduct that evaluation of estimate is the highest The initial optimal warehouse that distribution vehicle is stopped.

Specifically, described railway platform scheduling mathematical model Chinese style (3.1) based on multiple agent is as whole model Optimization aim, represents that vehicle finally completes the shortest time of task through sequence, and constraints (3.2) represents that vehicle arrives warehouse Time can not early than warehouse start provide service time, it is impossible to be later than warehouse close service time, little for plant area 24 Time work, this condition negligible；Formula (3.3) represents that the final carrying capacity of vehicle not can exceed that the carrying capacity of vehicle itself；Formula (3.4), (3.5) represent that a vehicle once and at most can only once can not surpass through same warehouse and the sum through warehouse Cross the sum in warehouse.

It is an advantage of the current invention that: whole scheme be designed with multiple theory and algorithm, and according to the application of event Demand is made that improvement to original algorithm, comprehensively embodies the advantage of three aspects.

1. the advantage of multi-agent theory.

Use multi-agent theory to build ordering system, the advantage of multi-agent theory can be given full play to.

Autonomy: each Agent can singly complete the work of data this Agent task, needs the number of other Agent According to time, can communicate acquisition by setting up with other Agent.

Pre-activity: each Agent can in time predict what this Agent was responsible for according to the scheduling of resource of whole system Resource status, the most more new resources, it is to avoid unnecessary is time-consuming.

Extensibility: along with the mode of production and the change of demand, system needs to make adjustment according to actual needs, the most intelligent Can be easy to extend new Agent on the basis of original structure, it is not necessary to do big change.

Social: problem can well be divided by multiple agent according to practical problem, by module low for coupling Give different Agent to go to process, clear workflow, improve work efficiency, improve systematic function.

2. the advantage of hybrid algorithm

The traditional approach processing vehicle dispatching problem is to use exact algorithm to solve, and including branch-bound algorithm, moves State laws of planning etc., these algorithms all receive the invariance of scale and the constraints being limited to problem.Processing extensive changeable constraint Under the conditions of problem be generally to use heuritic approach, common heuritic approach has genetic algorithm, tabu search algorithm, ant colony Algorithm etc., but there is intrinsic defect in single algorithm.

Although genetic algorithm has considerable flexibility and robustness, it is suitable for solving of extensive problem, but it exists too early Ground convergence and the defect of local search ability difference.And the hybrid algorithm adding tabu search algorithm on the basis of genetic algorithm can This defect is well compensate for the advantage of releasing tabu search algorithm local optimum.The thought of whole hybrid algorithm is led to exactly Cross the initial solution of genetic algorithm structure tabu search algorithm, as the neighborhood of TABU search module, entered by tabu search algorithm One-step optimization is individual, promotes the quality of whole population, makes more excellent individuality occur in minimum iterative steps.

3. the breakthrough that application is actual

Conventional vehicle scheduling based on heuritic approach is all applied in the logistics company dispensing to customer demand, it is considered to Be all the vehicle scheduling from logistics center to client's hands, and consider that restrictive condition is the most single and there is the biggest reason Opinion property.The design, from the actual demand of plant area, breaks through conventional logistics condition outside the venue, considers physical presence in plant area Constraints, use heuritic approach to complete plant area's interior vehicle and commute the scheduling problem in different warehouse, from original from Focus on scattered one-to-many dis-tribution model be changed into current from warehouse to the many-one logistic pattern of this vehicle, according to actual Original algorithm is carried out bold innovation, to solve actual logistical problem.

The design, based on actual items background, utilizes MAS multiple agent model buildings vehicle scheduling based on railway platform system System model, and on the basis of abundant analysis demand, devise and operate the shortest time vehicle scheduling mathematics as target with vehicle Model, has wherein applied to heuritic approach widely used on Logistics Scheduling Problem: genetic algorithm and TABU search are calculated Method, devises genetic algorithms based on both on the basis of advantage both extracting, and makes vehicle scheduling reach optimum.

Accompanying drawing explanation

Fig. 1 is logistics distribution model model schematic based on railway platform scheduling；

Fig. 2 is intelligent scheduling order models schematic diagram based on MAS；

Fig. 3 is vehicle scheduling order models schematic diagram；

Fig. 4 is the flow chart of vehicle scheduling matching algorithm；

Fig. 5 is that storehouse model analyzes plane graph；

Fig. 6 hybrid algorithm model schematic；

Fig. 7 hybrid algorithm algorithm flow schematic diagram.

Detailed description of the invention

The present invention is described in further detail with specific embodiment below in conjunction with the accompanying drawings.

The present invention provides a kind of multi-agent system Multi-Agent System, MAS theoretical, constructs tune based on railway platform Degree ordering system is as the Combinatorial Optimization complication system of multi-constraint condition, it is achieved multi-agent Technology is to real-time vehicle scheduling certainly Plan carries out studying and provides new solution, to improve freight operation efficiency.

The Disciplinary Frontiers that multiple agent (Multi-Agent System, MAS) is studied as distributed artificial intelligence and One of important method holding intelligent decision, the feature possessed because of it and the most important theories being acknowledged as study of various complication system Model.Scheduling system based on railway platform, as the Combinatorial Optimization complication system of multi-constraint condition, introduces multi-agent Technology Real-time vehicle scheduling decision-making is studied and provides new solution, possess good theory significance and practical value.

Multi-agent system (Multi-Agent System, MAS) is the multiple intelligence mutual in an environment by The calculating system of energy body composition.Multi-agent system also can be used in the intelligent body solving to separate and single-layer system is difficult to solve Problem.Intelligent body can be by certain methods, function, process, and searching algorithm or reinforcement study realize.Agent has autonomy Property, reactivity, pre-activity, social feature.MAS itself has collaborative, concurrency, vigorousness, expansibility and divides The features such as cloth solves so that it has natural superiority in terms of processing dynamic complex system.MAS is several the most autonomous or oneself The Agent controlled is according to certain agreement and certain language, and can communicate the one of a complex problem solving with other Agent Individual system.There is the features such as collaborative, concurrency, vigorousness, expansibility, distributivity.In MAS system, each intelligent body is logical Letter is by the communication shelving behavior realization each other coordinated with each other, is based on communication technology.Autonomy, dynamic, The feature such as distributivity and harmony is the advantage that the MAS of field of artificial intelligence research possesses, and the cooperation between MAS is also commonly used for Optimize allocation of resources and distributed problem cooperation solving.

The present invention includes ordering system based on railway platform scheduling and multi-agent algorithm two parts of design: wherein, described Based on railway platform scheduling ordering system also include dis-tribution model, Agent modeling, vehicle scheduling；

A, dis-tribution model: logistics distribution model based on railway platform scheduling is substantially as it is shown in figure 1, home-delivery center obtains from upstream The source of goods, through flow processs such as receiving, store, assemble and dispatch buses, finally delivers to client, vehicle scheduling based on railway platform by goods The main vehicle being responsible for on-site enters the process that car is freighted and dispatched from the factory.

The logistic distribution vehicle scheduling problem of any situation can be expressed as target letter according to the method for mathematical modeling The mathematical programming model that number and constraints two parts are constituted.According to the vehicle scheduling demand scheme dispatched based on railway platform, its number Learn model can be expressed as:

Min or max z=f (x)

According to the demand of actual items, can analyze and draw constraints: constraint that delivery period is relevant, haulage vehicle phase Constraint that the relevant constraint of the constraint closed, standard work force, the constraint that emergency response is relevant, human resources are correlated with, handling facilities phase The constraint etc. closed.

B, Agent model: process the clear superiority of multi-constraint condition objective optimisation problems based on MAS, and the design should by MAS Using in railway platform dispatching patcher, the main body (influence factor) of different levels can be described by the Agent of different levels And expression.The Agent of different levels connects each other, interacting has collectively constituted an actual dispatching patcher.

The thought modeled according to supply chain and Agent, whole railway platform intelligent system model based on MAS is as shown in Figure 2.

1, client Agent comprises customer name, client codes, shipment schedule, requires delivery time, the type of required bale packing With information such as quantity.

2, warehouse Agent comprises the information such as bale packing type, bale packing number, stock coefficient.

3, vehicle Agent with driver bind, comprise vehicle location state (vehicle wait, just in entrucking, install vehicle), car Useful load, car plate, driver, car such as criticize at the information.

4, road network Agent comprise the different Ku Men of warehouse point, warehouse compartment and they between switch time-consuming information.

5, labour force Agent comprises single packaging car standard work force, stevedore's information (total number of persons, is distributed number, treated point Join number) etc. information.

Order processing Agent can store present situation and the existing goods of order requirements entry evaluation of client according to the goods in warehouse Can thing amount meet the demand of client, and final assessment report informs the total activation Agent of the superiors, total activation Agent determines to take generate loading and transporting plan list (plan odd numbers is as order ID) or arrange complementary goods goods according to assessment report The production schedule of thing.

Emergency response Agent belongs to the intelligent body that one direction relies on, for the emergency in processing system.Act on and be System each transport instruction key link on, when reality perform confirm instruction not in planned time internal feedback to total activation Agent, early warning information can be displayed immediately on system billboard, and the process initiative of abnormal conditions is strong.

Total activation, according to the loading and transporting plan generated, calculates shipment shipment line and transport mileage (is used according to theory when way Between calculate on the time of making the product).Loading and transporting plan is supplied to vehicle scheduling Agent, vehicle scheduling Agent and completes vehicle scheduling.? March into the arena the time according to the result of calculation of the vehicle scheduling Agent vehicle that falls back afterwards.

Vehicle scheduling Agent is the key of whole dispatching patcher, when customer order generation and system do not occur accident In the case of, total activation Agent extracts the human resource information loading scheduling Agent, it is handed down to order vehicle in the lump and adjusts Degree Agent, vehicle scheduling Agent, according to the resource such as order and vehicle, to entrucking order rational management, are that system whole efficiency reaches To the highest.

Handling scheduling Agent is responsible for the transfer of labour force, reasonably arranges worker-hours and place, allows workman just The true time occurs in correct doorway, warehouse handling goods, calculates the entrucking used time, is returned to higher level Agent with regard to result of calculation, allows Vehicle is minimum in the warehouse time of staying.

Wagon loading plan is time-consumingly number * standard work force standard work force=bale packing

Total activation Agent is the brain of whole system, is responsible for synchronization and the management of whole system, coordinates each Agent, make Cooperate between them, it is ensured that the operation that whole system is orderly.

C, vehicle scheduling: for the target of railway platform scheduling intelligent sequencing, module the most key in whole system is vehicle Scheduling Agent module, corresponding influence factor relates generally to warehouse and two aspects of vehicle, and relevant influence factor's graph of a relation is such as Shown in Fig. 3.

In the environment in many warehouses, generation fortune vehicle scheduling is improved to suitable warehouse gate the efficiency of loading and unloading of whole plant area Relate to more influence factor, such as busy extent, the goods storage level etc. in warehouse in this warehouse.The factor of comprehensive each side, knot Closing 13 warehouse gates of on-site, vehicle is as shown in table 1 with corresponding warehouse coupling matrix.

Table 1 vehicle mates matrix with warehouse

In table 1, D represents warehouse,Represent the weight of different affecting factors, Δ_ijRepresent i-th warehouse jth is affected because of The evaluation situation of element, warehouse handling capacity represents the entrucking level of Current warehouse, loads the use of unit product under real-time status Time；Vehicle time-consumingly represents that to warehouse empty-car (not loading goods) is dispatched a car a little to the used time of opposite depot door in field；Storehouse Storehouse goods reserves represent the maximum of the corresponding goods that Current warehouse door can provide, and decide the demand whether meeting vehicle； Vehicle rate of actual loading represents the loading condition of Current vehicle, and residue is available for the space of product loading and also has how many, affects vehicle choosing Select suitable warehouse gate.Weight represents each factor significance level to vehicle scheduling.

The best match algorithm of vehicle initial rest position is (with warehouse gate D₀₁As a example by):

Step one: the index of standardization coupling matrix, becomes to can be used to the same amount compared by different unit conversions；

Δ₁₀₁=P_{Operation number}*0.6+P_{Driving is away from Ku Men position}*0.2+P_{Waiting time}*0.2

Δ₂₀₁=P_{Arrive the warehouse gate time}

Δ₃₀₁=P_{Corresponding product amount}*0.9+P_{Other products}*0.1

Δ₄₀₁=P_{Vehicle is real to be carried}/P_{Vehicular load}

To Δ₁₀₁、Δ₂₀₁、Δ₃₀₁、Δ₄₀₁、Δ₅₀₁、Δ₆₀₁Carry out unit-normalization

Step 2: the weight of each influence factor of normalized；

WeighValue, and make its meet:

${\∂}_1+{\∂}_2+{\∂}_3+{\∂}_4+{\∂}_5+{\∂}_6=1$

Step 3: calculate the comprehensive evaluation value that warehouse is stopped；

The computing formula of comprehensive evaluation value is:

$S_{01}={\mathrm{\Δ}}_{101}*{\∂}_1+{\mathrm{\Δ}}_{201}*{\∂}_2+{\mathrm{\Δ}}_{301}*{\∂}_3+{\mathrm{\Δ}}_{401}*{\∂}_4+{\mathrm{\Δ}}_{501}*{\∂}_5+{\mathrm{\Δ}}_{601}*{\∂}_6$

Step 4: when there being one or more vehicle to simultaneously participate in transport, considers the warehouse conduct that evaluation of estimate is the highest The initial optimal warehouse that distribution vehicle is stopped.The flow chart of matching algorithm is as shown in Figure 4.

There is a warehouse can not meet distribution vehicle fortune owing to there is an articles from the storeroom shortage or other reasons The situation of defeated demand, vehicle needs other warehouse gates to extract the goods needing to load, and extracts corresponding arriving other warehouses During goods, owing to loading and unloading the restriction of level, after needing vehicle below to wait that preceding vehicle is left the when that multiple vehicle occurring Could load, waste the time, reduce conevying efficiency.

Plant area's storehouse model: as a example by warehouse 1 (other two storehouses are with warehouse 1), storehouse model is analyzed, warehouse The plane graph of 1 is as shown in Figure 5.

Corresponding six the storehouse doors in warehouse 1, wherein No. 1 door and No. 7 doors share handling facilities, and No. 2 doors and No. 6 doors share handling and set Standby, No. 3 doors and No. 5 doors share handling facilities, for the different sink door in a warehouse, store different products, under normal circumstances, Each storehouse door correspondence one product (arrowband, plate etc.), the practical situation loaded in mixture due to different vehicle needs, it is understood that there may be this storehouse Ku Men deposits a small amount of product being originally not belonging to this storehouse door, after customer order completes, arranges vehicle to arrive according to customer order Corresponding warehouse Ku Men goes to take corresponding product.Busy extent according to each warehouse gate or the impact of handling horizontal factor, The possible different transportation route of vehicle arrives corresponding warehouse gate and completes entrucking work with the shortest time, convenient vehicle the most below Running, improves the operational efficiency that system is overall.

According to the target design scheduling model that the used time is optimum, the parameter related in model is explained as follows:

The set of G: warehouse gate

V: participate in the set of distribution vehicle

t_i: vehicle arrives the used time of warehouse i

L_i: the time (plant area's work in 24 hours, negligible) that warehouse i services the latest

E_i: the time (plant area's work in 24 hours, negligible) that warehouse i services the earliest

a_i: warehouse can provide the time (shortest time that vehicle needs to wait for when arriving warehouse) that entrucking services the earliest

b_i: warehouse can provide the time (maximum duration that vehicle needs to wait for when arriving warehouse) that entrucking services the latest

f₁: vehicle arrives warehouse and provides the penalty coefficient of service time early than warehouse

f₂: vehicle arrives warehouse and is later than the penalty coefficient of warehouse offer service time

c_ij: the vehicle time-consuming (scope is 5-10min) between warehouse i and j

P_i: warehouse gate i installs the time-consuming of a car required product

The capacity of q: vehicle

n_i: warehouse gate i is supplied to the density of cargo of car

Mathematical model:

$\begin{array}{c}{\mathrm{minZ}}_1=\underset{i=1}{\overset{G}{\Σ}}\underset{j=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{c}_{ij}{X}_{ijk}+\underset{i=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{P}_i{Y}_{ik}+f_1\underset{i=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{Y}_{ik}\max (a_i-t_i,0)+\\ f_2\underset{i=0}{\overset{G}{\Σ}}\underset{k=1}{\overset{V}{\Σ}}{Y}_{ik}\max (t_i-b_i,0)\end{array}\mathrm{...}\left(3.1\right)$

$\begin{array}{c}s.t.\\ E_i\≤t_i\≤L_i\end{array}\mathrm{...}\left(3.2\right)$

$\underset{i=0}{\overset{G}{\Σ}}{n}_i{Y}_{ik}\≤q,k\∈\{1,2,\mathrm{...}V\}\mathrm{...}\left(3.3\right)$

$\underset{i=0}{\overset{G}{\Σ}}{Y}_{ik}\≤G,k\∈\{1,2,\mathrm{...}V\}\mathrm{...}\left(3.4\right)$

$\underset{i=0}{\overset{G}{\Σ}}{Y}_{ik}^{*}+\underset{i=0}{\overset{G}{\Σ}}{Y}_{ik}=G,K\∈\{1,2,\mathrm{...}V\}\mathrm{...}\left(3.5\right)$

Formula (3.1), as the optimization aim of whole model, represents that vehicle finally completes the shortest time of task through sequence, Constraints (3.2) represents that the time in vehicle arrival warehouse can not start to provide early than warehouse the time of service, it is impossible to be later than storehouse The time of service is closed in storehouse, for plant area's work in 24 hours, this condition negligible；Formula (3.3) represents the carrying capacity that vehicle is final Not can exceed that the carrying capacity of vehicle itself；It is same that formula (3.4), (3.5) represent that a vehicle once and at most can only once pass through One warehouse and the sum through warehouse not can exceed that the sum in warehouse.

Present invention construction algorithm based on multiple agent railway platform scheduling model designs:

Hybrid algorithm:

In the solving of vehicle scheduling relevant issues, generally take is heuritic approach, and the design uses two kinds of application The wide heuritic approach sent out, the hybrid algorithm that genetic algorithm combines with tabu search algorithm realizes, and mixed algorithm both had There is the advantage of overall importance of genetic algorithm, have the ability of climbing the mountain with tabu search algorithm, precocity can be avoided largely, carry High-performance.

Genetic algorithm (GA) is according to Darwinian natural selection and theory of heredity, by raw for person suitable during biological evolution Deposit rule and same a group chromosome with entering the information intelligent algorithm that combines of exchange.

The performance of genetic algorithm is largely dependent upon intersects and the operation of variation, and this depends on how solving concentration Sample drawn solution.

Tabu search algorithm (TA) most important thought is some objects of the locally optimal solution that labelling correspondence has been searched for, and In further iterative search, avoid these objects (rather than absolute prohibition circulation) as far as possible, thus ensure different effective The exploration of search approach.

After mixing, the main policies of algorithm is exactly: first passes through genetic algorithm and carries out global search, uses natural number to institute Having warehouse and adjustable vehicle to encode, the path that with the carrying capacity of vehicle, the availability in each warehouse is carried out the overall situation is excellent Change；Then TABU search is used with certain probability, the individuality in population to be carried out Local Search, namely for same car All warehouses are carried out local transportation route optimization.The layout strategy of hybrid algorithm is as shown in Figure 6.

Just start only initial population (ten blue Magen Davids), then simulated biological evolution, produced between initial population Raw selection (adaptable individual reservation, leave six blue Magen Davids), variation (two corner stars occur), (appearance that intersects One red Magen David) as a new generation population, a new generation population is cooked only chess game optimization, the individuality stayed, through too much The heredity in generation, eventually forms the best individuality of fitness (a red Magen David and a blue five-pointed star).

Hybrid algorithm solves:

(1) warehouse gate directly arranges natural number coding

First can design the different natural number arrangement not sticking with paste repetition of multiple 1-G, the arrangement of this natural number just constitutes one Individual.Successively warehouse gate can be inserted in travel route according to constraints, such as, call two cars and arrive the warehouse of 4 Point, it is assumed that the driving path of car is 1234, traversal is numbered 1 the most successively, and the warehouse point of 2,3,4, first by first warehouse point Being inserted in travel route, if meeting above-mentioned institute Prescribed Properties, inserting second warehouse point, proceeding if meeting, when During beyond the carrying capacity of vehicle, call second car.

(2) initial population is set

It is randomly generated this G of 1-G mutual unduplicated natural number arrangement, i.e. generates body one by one.Assume initial population Number is N, then produce the most different N number of individuality.

(3) fitness evaluation standard determines

Because optimization aim is to minimize, and the fitness of genetic algorithm represents the individuality that adaptation ability is the strongest, therefore can Fitness is represented with the inverse of object function.

F=1/Z₁

(4) operation is replicated

The design completes the operation to the population at individual survival of the fittest by retaining optimized individual and gambling dish strategy.

(5) intersection operation

Exchange the individual Partial Fragment of two parents by certain probability and complete intersection operation, common crossover operator There are partial intersection operator, ordered crossover operator, recycling cross operator and class OX operator etc..

The design uses ordered crossover operator, such as one car completes to need through 1 according to shipping work, 2,3,4,5,6, 7 these seven warehouse gates load corresponding product, existing two kinds of different route or travel by vehicle: R₁=1234567, R₂= 3425167, R₁Represent that vehicle sequentially passes through No. 1 door, No. 2 door ..., No. 7 doors, R₂Represent vehicle sequentially pass through No. 3 doors, No. 4 Door ..., No. 7 doors, therefrom select a matching section,

According to the mapping relations of matching section, position corresponding outside matching section region is labeled as A, it may be assumed that

Shifted matching section is to original position again, and the reserved position identical with matching section space later, is labeled as A, it may be assumed that

Finally the matching section of two sequences is exchanged with each other, obtains two new offsprings, it may be assumed that

(6) mutation operation

Mutation operation embodies the thought of nature gene mutation, and common mutation operator has reverse variation, exchange mutation Make a variation with inserting.

The design uses reverse variation, randomly chooses two in a sequence and clicks on line position exchange, by 2 interior words Symbol inverted sequence is inserted in former sequence.The such as travel route for a car is R₁=1234567, by second position and the 5th Individual position carries out reversing variation, and the sequence obtained is R₁ ^*=1543267.

(7) utilize TS algorithm that current solution is improved

Tabu search algorithm uses

Current solution is evaluated

Wherein T (i) represents that vehicle arrives the time that warehouse point needs, E (i) represent vehicle at warehouse point handling goods and The time waited, W (i) represents the capacity of carriage of vehicle, and p represents penalty coefficient.

The execution step of tabu search algorithm is as follows:

Step one: selected initial solution (having genetic algorithm to obtain) x^now, make taboo list

Step 2: if meeting stop criterion, go to step four；Otherwise, at x^nowField N (x^nowSelect in) to meet and avoid wanting Candidate Set can_N (the x asked^now), perform step 3.

Step 3: at can_N (x^now) select one group of evaluation of estimate optimal solution x^best, make x^now=x^best, update taboo list, turn Step 2.

Step 4: output operation result.

(8) stop criterion

Because the dynamic factor affecting vehicle dispatch system stability is more, ageing to decision system wants height, originally sets Meter is intended using the stop criterion specifying algebraically step number to terminate.

Algorithm flow: combine the flow chart of hybrid algorithm of genetic algorithm and urgent searching algorithm as shown in Figure 7.

Algorithm designs:

Input parameter:

Population scale N, represents the running route of different initial vehicle

Evolutionary generation T, represents the algebraically that population is to be multiplied

Crossover probability P_c

Mutation probability P_m

Penalty coefficient p

Output result:

Vehicle scheduling route and optimization target values

Algorithm main body:

Producing multiple different initial population P (0) according to vehicle park coupling matrix, current algebraically is t=0；

Calculate the fitness of initial population

Innovation point:

Whole scheme be designed with multiple theory and algorithm, and according to the application demand of event, original algorithm is done Go out improvement, comprehensively embody the advantage of three aspects.

1, the advantage of multi-agent theory.

Use multi-agent theory to build ordering system, the advantage of multi-agent theory can be given full play to.

Autonomy: each Agent can singly complete the work of data this Agent task, needs the number of other Agent According to time, can communicate acquisition by setting up with other Agent.

Pre-activity: each Agent can in time predict what this Agent was responsible for according to the scheduling of resource of whole system Resource status, the most more new resources, it is to avoid unnecessary is time-consuming.

Extensibility: along with the mode of production and the change of demand, system needs to make adjustment according to actual needs, the most intelligent Can be easy to extend new Agent on the basis of original structure, it is not necessary to do big change.

Social: problem can well be divided by multiple agent according to practical problem, by module low for coupling Give different Agent to go to process, clear workflow, improve work efficiency, improve systematic function.

2, the advantage of hybrid algorithm

The traditional approach processing vehicle dispatching problem is to use exact algorithm to solve, and including branch-bound algorithm, moves State laws of planning etc., these algorithms all receive the invariance of scale and the constraints being limited to problem.Processing extensive changeable constraint Under the conditions of problem be generally to use heuritic approach, common heuritic approach has genetic algorithm, tabu search algorithm, ant colony Algorithm etc., but there is intrinsic defect in single algorithm.

Although genetic algorithm has considerable flexibility and robustness, it is suitable for solving of extensive problem, but it exists too early Ground convergence and the defect of local search ability difference.And the hybrid algorithm adding tabu search algorithm on the basis of genetic algorithm can This defect is well compensate for the advantage of releasing tabu search algorithm local optimum.The thought of whole hybrid algorithm is led to exactly Cross the initial solution of genetic algorithm structure tabu search algorithm, as the neighborhood of TABU search module, entered by tabu search algorithm One-step optimization is individual, promotes the quality of whole population, makes more excellent individuality occur in minimum iterative steps.

3, the breakthrough that application is actual

Conventional vehicle scheduling based on heuritic approach is all applied in the logistics company dispensing to customer demand, it is considered to Be all the vehicle scheduling from logistics center to client's hands, and consider that restrictive condition is the most single and there is the biggest reason Opinion property.The design, from the actual demand of plant area, breaks through conventional logistics condition outside the venue, considers physical presence in plant area Constraints, use heuritic approach to complete plant area's interior vehicle and commute the scheduling problem in different warehouse, from original from Focus on scattered one-to-many dis-tribution model be changed into current from warehouse to the many-one logistic pattern of this vehicle, according to actual Original algorithm is carried out bold innovation, to solve actual logistical problem.

The present invention, based on actual items background, utilizes MAS multiple agent model buildings vehicle scheduling based on railway platform system System model, and on the basis of abundant analysis demand, devise and operate the shortest time vehicle scheduling mathematics as target with vehicle Model, has wherein applied to heuritic approach widely used on Logistics Scheduling Problem: genetic algorithm and tabu search algorithm, On the basis of advantage both extracting, devise genetic algorithms based on both, make vehicle scheduling reach optimum.

Claims (4)

1. a structure based on multiple agent railway platform scheduling intelligent sequencing model, it is characterised in that: multi-agent system The calculating system that (Multi-Agent System, MAS) is made up of multiple intelligent bodies mutual in an environment；Many Multiagent system also can be used in insoluble problem in the intelligent body solving to separate and single-layer system；Intelligent body passes through one A little methods, function, process, searching algorithm realizes, and the communication of the most each intelligent body is to shelve behavior realization by coordinated with each other Communication each other, is based on communication technology；

The present invention includes ordering system based on railway platform scheduling and multi-agent algorithm two parts of design: wherein, described base Ordering system in railway platform scheduling also includes dis-tribution model, Agent modeling, vehicle scheduling；

Described dis-tribution model is logistics distribution model based on railway platform scheduling, and home-delivery center obtains the source of goods from upstream, through receiving, Flow processs such as storing, assemble and dispatch buses, finally delivers to client by goods；The logistic distribution vehicle scheduling problem of any situation is all Object function and the mathematical programming model of constraints two parts composition can be expressed as according to the method for mathematical modeling；This Inventing vehicle scheduling demand scheme based on railway platform scheduling, its mathematical model can be expressed as: min or max z=f (x) root According to the demand of actual items, can analyze and draw constraints: constraint that delivery period is relevant, the constraint that haulage vehicle is relevant, Relevant constraint standard work force, the constraint that emergency response is relevant, constraint that human resources is correlated with, constraint that handling facilities are relevant；

Described Agent modeling refers to: process the clear superiority of multi-constraint condition objective optimisation problems based on MAS, and the present invention will MAS is applied in railway platform dispatching patcher, and the main body (influence factor) of different levels can be carried out by the Agent of different levels Describe and express；The Agent of different levels connects each other, interacting has collectively constituted an actual dispatching patcher；

Described vehicle scheduling is the most key module designed for the railway platform scheduling target of intelligent sequencing, affects accordingly Factor relates generally to warehouse and two aspects of vehicle, in the environment in many warehouses, by as ready vehicle scheduling to suitable warehouse gate The efficiency of loading and unloading improving whole plant area relates to more influence factor, such as busy extent, the goods deposit in warehouse in this warehouse Amount etc.；The factor of comprehensive each side, in conjunction with each warehouse gate of on-site, vehicle and corresponding warehouse Intelligent Matching；For one The different sink door in individual warehouse, stores different products, under normal circumstances, each storehouse door correspondence one product (arrowband, plate etc.), by The practical situation loaded in mixture is needed, it is understood that there may be this warehouse gate deposits a small amount of product being originally not belonging to this storehouse door in different vehicle Product, after customer order completes, arrange vehicle to go to take corresponding product to corresponding warehouse Ku Men according to customer order；According to respectively The busy extent of individual warehouse gate or the impact of handling horizontal factor, the transportation route that vehicle may be different arrives corresponding warehouse Door completes entrucking work, the running of convenient vehicle below with the shortest time, improves the operational efficiency that system is overall；

Being constructed as follows of present invention railway platform scheduling mathematical model based on multiple agent:

The parameter related in above-mentioned model is explained as follows:

The set of G: warehouse gate

V: participate in the set of distribution vehicle

t_i: vehicle arrives the used time of warehouse i

L_i: the time (plant area's work in 24 hours, negligible) that warehouse i services the latest

E_i: the time (plant area's work in 24 hours, negligible) that warehouse i services the earliest

a_i: warehouse can provide the time (shortest time that vehicle needs to wait for when arriving warehouse) that entrucking services the earliest

b_i: warehouse can provide the time (maximum duration that vehicle needs to wait for when arriving warehouse) that entrucking services the latest

f₁: vehicle arrives warehouse and provides the penalty coefficient of service time early than warehouse

f₂: vehicle arrives warehouse and is later than the penalty coefficient of warehouse offer service time

c_ij: the vehicle time-consuming (scope is 5-10min) between warehouse i and j

P_i: warehouse gate i installs the time-consuming of a car required product

The capacity of q: vehicle

n_i: warehouse gate i is supplied to the density of cargo of car

The algorithm of present invention railway platform based on multiple agent scheduling model uses two kinds of wide heuritic approaches sent out of application, loses Hybrid algorithm that propagation algorithm combines with tabu search algorithm realizes, genetic algorithm (GA) be according to Darwinian natural selection and Theory of heredity, by survival of the fittest rule during biological evolution and same a group chromosome with entering the information intelligence that combines of exchange Algorithm；The performance of genetic algorithm is largely dependent upon intersects and the operation of variation, and this depends on solving how concentration takes out Sample this solution；Tabu search algorithm (TA) most important thought is some objects of the locally optimal solution that labelling correspondence has been searched for, And in further iterative search, avoid these objects rather than absolute prohibition circulation, thus ensure that different are had as far as possible The exploration of efficient search approach；

After mixing, the main policies of algorithm is exactly: first passes through genetic algorithm and carries out global search, uses natural number to all storehouses Storehouse and adjustable vehicle encode, and the availability in each warehouse carries out the path optimization of the overall situation with the carrying capacity of vehicle； Then TABU search is used with certain probability, the individuality in population to be carried out Local Search, namely for same car to institute Warehouse is had to carry out local transportation route optimization；First the present invention arranges initial population, then simulates biological evolution, in initial population Between produce and select, make a variation, intersect as a new generation population, population of new generation is cooked only chess game optimization, the individuality stayed, Through the heredity in too much generation, eventually form the individuality that fitness is best；

Hybrid algorithm solution procedure of the present invention is as follows:

(1) warehouse gate directly arranges natural number coding

First can design the different natural number arrangement not sticking with paste repetition of multiple 1-G, the arrangement of this natural number just constitutes body one by one. Successively warehouse gate can be inserted in travel route according to constraints, such as, call two cars and arrive the warehouse point of 4, it is assumed that The driving path of car is 1234, and traversal is numbered 1 the most successively, and first first warehouse point is inserted into row by the warehouse point of 2,3,4 Sailing in route, if meeting above-mentioned institute Prescribed Properties, inserting second warehouse point, proceeding, when beyond vehicle if meeting Carrying capacity time, call second car.

(2) initial population is set

It is randomly generated this G of 1-G mutual unduplicated natural number arrangement, i.e. generates body one by one.Assume the number of initial population For N, then produce the most different N number of individuality.

(3) fitness evaluation standard determines

Because optimization aim is to minimize, and the fitness of genetic algorithm represents the individuality that adaptation ability is the strongest, therefore can use mesh The reciprocal of scalar functions represents fitness.

F=1/Z₁

(4) operation is replicated

The design completes the operation to the population at individual survival of the fittest by retaining optimized individual and gambling dish strategy.

(5) intersection operation

Exchanging the individual Partial Fragment of two parents by certain probability and complete intersection operation, common crossover operator has portion Divide crossover operator, ordered crossover operator, recycling cross operator and class OX operator etc..

The design uses ordered crossover operator, such as one car complete according to shipping work need through 1,2,3,4,5,6,7 this Seven warehouse gates load corresponding product, existing two kinds of different route or travel by vehicle: R₁=1234567, R₂=3425167, R₁ Represent that vehicle sequentially passes through No. 1 door, No. 2 door ..., No. 7 doors, R₂Represent vehicle sequentially pass through No. 3 doors, No. 4 doors ..., No. 7 Door, therefrom selects a matching section,

According to the mapping relations of matching section, position corresponding outside matching section region is labeled as A, it may be assumed that

Shifted matching section is to original position again, and the reserved position identical with matching section space later, is labeled as A, it may be assumed that

Finally the matching section of two sequences is exchanged with each other, obtains two new offsprings, it may be assumed that

(6) mutation operation

Mutation operation embodies the thought of nature gene mutation, and common mutation operator has reverse variation, exchange mutation and inserts Enter variation.

The design uses reverse variation, randomly chooses two in a sequence and clicks on line position exchange, by anti-for 2 interior characters Sequence is inserted in former sequence.The such as travel route for a car is R₁=1234567, by second position and the 5th position Putting and carry out reversing variation, the sequence obtained is R₁ ^*=1543267.

(7) utilize TS algorithm that current solution is improved

Tabu search algorithm uses

Current solution is evaluated

Wherein T (i) represents that vehicle arrives the time that warehouse point needs, and E (i) represents that vehicle is in warehouse point handling goods and wait Time, W (i) represents the capacity of carriage of vehicle, and p represents penalty coefficient.

The execution step of tabu search algorithm is as follows:

Step one: selected initial solution (having genetic algorithm to obtain) x^now, make taboo list

Step 2: if meeting stop criterion, go to step four；Otherwise, at x^nowField N (x^nowSelect in) and meet what taboo required Candidate Set can_N (x^now), perform step 3.

Step 3: at can_N (x^now) select one group of evaluation of estimate optimal solution x^best, make x^now=x^best, update taboo list, go to step Two.

Step 4: output operation result.

(8) stop criterion

Because the dynamic factor affecting vehicle dispatch system stability is more, ageing to decision system wants height, and the design intends Use the stop criterion specifying algebraically step number to terminate；

The algorithm design of present invention railway platform based on multiple agent scheduling model is as follows:

Input parameter:

Population scale N, represents the running route of different initial vehicle

Evolutionary generation T, represents the algebraically that population is to be multiplied

Crossover probability P_c

Mutation probability P_m

Penalty coefficient p

Output result:

Vehicle scheduling route and optimization target values

Algorithm main body:

Producing multiple different initial population P (0) according to vehicle park coupling matrix, current algebraically is t=0；

Calculate the fitness of initial population

While(t<T)

{

Carry out individuality the highest for current algebraically fitness replicating operation, be inserted into P in a new generation (t+1)；

According to fitness and gambling dish selection strategy, calculate the select probability P of each individuality_i；

For (k=0；K≤N；K+=2)

{

According to select probability P_iTwo parent individualities are selected from parent population；

Random value between r=[0,1]

If (r≤P_c)

Carry out two parent individualities intersecting and operate, add in a new generation population P (t+1)；

else

{

Random value between r=[0,1]

If (r≤P_m)

Two parent individualities 1 are carried out mutation operation, adds in a new generation population P (t+1)；

Else parent individuality 1 directly replicates, and adds in a new generation population P (t+1)；

Random value between r=[0,1]

If (r≤P_m)

Two parent individualities 2 are carried out mutation operation, adds in a new generation population P (t+1)；

Else parent individuality 2 directly replicates, and adds in a new generation population P (t+1)；

Tabu search algorithm；

}

}

Calculate P (t+1) for population's fitness；

T=t+1；

}

Output result.

Structure based on multiple agent railway platform scheduling intelligent sequencing model the most according to claim 1, it is characterised in that: institute State Agent modeling to include: client Agent, warehouse Agent, vehicle Agent, road network Agent, labour force Agent, order processing Agent, emergency response Agent, vehicle scheduling Agent, handling scheduling Agent and total activation Agent, wherein:

1) client Agent comprise customer name, client codes, shipment schedule, require delivery time, the type of required bale packing and The information such as quantity；

2) warehouse Agent comprises the information such as bale packing type, bale packing number, stock coefficient；

3) vehicle Agent with driver bind, comprise vehicle location state (vehicle wait, just in entrucking, install vehicle), vehicle dress Carrying capacity, car plate, driver, car such as criticize at the information；

4) road network Agent comprise the different Ku Men of warehouse point, warehouse compartment and they between switch time-consuming information；

5) labour force Agent comprises single packaging car standard work force, stevedore's information (total number of persons, distributes number, people to be allocated Number) etc. information；

6) order processing Agent can store present situation and the existing goods of order requirements entry evaluation of client according to the goods in warehouse Can amount meet the demand of client, and final assessment report informs the total activation Agent, total activation Agent of the superiors Determine to take generate loading and transporting plan list (plan odd numbers is as order ID) or arrange complementary goods goods according to assessment report The production schedule；

7) emergency response Agent belongs to the intelligent body that one direction relies on, for the emergency in processing system；Act on system Each transport instruction key link on, when reality perform confirm instruction not in planned time internal feedback to total activation Agent, early warning information can be displayed immediately on system billboard, and the process initiative of abnormal conditions is strong；

Total activation, according to the loading and transporting plan generated, calculates shipment shipment line and transport mileage (is used in and counts in the way time according to theory Calculate on the time of making the product)；Loading and transporting plan is supplied to vehicle scheduling Agent, vehicle scheduling Agent and completes vehicle scheduling；Last root March into the arena the time according to the result of calculation of the vehicle scheduling Agent vehicle that falls back；

8) vehicle scheduling Agent is the key of whole dispatching patcher, when customer order generation and system do not occur unexpected feelings Under condition, total activation Agent extracts the human resource information loading scheduling Agent, and it is handed down in the lump vehicle scheduling with order Agent, vehicle scheduling Agent are according to the resource such as order and vehicle, to entrucking order rational management, are that system whole efficiency reaches The highest；

9) handling scheduling Agent is responsible for the transfer of labour force, reasonably arranges worker-hours and place, allows workman just The true time occurs in correct doorway, warehouse handling goods, calculates the entrucking used time, is returned to higher level Agent with regard to result of calculation, allows Vehicle is minimum in the warehouse time of staying；

Total activation Agent is the brain of whole system, is responsible for synchronization and the management of whole system, coordinates each Agent, make them Between cooperate, it is ensured that the operation that whole system is orderly.

Structure based on multiple agent railway platform scheduling intelligent sequencing model the most according to claim 1, it is characterised in that: institute Stating in vehicle scheduling module, D represents warehouse,Represent the weight of different affecting factors, Δ_ijRepresent that i-th warehouse is to jth shadow The evaluation situation of the factor of sound, warehouse handling capacity represents the entrucking level of Current warehouse, loads unit product under real-time status Used time；Vehicle time-consumingly represents that to warehouse empty-car (not loading goods) is dispatched a car a little to the use of opposite depot door in field Time；Articles from the storeroom reserves represent the maximum of the corresponding goods that Current warehouse door can provide, and decide and whether meet vehicle Demand；Vehicle rate of actual loading represents the loading condition of Current vehicle, and residue is available for the space of product loading and also has how many, affects car Select suitable warehouse gate.Weight represents each factor significance level to vehicle scheduling.

The best match algorithm of vehicle initial rest position is (with warehouse gate D₀₁As a example by):

Step one: the index of standardization coupling matrix, becomes to can be used to the same amount compared by different unit conversions；

Δ₁₀₁=P_{Operation number}*0.6+P_{Driving is away from Ku Men position}*0.2+P_{Waiting time}*0.2

Δ₂₀₁=P_{Arrive the warehouse gate time}

Δ₃₀₁=P_{Corresponding product amount}*0.9+P_{Other products}*0.1

Δ₄₀₁=P_{Vehicle is real to be carried}/P_{Vehicular load}

To Δ₁₀₁、Δ₂₀₁、Δ₃₀₁、Δ₄₀₁、Δ₅₀₁、Δ₆₀₁Carry out unit-normalization

Step 2: the weight of each influence factor of normalized；

WeighValue, and make its meet:

Step 3: calculate the comprehensive evaluation value that warehouse is stopped；

The computing formula of comprehensive evaluation value is:

Step 4: when there being one or more vehicle to simultaneously participate in transport, considers the highest warehouse of evaluation of estimate as dispensing The initial optimal warehouse of vehicle parking.

Structure based on multiple agent railway platform scheduling intelligent sequencing model the most according to claim 1, it is characterised in that: institute State railway platform scheduling mathematical model Chinese style (3.1) based on the multiple agent optimization aim as whole model, represent vehicle Finally complete the shortest time of task through sequence, constraints (3.2) represents that the time in vehicle arrival warehouse can not be early than storehouse Storehouse starts to provide the time of service, it is impossible to is later than warehouse and closes the time of service, for plant area 24 hours work, negligible this Part；Formula (3.3) represents that the final carrying capacity of vehicle not can exceed that the carrying capacity of vehicle itself；Formula (3.4), (3.5) represent one Individual vehicle once and at most once can only not can exceed that the sum in warehouse through same warehouse and the sum through warehouse.