CN108334986A - A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism - Google Patents
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism Download PDFInfo
- Publication number
- CN108334986A CN108334986A CN201810115255.0A CN201810115255A CN108334986A CN 108334986 A CN108334986 A CN 108334986A CN 201810115255 A CN201810115255 A CN 201810115255A CN 108334986 A CN108334986 A CN 108334986A
- Authority
- CN
- China
- Prior art keywords
- plastic mechanism
- individual
- rescheduling
- scheduling
- method based
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 115
- 230000007246 mechanism Effects 0.000 title claims abstract description 54
- 239000003795 chemical substances by application Substances 0.000 title claims abstract description 45
- 238000005457 optimization Methods 0.000 claims abstract description 47
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims abstract description 3
- 238000012216 screening Methods 0.000 claims description 19
- 210000003127 knee Anatomy 0.000 claims description 15
- 239000012141 concentrate Substances 0.000 claims description 7
- 230000006872 improvement Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 4
- 238000009826 distribution Methods 0.000 description 8
- 210000002569 neuron Anatomy 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- WTTWSMJHJFNCQB-UHFFFAOYSA-N 2-(dibenzylamino)ethanol Chemical compound C=1C=CC=CC=1CN(CCO)CC1=CC=CC=C1 WTTWSMJHJFNCQB-UHFFFAOYSA-N 0.000 description 3
- 230000000452 restraining effect Effects 0.000 description 3
- 238000005520 cutting process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 230000000638 stimulation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012856 packing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009257 reactivity Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- General Factory Administration (AREA)
Abstract
The present invention relates to a kind of multiple agent Multipurpose Optimal Methods of the rescheduling method based on plastic mechanism, sequential encoding is carried out to all intelligent bodies first, using intelligent body as population gene initialization population, population at individual is the scheduling strategy for all intelligent bodies, the context completed according to individual establishes Work flow model, the rescheduling method based on plastic mechanism is reapplied Work flow model is solved to obtain optimal policy model, then the NSGA III optimization algorithms of application enhancements carry out multiple-objection optimization processing and acquire Pareto disaggregation and target disaggregation, optimal element is determined in target disaggregation and chooses the correspondence scheduling strategy of Pareto solution concentrations, the scheduling strategy scheduling multiple agent finally chosen according to preceding step executes task and completes multiple-objection optimization.The present invention can effectively realize multiple-objection optimization, and effect of optimization is good, HV values >=0.450643, IGD value≤0.229190 of inventive algorithm.
Description
Technical field
The invention belongs to multiple agent cooperative scheduling field, it is related to a kind of the more of the rescheduling method based on plastic mechanism
Intelligent body Multipurpose Optimal Method.
Background technology
Intelligent body is exactly as its name suggests the entity with intelligence, the entitled Agent of English.Intelligent body refer to reside in it is a certain
Under environment, it can continue automatically to play a role, have the computational entity of the features such as presence, reactivity, social, initiative.
Intelligent body had not only included robot, but also including computational entities such as computer, computer clusters.Multi-agent system is multiple intelligent bodies
The set of composition, the purpose is to will be big and complicated system Construction at it is small, communicate and coordinate each other, be easily managed
System.With the continuous development of society, the application of multiple agent is more extensive.
Multi-robot system (Multi-robot systems, MRS) is relatively easy by multiple structures, by mutually assisting
Make to execute the total system of some or a certain group of complex task parallel.Conjunction by task resource to each robot in MRS
Reason distribution, can effectively promote the execution speed of overall task, reach the unapproachable mission requirements of individual machine people.Therefore,
The task reasonable distribution between multirobot is studied, is realized between individual robot and effectively collaboration, the task of MRS entirety is improved
Executive capability has critically important research significance and actual application value.
In the cooperative scheduling of MRS, mostly important challenge be how the phase between the multiple robots of reasonable consideration
The conflict relationship mutually competed, by reasonable effective resource allocation, farthest to promote the overall execution ability of MRS.And
Under the background of the overall task diversification of demand, also need to consider how according to set multiple targets, in conjunction with multiple mesh
Restriction condition between mark, to rationally provide alternate strategies set.Usually in the system of this cooperation collaboration, robot
Demand between body, between multiple targets is mutual restriction and conflict, such as RoboCup robot worlds Cup, robot
It plays chess to wait in multi-robot systems and usually can take into account these problems.
In practical application, a certain group of complex task is often made of multiple simple subtask mixing, and each simple son
Task often has the completion condition mutually restricted.Such as only in the case where all prime tasks of the task are completed, this
Business just has the condition that can be performed.Only in the case that all tasks are all completely completed, the complex task group ability quilt
It is fully finished.It is such that there is the complex task demand environment for mutually restraining subtask, it often can correspond to the machine of industrial process
The application environments such as workflow schedule in device people scheduling, cloud computing.
During actual multirobot execution task, consider that the accidental error factor occurred has critically important grind
Study carefully value.In the case that mission failure rate obeys certain distribution, integration and publication again to failed tasks progress are needed, and
By calling the robot resource of existing current task free time, task is re-executed, realizes certain execution fault rate
In the case of, the supplement that failed tasks execute is dispatched so that entire task groups are effectively completed.
In different application environments, multirobot cooperative system can have different mission requirements, as multirobot is assisted
It is often most short as main target using the whole search and rescue time with searching and rescuing；In the application background that multirobot cooperative surroundings are explored
Under, it is most short as tactful good and bad Main Basiss that distance is often expended using whole robot.So according to application scenarios
Difference, selected primary evaluation index then respectively have weighting.By the way that multiple evaluation indexes to condition each other are established as target letter
Number, is sought optimal policy disaggregation using multi-objective optimization algorithm, can both be closed with the restriction between each object function of objective description
System, while also can provide reference set for the policy selection under certain special scenes.
The scheduling of multiple agent and task mechanism are similar with Work flow model, therefore the research of Workflow Management is to mostly intelligent
Body cooperative scheduling has certain reference significance.CN 102509197A propose a kind of concentrate tube to multiple workflow engines
Reason；CN 101615269 is optimized the fallback mechanism of task under the task context of workflow, realizes in workflow
The operation that task can arbitrarily retract in implementation procedure enhances the flexibility of task execution realization；CN 106845642A are provided
The adaptive multi-target evolution method of belt restraining cloud workflow schedule a kind of can improve the global of multi-target evolution method and visit
Survey and local producing capacity, are suitable for solving the multi-objective optimization question of belt restraining, and can be applied in cloud computing environment work
Flow dispatching technique field.Although foregoing invention has carried out Work flow model certain optimization, it is capable of providing to a certain extent more
The solution of objective optimisation problems, but its be directed to be optimization single link, it is excellent to provide complete reliable multiple target
Change method.
Therefore, a kind of complete and good effect of optimization great realistic meaning of multiple agent Multipurpose Optimal Method is studied.
Invention content
It is a kind of based on the more of dynamic accidentally failed tasks environment the purpose of the present invention is overcoming the deficiencies of the prior art and provide
The readjustment degree of intelligent body Multipurpose Optimal Method, rescheduling method of this method based on plastic mechanism, the rescheduling method draws
It holds up and is optimized by plastic mechanism, the failure of readjustment degree is adjusted come dynamic by the fluctuation for frequency of slipping up in the unit interval
Task merging ratio keeps Mission Success rate to stablize in the case of Random Task failure；And on this basis by improved
NSGAIII objective optimization algorithms seek Pareto disaggregation and target disaggregation has obtained strategy set and policy execution result collection
It closes, then determines optimal element Selection Strategy, and multiple-objection optimization is completed according to Selection Strategy scheduling multiple agent.
In order to achieve the above object, the technical solution adopted by the present invention is：
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism, the present invention is by exchanging
Degree strategy optimize realize multi-agent system multiple-objection optimization (shorten whole completion date, individual Maximal Makespan and
Total consuming time of execution task), for the environment of each intelligent body random error, first to all intelligent body carry out sequence volumes
Code, using intelligent body as population gene initialization population, population at individual is the scheduling strategy for all intelligent bodies, according to individual
The precedence relationship of context, that is, task of completion establishes Work flow model, reapplies the rescheduling method based on plastic mechanism
Work flow model is solved to obtain optimal policy model so that under any preassigned Task Assigned Policy, exported
Multiple target (whole completion date, it is whole expend and individual Maximal Makespan) stablize as possible to realize fault-tolerant there is appearance
The ability for bearing random error is mainly manifested in multiple target (whole completion date, whole consuming and individual Maximal Makespan)
It will not be acquired with error larger fluctuation, then the NSGA III optimization algorithms progress multiple-objection optimization processing of application enhancements
Pareto disaggregation and target disaggregation, target disaggregation determine optimal element and choose Pareto solution concentrate correspondence scheduling strategy,
The scheduling strategy scheduling multiple agent finally chosen according to preceding step executes task and completes multiple-objection optimization；Pareto solutions are concentrated every
A element represents the scheduling strategy after one group of optimization, and target solution concentrates each element to represent the scheduling plan after one group of target optimizes
It is slightly executing as a result, each element that Pareto disaggregation and target solution are concentrated corresponds；In the case where considering all targets, into
Row multiple-objection optimization is handled, then obtained final result should be then a group policy disaggregation and corresponding target disaggregation,
One group of set being made of element with strategy that tactful disaggregation, that is, table obtains under the tradeoff of multiple targets；
The multiple target includes whole completion date W1, individual Maximal Makespan W2And execute total consuming time of task
W3；The optimal element of determination refers to for the W in single-element1、W2And W3Weighted sum obtains overall targetThe element of M minimums is optimal element, wherein NiFor WiWeight coefficient, N1、N2And N3It is 0.3333；
The rescheduling method based on plastic mechanism refers to being optimized using plastic mechanism counterweight scheduling engine
The method of obtained rescheduling method, the optimization is：In true environment, task often has the characteristic successively accepted, certain
A task only in the case where all prime tasks are completed, can be just performed, accidental for failed tasks in true environment
There is a situation where according to the time interval that adjacent failed tasks twice generate, dynamically adjust weight scheduling engine to failed tasks
Merging ratio, mission failure rate distribution remain unchanged under the premise of, when adjacent failed tasks twice generate time interval compared with
It is small, it reduces failed tasks and merges ratio, be equivalent to the successful number for improving task execution in this time indirectly；
The optimal policy model refers to reflection scheduling strategy and the target i.e. model of scheduling strategy implementing result relationship；
The improved NSGA III multi-objective optimization algorithms refer to using Knee Point thoughts to the more mesh of NSGA III
The algorithm that mark optimization algorithm obtains after being improved, the improved method are：In conjunction with the screening technique based on reference point distance
Individual picking rule with the screening technique of Knee Point as NSGAIII, from the last one non-dominant level picking individual
Into filial generation.
As preferred technical solution：
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, the party
Method the specific steps are：
(1) first to all intelligent body carry out sequence integer codings, using intelligent body as population gene initialization population, kind
Group's individual is the strategy for all intelligent bodies, and the context completed according to individual establishes Work flow model；
(2) the case where accidentally failing in the dispatch environment of multitask is executed parallel for multiple agent, the present invention is based on god
The heavy scheduling engine in rescheduling method is optimized through inherent plastic mechanism in science, is obtained based on plastic mechanism
Rescheduling method；
(3) Work flow model is solved to obtain optimal policy model using the rescheduling method based on plastic mechanism, is made
The final result for obtaining the scheduling of pinned task allocation strategy Imitating keeps stable as possible；
(4) it is directed to such dynamic environment, NSGAIII objective optimizations algorithm improved to optimal policy model use calculates
To the better Pareto disaggregation of index and corresponding target disaggregation, improvement refers to being selected in the last one non-dominant level
When body enters filial generation, in conjunction with screening technique and Knee Point based on reference point distance screening technique as NSGAIII's
Individual picking rule；
(5) it determines optimal element in target disaggregation, and chooses the correspondence scheduling strategy that Pareto solutions are concentrated, finally according to choosing
The scheduling strategy scheduling multiple agent taken executes task and completes multiple-objection optimization.
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, it is described
Counterweight scheduling engine optimizes specially：
According to the time interval that adjacent failed tasks twice generate, conjunction of the weight scheduling engine to failed tasks is dynamically adjusted
And ratio, time interval are as follows with the relationship for merging ratio：
Wherein T is the time interval that adjacent failed tasks twice generate, and s is the merging ratio of failed tasks, and p is correlation
Coefficient depends primarily on the number of overall task, and R is adjustment thresholding, determines cutting for the corresponding adjustment formula during adjustment
Away from C is adjustment sensitivity, determines the size changed caused by T variations during adjusting.
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, it is described
Rescheduling method is MaxMin dispatching methods.
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, it is described
Work flow model is solved to obtain optimal policy model using the rescheduling method based on plastic mechanism refer to：By workflow mould
Type, which is read to be added in the queue of the rescheduling method based on plastic mechanism by model explanation device, obtains optimal policy model.
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, it is described
Screening technique in conjunction with screening technique and Knee Point based on reference point distance is as the individual picking rule of NSGAIII
Refer to：First the nearest n% of distance reference point is selected from the last one non-dominant level according to the screening technique based on reference point distance
Individual, then the screening technique of Knee Point chooses Knee Point individuals as update in wherein according to KnEA algorithms
Body is that the individual for the hypervolume minimum that selection is surrounded with reference axis enters filial generation, is selected successively, until the number of more new individual
Amount is equal with population scale number, and the effect achieved is better than NSGAIII and other multi-objective Evolutionary Algorithms.
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, it is described
Scheduling strategy scheduling multiple agent according to selection refers to by the substrategy in scheduling strategy according to its pass corresponding with intelligent body
System is sequentially inputted to each intelligent body, and for each intelligent body according to input instruction execution order, the substrategy is single in scheduling strategy
The corresponding strategy of intelligent body.
A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism as described above, it is described
HV values >=0.450643, IGD value≤0.229190 of improved NSGAIII objective optimizations algorithm, the present invention is compared to existing skill
Art significantly improves HV values, reduces IGD values, i.e., algorithm of the invention obtain disaggregation diversity it is more preferable, algorithm the convergence speed
Faster.
Advantageous effect：
(1) the multiple agent Multipurpose Optimal Method of a kind of rescheduling method based on plastic mechanism of the invention,
Under the conditions of fixed schedule strategy, that is, task is fixed, readjustment degree can be dynamically adjusted according to the fluctuation of error rate in the unit interval
Engine merges ratio to issuing again with appoint failed tasks, to as keep optimization aim (whole completion date, a as possible
Body Maximal Makespan and execute the total of task and expend the time) stabilization；
(2) the multiple agent Multipurpose Optimal Method of a kind of rescheduling method based on plastic mechanism of the invention is right
NSGAIII is improved, and is solved to model using the innovatory algorithm, and NSGAIII is better than in the HV values and IGD of algorithm
With other multi-objective Evolutionary Algorithms, the convergence rate of innovatory algorithm of the present invention and the diversity for obtaining disaggregation are more preferable；
(3) the multiple agent Multipurpose Optimal Method of a kind of rescheduling method based on plastic mechanism of the invention, can
Effectively the Mission Capability of multi-agent system is optimized, dynamic adjusts the task execution process of multiple agent.
Description of the drawings
Fig. 1 is the flow chart of the present invention；
Fig. 2 is the flow chart of the rescheduling method based on plastic mechanism of the present invention；
Fig. 3 is the flow chart of the improvement NSGAIII algorithms of the present invention；
Fig. 4 is workflow test model Montage-100 schematic diagrames；
Fig. 5 is the test result of the optimal policy model of the present invention；
Fig. 6 is the HV Data-Statistics results that the present invention improves NSGAIII algorithms；
Fig. 7 is the IGD Data-Statistics results that the present invention improves NSGAIII algorithms.
Specific implementation mode
Present invention will be further explained below with reference to specific examples.It should be understood that these embodiments are merely to illustrate the present invention
Rather than it limits the scope of the invention.In addition, it should also be understood that, after reading the content taught by the present invention, people in the art
Member can make various changes or modifications the present invention, and such equivalent forms equally fall within the application the appended claims and limited
Range.
A kind of multiple agent Multipurpose Optimal Method of rescheduling method based on plastic mechanism of the present invention, such as Fig. 1
It is shown, it is as follows：
(1) first to all intelligent body carry out sequence integer codings, using intelligent body as population gene initialization population, kind
Group's individual is the strategy for all intelligent bodies, and the context completed according to individual establishes Work flow model；
(2) the case where accidentally failing in the dispatch environment of multitask is executed parallel for multiple agent, the present invention is based on god
The heavy scheduling engine in rescheduling method, that is, MaxMin dispatching methods is optimized through inherent plastic mechanism in science, is obtained
Rescheduling method based on plastic mechanism as shown in Figure 2；
In Neuscience, inherent plasticity is a kind of mechanism of objective reality of neuron self-control.Work as neuron
When the stimulus signal of receiving is excessively frequent, neuron itself can gradually improve the threshold value of the receiving stimulation of itself；Work as neuron
When the stimulus signal of receiving becomes rareness, neuron itself can be gradually lowered the threshold value that itself receives stimulation；In task scheduling
In environment, consider when the error rate that a certain section of moment executes task is higher, the Mission Capability of corresponding whole system
It can decline to a great extent；It is optimized based on inherent plastic mechanism counterweight scheduling engine, dynamic reduces the packing size of task, indirectly
Improve the number of successful execution task so that still maintained a degree of Mission Capability in this case, specially：
According to the time interval that adjacent failed tasks twice generate, conjunction of the weight scheduling engine to failed tasks is dynamically adjusted
And ratio, time interval are as follows with the relationship for merging ratio：
Wherein T is the time interval that adjacent failed tasks twice generate, and s is the merging ratio of failed tasks, and p is correlation
Coefficient depends primarily on the number of overall task, and R is adjustment thresholding, determines cutting for the corresponding adjustment formula during adjustment
Away from C is adjustment sensitivity, determines the size changed caused by T variations during adjusting；
(3) Work flow model is solved to obtain optimal policy model using the rescheduling method based on plastic mechanism, i.e.,
It is added in the queue of the rescheduling method based on plastic mechanism as shown in Fig. 2, Work flow model is read by model explanation device
Obtain optimal policy model so that the final result of pinned task allocation strategy Imitating scheduling keeps stable as possible；
Under invocation framenort, scheduling strategy is among being passed to " current queue " module by task distributor, and model
Whether interpreter is completed, will have can by reading preset Work flow model by the prime task of each subtask
Task to be performed condition is passed among " current queue " module；Only in current queue some task there are the case where
Under, task distributor could execute the task in current queue by calling some execution unit；In addition to this,
Remaining execution unit is in idle state；In simulated environment, error monitor is whether execution unit executes times for completing distribution
Business provides judgment criteria, to distinguish the task of the task that is completed and failure；The task of failure is carried out by weight scheduling engine
It repacks, is added in " current queue " module again, the heavy scheduler task rejoined will be still by the shadow of mission failure rate
It rings；The maximum execution unit for executing task ability of current idle goes to execute the task with minimum time-consuming demand, so recycles
Back and forth until all tasks are finished completely；
(4) NSGA III multi-objective optimization algorithms are improved, concrete structure is with improvement rule as shown in figure 3, at last
When a non-dominant level picking individual enters filial generation, in conjunction with the sieve of screening technique and Knee Point based on reference point distance
Individual picking rule of the choosing method as NSGAIII enters filial generation, specially from the last one non-dominant level picking individual：
First distance reference point to be selected from the last one non-dominant level nearest according to the screening technique based on reference point distance
N% individual, then in wherein according to KnEA algorithms Knee Point screening technique choose Knee Point individual conducts
More new individual is that the individual for the hypervolume minimum that selection is surrounded with reference axis enters filial generation, is selected successively, until update
The quantity of body is equal with population scale number；
(5) it is better that index is calculated in NSGAIII objective optimizations algorithm improved to optimal policy model use
Pareto disaggregation and corresponding target disaggregation, Pareto solutions concentrate each element to represent the scheduling strategy after one group of optimization,
Target solution concentrate each element represent the scheduling strategy after one group of target optimizes execution as a result, Pareto disaggregation and target
The each element that solution is concentrated corresponds, and target includes whole completion date W1, individual Maximal Makespan W2And execute the total of task
Expend time W3；
(6) it determines optimal element in target disaggregation and chooses the correspondence scheduling strategy of Pareto solution concentrations, determine optimal member
Element is for the W in single-element1、W2And W3Weighted sum obtains overall targetThe element of M minimums is most
Excellent element, wherein NiFor WiWeight coefficient, N1、N2And N3It is 0.3333；
(7) it executes task according to the scheduling strategy of selection scheduling multiple agent and completes multiple-objection optimization, i.e., by scheduling strategy
In substrategy be sequentially inputted to each intelligent body according to the correspondence of itself and intelligent body, each intelligent body is according to input instruction execution
Order, substrategy are the corresponding strategy of single intelligent body in scheduling strategy.
The present embodiment calls Montage_100 Work flow models input step (2) to be tested, and Montage models are one
The generally acknowledged multiple agent Work flow model of kind, is created by NASA/IPAC, and star is taken the photograph to multiple in outer space exploration by one kind
Null images carry out the integrated Work flow model of splicing.
In Montage_100, a total of 100 mission requirements nodes, specific each task relies on relationship such as Fig. 4 institutes
Signal.As it can be seen that other than first order task, each task has at least one relied on prime task.In this case, only
Have and be completed completely in prime task, then this task just has the condition executed.And set 20 available task executions
Unit, i.e. be up to 20 tasks are run parallel simultaneously.
In its workflow execution system, it is assumed that execute the mortality Follow Weibull Distribution of task：
Wherein x is stochastic variable, λ>0 is scale parameter, k>0 is form parameter.
During the test, the scheduling relationship between some execution unit and task node is fixed in advance.It is only in office
When business failure, in the case of carrying out rescheduling strategy, just using the heavy scheduling engine after being optimized based on plastic mechanism at that time
Situation made up.
Form parameter k=1 is set, task execution fault rate obeys exponential distribution under the setting condition.To simulation result
Statistical experiment is carried out, the optimal policy Model Independent of step (3) is run 5000 times and counted under every group of scale parameter situation
Analysis.The visible Fig. 5 of concrete outcome, by Fig. 5 it can be found that with fault rate growth, implementing result is in whole completion date, whole
Body expend and individual these three targets of Maximal Makespan on fluctuation it is little.
Set Weibull distribution parameter as：K=1, λ=100, select NSGAII, MOEA/D, DBEA, NSGAIII,
RVEA and improvement NSGAIII (K-NSGAIII) algorithms of the present invention independently run optimal policy model 50 times, and statistics is real
It tests shown in result such as Fig. 6 (HV), Fig. 7 (IGD), table 1 (HV) and table 2 (IGD).By table 1 and table 2 it can be found that the present invention's changes
HV values into NSGAIII algorithms are much larger than other algorithms, and IGD values, which are less than other algorithms, has the disaggregation diversity calculated high, receives
Fast advantage is held back, the algorithm of the present invention has on multiple statistical experiment in HV values and IGD values it can be seen from Fig. 6 and 7
Relative fluctuation is little, that is, variance is smaller, and stability is good.It can be seen that the present invention is establishing one kind based on plastic mechanism
Rescheduling method, and with the dispatching method Work flow model is solved to obtain optimal policy model, and in optimal plan
Slightly on the basis of model, it is proposed that the multi-objective optimization algorithm of improvement NSGAIII for the model a kind of, using algorithm pair
Optimal policy model solution, disaggregation diversity height and fast convergence rate in solution procedure.The present invention selects more true imitative
True environment has relatively significant practice significance, on this basis, system is dispatched by being applied to multiple agent, can be with
The restricting relation between each target is described, while effectively iterating to calculate out Pareto disaggregation, it is optimal mostly intelligent to select
Body task scheduling strategy provides suitable reference policy set.
| HV | NSGAII | MOEA/D | DBEA | NSGAIII | RVEA | K-NSGAIII |
| Min: | 0.025812 | 0.0 | 0.004698 | 0.071828 | 0.0 | 0.450643 |
| Median: | 0.067357 | 0.026648 | 0.037969 | 0.206127 | 0.000852 | 0.562194 |
| Max: | 0.120648 | 0.165798 | 0.107499 | 0.368534 | 0.092078 | 0.647501 |
Table 1
| IGD | NSGAII | MOEA/D | DBEA | NSGAIII | RVEA | K-NSGAIII |
| Min: | 0.325454 | 0.297799 | 0.346063 | 0.133299 | 0.407729 | 0.094520 |
| Median: | 0.420769 | 0.536281 | 0.491865 | 0.220765 | 0.748984 | 0.127628 |
| Max: | 0.547381 | 0.913793 | 0.663375 | 0.396861 | 1.249518 | 0.229190 |
Table 2.
Claims (8)
1. a kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism, it is characterized in that：It is right first
All intelligent bodies carry out sequential encoding, and using intelligent body as population gene initialization population, population at individual is for all intelligence
The scheduling strategy of body, the context completed according to individual establish Work flow model, reapply the readjustment based on plastic mechanism
Degree method solves Work flow model to obtain optimal policy model, and then the NSGAIII optimization algorithms of application enhancements carry out more mesh
Mark optimization processing acquires Pareto disaggregation and target disaggregation, determines optimal element in target disaggregation and chooses what Pareto solutions were concentrated
Corresponding scheduling strategy, the scheduling strategy scheduling multiple agent finally chosen according to preceding step execute task and complete multiple-objection optimization；
Pareto solutions concentrate each element to represent the scheduling strategy after one group of optimization, and target solution concentrates each element to represent one group of target,
Pareto disaggregation is corresponded with each element that target solution is concentrated；
The multiple target includes whole completion date W1, individual Maximal Makespan W2And execute total consuming time W of task3；Institute
It states and determines that optimal element refers to for the W in single-element1、W2And W3Weighted sum obtains overall targetM is most
Small element is optimal element, wherein NiFor WiWeight coefficient, N1、N2And N3It is 0.3333；
The rescheduling method based on plastic mechanism refers to optimizing to obtain using plastic mechanism counterweight scheduling engine
Rescheduling method, the method for the optimization is：According to the time interval that adjacent failed tasks twice generate, weight is dynamically adjusted
Merging ratio of the scheduling engine to failed tasks；
The optimal policy model refers to the model for reflecting scheduling strategy and relationship by objective (RBO)；
The improved NSGA III multi-objective optimization algorithms refer to excellent to NSGA III multiple targets using Knee Point thoughts
Change the algorithm obtained after algorithm is improved, the improved method is：In conjunction with based on reference point distance screening technique with
Individual picking rule of the screening technique of Knee Point as NSGAIII, from the last one non-dominant level picking individual into
Enter filial generation.
2. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 1
Method, which is characterized in that this method the specific steps are：
(1) first to all intelligent body carry out sequence integer codings, using intelligent body as population gene initialization population, population
Body is the strategy for all intelligent bodies, and the context completed according to individual establishes Work flow model；
(2) the heavy scheduling engine in rescheduling method is optimized based on inherent plastic mechanism in Neuscience, obtains base
In the rescheduling method of plastic mechanism；
(3) Work flow model is solved to obtain optimal policy model using the rescheduling method based on plastic mechanism；
(4) Pareto disaggregation and right with it is calculated in NSGAIII objective optimizations algorithm improved to optimal policy model use
The target disaggregation answered, improvement refer to when the last one non-dominant level picking individual enters filial generation, in conjunction with remote based on reference point
Individual picking rule of the screening technique of close screening technique and Knee Point as NSGAIII；
(5) it determines optimal element in target disaggregation, and chooses the correspondence scheduling strategy that Pareto solutions are concentrated, finally according to selection
Scheduling strategy dispatches multiple agent and executes the i.e. completion multiple-objection optimization of task.
3. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 2
Method, which is characterized in that the counterweight scheduling engine optimizes specially：
According to the time interval that adjacent failed tasks twice generate, merging ratio of the weight scheduling engine to failed tasks is dynamically adjusted
Example, time interval are as follows with the relationship for merging ratio：
Wherein T is the time interval that adjacent failed tasks twice generate, and s is the merging ratio of failed tasks, and p is related coefficient, R
To adjust thresholding.
4. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 3
Method, which is characterized in that the rescheduling method is MaxMin dispatching methods.
5. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 2
Method, which is characterized in that the application solves Work flow model based on the rescheduling method of plastic mechanism to obtain optimal policy
Model refers to：Work flow model is read to be added in the queue of the rescheduling method based on plastic mechanism by model explanation device and is obtained
To optimal policy model.
6. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 2
Method, which is characterized in that the screening technique conduct of the combination screening technique based on reference point distance and Knee Point
The individual picking rule of NSGAIII refers to：First according to the screening technique based on reference point distance from the last one non-dominant level
N% nearest individual of distance reference point is selected, then chooses Knee Point according to the screening technique of Knee Point wherein
Individual is used as more new individual, is selected successively, until the quantity of more new individual is equal with population scale number.
7. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 2
Method, which is characterized in that the scheduling strategy scheduling multiple agent according to selection refer to by the substrategy in scheduling strategy according to
The correspondence of itself and intelligent body is sequentially inputted to each intelligent body, and each intelligent body is according to input instruction execution order, the sub- plan
The corresponding strategy of single intelligent body slightly in scheduling strategy.
8. a kind of multiple agent multiple-objection optimization side of rescheduling method based on plastic mechanism according to claim 1
Method, which is characterized in that HV values >=0.450643, IGD value≤0.229190 of the improved NSGAIII objective optimizations algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810115255.0A CN108334986B (en) | 2018-02-06 | 2018-02-06 | Multi-agent multi-target optimization method of rescheduling method based on plasticity mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810115255.0A CN108334986B (en) | 2018-02-06 | 2018-02-06 | Multi-agent multi-target optimization method of rescheduling method based on plasticity mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108334986A true CN108334986A (en) | 2018-07-27 |
| CN108334986B CN108334986B (en) | 2020-06-02 |
Family
ID=62928120
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810115255.0A Expired - Fee Related CN108334986B (en) | 2018-02-06 | 2018-02-06 | Multi-agent multi-target optimization method of rescheduling method based on plasticity mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108334986B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110070235A (en) * | 2019-05-01 | 2019-07-30 | 湖南大学 | A kind of flexible dispatching method of multiple mobile robot |
| CN111324429A (en) * | 2019-10-09 | 2020-06-23 | 中国人民解放军国防科技大学 | Micro-service combination scheduling method based on multi-generation ancestry reference distance |
| CN112181608A (en) * | 2019-07-03 | 2021-01-05 | 中南大学 | Distributed distribution algorithm for multipoint dynamic aggregation tasks based on local information |
| CN113031650A (en) * | 2021-03-04 | 2021-06-25 | 南京航空航天大学 | Unmanned aerial vehicle cluster cooperative target distribution design method under uncertain environment |
| CN113419854A (en) * | 2021-06-23 | 2021-09-21 | 东北大学 | Cloud resource scheduling method oriented to unbalanced multi-objective optimization |
| CN116521344A (en) * | 2023-05-12 | 2023-08-01 | 广州卓勤信息技术有限公司 | AI algorithm scheduling method and system based on resource bus |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550771A (en) * | 2015-12-09 | 2016-05-04 | 中冶南方工程技术有限公司 | Multi-objective optimization method of steelmaking-continuous casting production scheduling based on NSGA-II |
| CN105929689A (en) * | 2016-04-22 | 2016-09-07 | 江南大学 | Machine tool manufacturing system processing and energy saving optimization method based on particle swarm algorithm |
-
2018
- 2018-02-06 CN CN201810115255.0A patent/CN108334986B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550771A (en) * | 2015-12-09 | 2016-05-04 | 中冶南方工程技术有限公司 | Multi-objective optimization method of steelmaking-continuous casting production scheduling based on NSGA-II |
| CN105929689A (en) * | 2016-04-22 | 2016-09-07 | 江南大学 | Machine tool manufacturing system processing and energy saving optimization method based on particle swarm algorithm |
Non-Patent Citations (3)
| Title |
|---|
| YUE SONG 等: ""A multi-objective model for regional bus timetable based on NSGA-II"", 《2012 IEEE INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND AUTOMATION ENGINEERING (CSAE)》 * |
| 王凯 等: ""基于pareto最优原理的钻机钻进参数多目标优化"", 《中国机械工程》 * |
| 黄晓敏 等: ""基于NSGA-Ⅱ的水文模型参数多目标优化研究"", 《人民长江》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110070235A (en) * | 2019-05-01 | 2019-07-30 | 湖南大学 | A kind of flexible dispatching method of multiple mobile robot |
| CN112181608A (en) * | 2019-07-03 | 2021-01-05 | 中南大学 | Distributed distribution algorithm for multipoint dynamic aggregation tasks based on local information |
| CN112181608B (en) * | 2019-07-03 | 2023-10-31 | 中南大学 | Distributed distribution method for multipoint dynamic aggregation tasks based on local information |
| CN111324429A (en) * | 2019-10-09 | 2020-06-23 | 中国人民解放军国防科技大学 | Micro-service combination scheduling method based on multi-generation ancestry reference distance |
| CN111324429B (en) * | 2019-10-09 | 2023-03-24 | 中国人民解放军国防科技大学 | Micro-service combination scheduling method based on multi-generation ancestry reference distance |
| CN113031650A (en) * | 2021-03-04 | 2021-06-25 | 南京航空航天大学 | Unmanned aerial vehicle cluster cooperative target distribution design method under uncertain environment |
| CN113031650B (en) * | 2021-03-04 | 2022-06-10 | 南京航空航天大学 | Unmanned aerial vehicle cluster cooperative target distribution design method under uncertain environment |
| CN113419854A (en) * | 2021-06-23 | 2021-09-21 | 东北大学 | Cloud resource scheduling method oriented to unbalanced multi-objective optimization |
| CN113419854B (en) * | 2021-06-23 | 2023-09-22 | 东北大学 | Unbalanced multi-objective optimization-oriented cloud resource scheduling method |
| CN116521344A (en) * | 2023-05-12 | 2023-08-01 | 广州卓勤信息技术有限公司 | AI algorithm scheduling method and system based on resource bus |
| CN116521344B (en) * | 2023-05-12 | 2023-10-03 | 广州卓勤信息技术有限公司 | AI algorithm scheduling method and system based on resource bus |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108334986B (en) | 2020-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108334986A (en) | A kind of multiple agent Multipurpose Optimal Method of the rescheduling method based on plastic mechanism | |
| Ewees et al. | Enhanced salp swarm algorithm based on firefly algorithm for unrelated parallel machine scheduling with setup times | |
| CN106020230B (en) | A kind of multiple no-manned plane method for allocating tasks under power consumption constraint | |
| Dai et al. | Multi-robot dynamic task allocation for exploration and destruction | |
| CN104880949B (en) | Method based on chicken group's algorithm acquisition work pieces process optimal scheduling is improved | |
| CN104408518B (en) | Based on the neural network learning optimization method of particle swarm optimization algorithm | |
| CN107831685A (en) | Group robot control method and system | |
| CN110533301A (en) | A kind of population dispatching method based on dynamic constrained matrix | |
| CN109857534A (en) | A kind of intelligent task scheduling strategy training method based on Policy-Gradient Reinforcement Learning | |
| CN103235743A (en) | Method for scheduling multi-target testing task based on decomposition and optimal solution following strategies | |
| CN108769182A (en) | A kind of prediction executes the Combinatorial Optimization dispatching method of task execution time | |
| JP2000020103A (en) | Evaluating method for genetic algorithm | |
| CN111160511A (en) | Group intelligent method for consensus active learning | |
| Yu et al. | A novel congruent organizational design methodology using group technology and a nested genetic algorithm | |
| CN109684088A (en) | The remote sensing big data of cloud platform resource constraint quickly handles method for scheduling task | |
| CN112734127A (en) | Multi-AUV task allocation method based on dynamic ant colony labor division model | |
| Liu et al. | UAV-USV cooperative task allocation for smart ocean networks | |
| CN108769105A (en) | A kind of scheduling system of knowledge services multi-task scheduling optimization method and its structure under cloud environment | |
| Liang et al. | Recent advances in particle swarm optimization via population structuring and individual behavior control | |
| Dorronsoro et al. | Evolutionary algorithms based on game theory and cellular automata with coalitions | |
| Ho et al. | Learning coordination strategies for cooperative multiagent systems | |
| Jianping et al. | Multi-robot task allocation based on robotic utility value and genetic algorithm | |
| Chen et al. | Group competition-cooperation optimization algorithm | |
| CN114118374A (en) | Multi-wisdom reinforcement learning method and system based on hierarchical consistency learning | |
| Bastos et al. | Variable utility in multi-robot task allocation systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200602 |