CN103345384A - Multi-target test task scheduling method and platform based on decomposed variable neighborhoods - Google Patents

Abstract

The invention discloses a multi-target test task scheduling method and platform based on decomposed variable neighborhoods, and belongs to the field of parallel test task scheduling. The platform comprises a user management module, a test task scheduling example input module, a database module, a test example importing module, a scheduling method module, a scheduling method extension module, a scheduling result decision module and a scheduling result displaying and outputting module. The platform extracts test task sample information from a database, calculation is conducted through the scheduling method based on the decomposed variable neighborhoods, decision is made on an obtained non-inferior solution through the analytic hierarchy process, and the scheduling result is output. The invention further provides the scheduling method extension module which is beneficial to the updating of the whole platform functions. The unified research platform is provided for test task scheduling and has good expandability. In addition, the invention provides a neighborhood variable strategy so that the individual cross range can be more reasonable, and the performance of the scheduling method is effectively improved.

Description

A kind of based on the change neighborhood multiple goal test assignment dispatching method and the platform that decompose

Technical field

The invention belongs to Auto-Test System concurrent testing task scheduling field, be specifically related to a kind of based on the change neighborhood multiple goal test assignment dispatching method and the platform that decompose.

Background technology

Fast development along with modern industry, the developing trend of electronic product is in integrated and complicated, the demand of test is more and more, and test macro resource needed and maintenance cost also constantly increase, and Auto-Test System has become the necessary means of modern equipment repair.The tradition Auto-Test System, serial test job pattern is continued to use in the restriction of tested person interface capacity and method of testing scheduling mode mostly, can not test equipment under test simultaneously, and test throughput is lower.So the serial test can not be satisfied actual requirement, concurrent testing has become the main trend of field of automatic testing.The concurrent testing technology depends on parallel processing technique, shows as a plurality of unit under tests and tests simultaneously under the concurrent testing scheduling mechanism.The concurrent testing technology improves the throughput of system by the quantity that increases measured piece in the unit interval, by improving usage ratio of equipment the standby time that reduces instrument and CPU, and by valuable equipment shared to save testing cost.The concurrent testing technology has improved testing efficiency effectively by the reasonable utilization to test resource, has reduced test and has consumed.

Concurrent testing requires different test assignments is selected suitable testing scheme, to reach the utilization of testing tool reasonable resources.The concurrent testing core technology is the test dispatching method, and the selection of dispatching method has very important effect for final output result.The concurrent testing problem belongs to typical NP-Complete problem, and solution space is huge, is difficult to obtain all disaggregation by exhaustive method.For the concurrent testing scheduling problem, the intelligent search method can reduce computing time effectively, and obtains approximate optimal solution or optimum solution.The method that is successfully applied to the concurrent testing scheduling problem at present has methods such as heredity, simulated annealing, ant group, figure taboo.Research mainly is conceived to test assignment method itself about concurrent testing mostly, and is confined to the single goal problem, is about to the test duration as final evaluation criterion.And along with the complexity of equipment increases, the maintenance cost of testing tool also increases thereupon.The decision maker should take all factors into consideration factors such as the shortest test duration, testing tool average load when formulating testing scheme.Therefore the research of multiple goal test assignment scheduling has great significance to parallel test system.

Multiple goal test assignment Research of Scheduling Method mainly concentrates on the intelligent optimization method.Present most of intelligent optimization method mostly adopts the weighted sum method that multi-objective problem is converted to the single goal problem when solving multi-objective problem, and being set in the whole search procedure of weighted sum parameter plays an important role.Be not suggested and have a kind of generally acknowledged method how the weighted sum parameter effectively is set at present, human factor will influence the effect of intelligent optimization method greatly.

Lack a complete design platform for concurrent testing research both at home and abroad at present and realize concurrent testing scheduling holistic approach.Therefore, build a complete multiple goal test assignment dispatching platform with extensibility, for improving testing efficiency, accelerating concurrent testing task scheduling research progress has vital role.

Summary of the invention

Lack the problem that complete design platform is realized concurrent testing scheduling holistic approach at having now, the invention provides a unified multiple goal test assignment dispatching platform.In order to solve the problem that existing dispatching method is absorbed in local optimum, bad adaptability easily, the present invention proposes a kind of based on the change neighborhood multiple goal test assignment dispatching method that decomposes simultaneously.

A kind of multiple goal test assignment dispatching platform provided by the invention comprises as lower module: user management module, test assignment scheduling example interpolation and maintenance module, database module, test case import module, dispatching method module, dispatching method expansion module, scheduling result decision-making module and scheduling result and show output module.

User management module is used for information and the password of leading subscriber, is the entrance of test assignment dispatching platform.Test assignment scheduling example interpolation and maintenance module add for five kinds of information tables to test and safeguard, and store data in the database module; Described five kinds of information tables are: the temporal constraint relation table between test assignment collection table, test assignment table, instrument resource table, test assignment and testing scheme collection table.Test case imports the test assignment collection that module is selected according to the user, obtains the needed information table of scheduling from database module, and outputs to the dispatching method module.Integrated more than one dispatching method in the dispatching method module, the user chooses a kind of dispatching method from the dispatching method module, and the parameter of selected dispatching method set, dispatching method module operation test assignment collection is with all noninferior solution input schedulings that obtain decision-making module as a result.The scheduling result decision-making module carries out final decision by analytical hierarchy process to noninferior solution, and the output optimum solution shows output module to scheduling result.Scheduling result shows that output module is shown to the user with the objective function point diagram of optimum solution and corresponding scheduling Gantt chart.The dispatching method expansion module is used for new dispatching method is added the dispatching method module.

Be integrated with provided by the invention based on the change neighborhood multiple goal test assignment dispatching method that decomposes in the dispatching method module.

Provided by the invention based on the change neighborhood multiple goal test assignment dispatching method that decomposes, specifically comprise the steps:

Step 1: obtain the information of selected test assignment collection, comprising:

Test assignment table r={t ₁, t ₂,, t _j..., t _N, t _jBe j test assignment, N is the test assignment sum;

Instrument resource table R={r ₁, r ₂,, r _i..., r _M, r _iBe i instrument, M is the instrument sum;

Testing scheme collection table, test assignment t _jThe testing scheme collection be defined as

Be test assignment t _jK _jIndividual testing scheme, k _jBe test assignment t _jThe testing scheme sum, j=1,2 ..., N;

And the temporal constraint relation table between test assignment;

Step 2: initialization comprises step 2.1: determine objective function～step 2.4;

Step 2.1: determine objective function: test duration f ₁(x) and each step average load f ₂(x);

Test duration

Wherein,

Expression test assignment t _jSelection scheme

The time deadline;

If d represents parallel step number, initial value is 1, after all test assignments are all arranged instrument resource, when the instrument resource of two test assignments of every existence has repetition, upgrades d=d+1, calculates total parallel step number, determines respectively to go on foot average load then:

$f_2\left(x\right)=\frac{1}{d}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{P}_j^i{Q}_j^i;$

Wherein,

Expression test assignment t _jAt instrument r _iElapsed time when carrying out;

Expression test assignment t _jWith instrument r _iDemand relation, as test assignment t _jNeed take instrument r _iSituation under,

Otherwise

Step 2.2: variable and parameter initialization;

The set that note is preserved noninferior solution is EP, and initial

Initialization reference point optimal solution set z=(z ₁, z ₂) ^T, z ₁=min{f ₁(x), x ∈ Ω }, z ₂=min{f ₂(x), x ∈ Ω }, wherein Ω represents solution space.

Initialized parameter comprises: maximum iteration time M'; Population size N' also claims the weight vector number; Neighborhood variation range, maximal value are T _High, minimum value is T _LowCrossover probability CR and variation Probability p; When iterations is ger, the field in corresponding generation size T=floor (((T _High-T _Low)/(M') ²) * (ger-M') ²)+T _Low, floor represents to round downwards.

Step 2.3: calculate weight vector neighborhood indexed set; Find and i T the weight vector that weight vector is nearest, defining i weight vector neighborhood indexed set is B (i)={ i ₁,, i _T, i ∈ [1, N'].

Step 2.4: produce initial solution; Produce initial population at random, and determine the value of the objective function of each individual correspondence; Comprise N decision variable in each individuality, wherein the execution time of j test assignment of the more big expression correspondence of the value of j decision variable is more forward.

Step 3, the individual intersection; I the individuality of t in generation

The individuality that produces after intersecting

For:

$x_{t+1}^i=\left\{\begin{array}{cc}{x}_t^i+F_1\&times;(x_t^i-x_t^{i1})+{\mathrm{<figure-callout\; id="2"\; label="F"\; filenames="HDA00003533436800042.png"\; state="\{\{state\}\}">F</figure-callout>}}_2\&times;(x_t^i-x_t^{i2})& \mathrm{rand}\left(1\right)<\mathrm{CR}\\ x_t^i& \mathrm{rand}\left(1\right)\&GreaterEqual;\mathrm{CR}\end{array}\right.$

Wherein,

With

Be the weight vector of random choose among the weight vector B (i), constant F ₁With F ₂All be set to 1, rand (1) and for variation range be 0 to 1 decimal at random.

Step 4: individual variation;

If i the individuality in t generation is

Each decision variable in the individuality

Make a variation according to following formula:

$x_{\left(j\right)}^{i*}=\left\{\begin{array}{cc}\mathrm{normal}(x_{\left(j\right)}^i,\mathrm{\&sigma;})& \mathrm{rand}\left(1\right)<p\\ x_{\left(j\right)}^i& \mathrm{rand}\left(1\right)\&GreaterEqual;p\end{array}\right.,j=\mathrm{1,2},...,N$

Wherein,

Be that an obedience average is

Variance is the number of the normal distribution of σ, and σ is set to 1/20 of decision variable variation range;

Expression x ⁱNew individuality after the variation.

Step 5: upgrade neighborhood and reference point;

For j ∈ B (i), the definition reference point is z _iThe fitness of j subproblem be

Wherein

Represent i j weight vector value in the weight vector neighborhood indexed set; If individual y' is the solution that obtains behind i the individual variation, to j ∈ B (i), if fitness value F (y')≤F (x ^j), the individual x of new population more then ^j=y', F (x ^j)=F (y'); Otherwise, keep the individual x of population ^jConstant.

Upgrade z: to any i=1,, if m is z _i＜f _i(y'), assignment z then _i=f _i(y').

Step 6: preserve noninferior solution; After each is finished for evolution, all noninferior solutions that obtain are saved among the set EP, and the inferior solution that will gather among the EP is removed;

If two are separated and target function value is respectively: A ₁(f ₁(A ₁), f ₂(A ₁)), A ₂(f ₁(A ₂), f ₂(A ₂)); If f ₁(A ₁) 〉=f ₁(A ₂) and f ₂(A ₁) 〉=f ₂(A ₂), then separate A ₁Be inferior solution, A is separated in deletion ₁

Step 7: carry out final decision by the noninferior solution among the analytical hierarchy process pair set EP;

Step 8: final plan is exported.

Described step 7 specific implementation method is: set up the tri-layer structure: top is the final decision result, and the middle layer is dimension n ₂=2 objective function, the bottom is the noninferior solution among the set EP, the noninferior solution number is n ₁

At first, each noninferior solution of bottom importance about the objective function of each dimension of middle layer is compared in twos, the structure judgment matrix, matrix is n ₁* n ₁Matrix; For a certain dimension, the capable j column element of i a in the judgment matrix _IjA is separated in expression _iWith respect to separating A _jSignificance level, f _iAnd f _jBe to separate A _iConciliate A _jAt the target function value of corresponding dimension, f _MaxAnd f _MinBe respectively maximum and the minimum value of objective function on the corresponding dimension; Element a _IjValue be:

Wherein, round represents the result of calculation fraction part is rounded up.

Then, judgment matrix is carried out consistency check, specifically: calculate coincident indicator CI=(λ earlier _Max-n)/(n-1), λ _MaxBe the eigenvalue of maximum of judgment matrix, n is the exponent number of judgment matrix; Calculate Consistency Ratio CRT=CI/RI again, RI is the mean random coincident indicator; If CRT≤0.10, then judgment matrix is by consistency check, otherwise needs the correction judgment matrix.

After consistency check is passed through, to the eigenvalue of maximum characteristic of correspondence vector u=(u of judgment matrix ₁, u ₂..., u _n) ^T, carry out normalized, obtain the relative ordering weight vectors w=(w of element under the corresponding single criterion ₁, w ₂..., w _n) ^T,

Corresponding two objective functions, weight vectors w obtains sorting ³

Equally, according to the importance of objective function to the final decision result, set up the middle layer for top judgment matrix, matrix is n ₂* n ₂Matrix; Obtain the synthetic weight vector W in middle layer ², finally obtain bottom synthetic weight vector

W ³=w ³W ²Find the maximal value in the bottom synthetic weight vector, the corresponding scheme of this value is exactly optimal case.

Major advantage of the present invention is:

(1) this platform carries out maintenances of user management, test assignment scheduling example, the scheduling of multiple goal test assignment etc. integrated, and resource information has been carried out highly integration, is convenient to maintenance and the upgrading of system.

(2) platform is with good expansibility, and the extension mechanism of dispatching method is provided, and is convenient to the lifting of platform and integrally performance.

(3) make individual crossover range more reasonable based on the change neighborhood test assignment dispatching method that decomposes, improve the quality of understanding effectively, obtain more excellent test assignment scheduling sequence.

Description of drawings

Fig. 1 is the structural representation of multiple goal test assignment dispatching platform of the present invention;

Fig. 2 is of the present invention based on the change neighborhood multiple goal test assignment dispatching method process flow diagram that decomposes;

Fig. 3 is the control curve synoptic diagram that becomes neighborhood size in the neighborhood multiple goal test assignment dispatching method among the present invention;

Fig. 4 is the synoptic diagram of differentiating noninferior solution among the present invention;

Fig. 5 is the hierarchical structure synoptic diagram of the middle-level analytic approach of the present invention;

Fig. 6 is net result output synoptic diagram among the present invention.

Wherein:

1A. user management module 2B. test assignment scheduling example adds and maintenance module 3C. database module

4D. test case imports module 5E. dispatching method module 6F. dispatching method expansion module

7G. scheduling result decision-making module 8H. scheduling result shows output module

Embodiment

The present invention is described in further detail below in conjunction with drawings and Examples.

The present invention is a kind of concurrent testing task scheduling platform, the structure of this platform comprises that user management module 1A, test assignment scheduling example interpolation and maintenance module 2B, database module 3C, test case import module 4D, dispatching method module 5E, method of testing expansion module 6F, scheduling result decision-making module 7G and scheduling result and show output module 8H as shown in Figure 1.

User management module 1A is the entrance of test assignment dispatching platform.The user can register new user, deletion user profile, revise user profile in this module, and carries out password modification etc.

Test assignment scheduling example adds and maintenance module 2B is connected with database module 3C.The information that test assignment scheduling example adds and maintenance module 2B can import comprises temporal constraint relation table and testing scheme collection table five part between test assignment collection table, test assignment table, instrument resource table, test assignment, and the interpolation that every partial information can be corresponding, deletion action are to adapt to various scheduling requirements.The data of input are saved to database module 3C.

In test assignment collection table, can add the test assignment collection, and be described, user-friendly.As shown in table 1:

Table 1 test assignment collection table

The task set name	Remarks
		None_Set1	There are not constraint 6 tasks 8 resources
None_Set2	There are not constraint 20 tasks 8 resources
		None_Set3	There are not constraint 40 tasks 12 resources

Strong_Set1	Strong constraint	20 tasks 8 resources
			Srong_Set2	Strong constraint	40 tasks 12 resources

Task number and the resource needed quantity having indicated this task-set in the remarks of each task-set and whether have restriction relation, comprise.Each task-set is to having a test assignment table, instrument resource table and testing scheme collection table.

Be example with task-set None_Set1, the test assignment table of this task-set correspondence, instrument resource table and testing scheme collection table are respectively shown in following table 2, table 3 and table 4.The user can make amendment and adds equally the information of each table.

Table 2 test assignment table

The task set name	Task ID	Test assignment is described
			None_Set1	1	Direct-flow signal voltage
None_Set1
		2	Receiving sensitivity
None_Set1				3	The AC signal frequency
	None_Set1
		4	The AC signal amplitude
	None_Set1
		5	Resistance 1 over the ground
	None_Set1			6	Resistance 2 over the ground

Comprising task set name, task ID and test assignment in the test assignment table describes.

Table 3 instrument resource table

Test resource ID	Instrument resource is described
		1	Direct supply
2	Channel oscilloscope 1
		3	Channel oscilloscope 2
4	Function generator
		5	Radio-frequency signal source
6	The Modulation analysis instrument
		7	Counter
8	Frequency spectrograph

Comprise the description of test resource ID and instrument resource in the instrument resource table.

Table 4 testing scheme collection table

The test assignment collection	Task ID	Scheme ID	Used instrument 1ID	Used instrument 2ID	Test duration
						None_Set1	1	1	1	4	2
None_Set1	1	2	3	5	5
						None_Set1	2	1	2	8	3

None_Set1	2	2	1	5	4
						None_Set1	3	1	1	3	6
None_Set1	4	1	3	7	10
						None_Set1	4	2	1	2	8
None_Set1	5	1	2	5	13
						None_Set1	6	1	1	6	17
None_Set1	6	2	3	6	22

Comprise test assignment collection, task ID, scheme ID, used instrument and test duration in the testing scheme collection table.Wherein, scheme refers to set two kinds of schemes in the table 4 at the testing scheme of this test assignment collection setting.Be example with first testing scheme in the testing scheme collection table, the test assignment 1 of expression test assignment collection None_Set1 is selected first kind of testing scheme, and selected instrument is instrument 1 and instrument 4, and the used test duration is 2 seconds.

The restriction relation of test assignment collection is recorded in the temporal constraint relation table between test assignment, is example with the test assignment collection Strong_Set1 in the table 1, and the temporal constraint relation table between test assignment is as shown in table 5.

Temporal constraint relation table between table 5 test assignment

The test assignment collection	Preceding item constraint	Back item constraint
			Strong_Set1
	2	13
			Strong_Set1	2	15
Strong_Set1	3	11
			Strong_Set1	4	16
Strong_Set1	5	7

Comprise test assignment collection, preceding item constraint and back item constraint in the temporal constraint relation table between test assignment, with the preceding item constraint 2 in first row in the table 5, back constraint 13 is example, task 13 among the expression test assignment collection Strong_Set1 only just can be carried out after task 2 is finished, and has embodied the temporal constraint that test assignment is concentrated.

Database module 3C can dispatch that example adds and maintenance module 2B obtains data message from test assignment, and is stored in the database.

Test case imports and connects database module 3C and dispatching method module 5E among the module 4D.Test case imports the test assignment collection that module 4D selectes according to the user, from database module 3C, obtain temporal constraint relation table and testing scheme collection table between the needed test assignment table of scheduling, instrument resource table, test assignment, and it is outputed to dispatching method module 5E.

Integrated more than one dispatching method among the dispatching method module 5E, the user can select each integrated dispatching method of platform, and selected dispatching method parameter is set.The selected test assignment collection of dispatching method module 5E operation behind end of run, is preserved also input scheduling as a result decision-making module 7G with all noninferior solutions that obtain (non-domination solution) at dispatching method.Change neighborhood multiple goal test assignment dispatching method based on decomposition provided by the invention also is integrated among the dispatching method module 5E.

Dispatching method expansion module 6F provides a kind of extension mechanism for dispatching method module 5E, new dispatching method can be added among the dispatching method module 5E, and the various aspects of performance of initiate dispatching method and original method can be compared, select optimum dispatching method as platform master scheduling method, improve the performance of whole platform.

Scheduling result decision-making module 7G is connected with dispatching method module 5E.After the dispatching method evolutionary process is finished, by analytical hierarchy process a plurality of noninferior solutions are carried out final decision.

Scheduling result shows that output module 8H shows output with the final optimal solution that scheduling result decision-making module 7G determines.The corresponding objective function point diagram of output optimum solution and corresponding scheduling Gantt chart thereof.

It is a kind of based on the change neighborhood multiple goal test assignment dispatching method that decomposes that the present invention also provides.Based on the change neighborhood Multipurpose Optimal Method of decomposing a multi-objective optimization question is decomposed into one group of sub-optimization problem by weight vector in its objective function space.Finish renewal by the information that obtains its neighborhood between the sub-optimization problem during evolution and evolve, and a plurality of target is parallel optimization.Adopt this optimal way can avoid the use of weighted sum method, thereby reduced the influence of human factor in optimizing the result, and the employing that becomes the neighborhood strategy can make the individual crossover range more reasonable, improve the quality of scheduling result solution.

As shown in Figure 2, of the present invention a kind of based on the change neighborhood multiple goal test assignment dispatching method that decomposes, specifically comprise the steps:

Step 1: obtain the test assignment instance data.

Existing test assignment example calls selection in the database to multiple goal test assignment dispatching platform.As selecting None_Set1(not have constraint 6 tasks 8 resources), Strong_Set1(strong constraint 20 tasks 8 resources), Strong_Set2(strong constraint 40 tasks 12 resources) etc. the test assignment collection.And obtain the information table of selected test assignment collection related data.The information that obtains comprises:

Test assignment table T={t ₁, t ₂, t _j..., t _N, t _jBe j test assignment, N is the test assignment sum;

Instrument resource table R={r ₁, r ₂,, r _i... r _M, r _iBe i instrument, M is the instrument sum;

Testing scheme collection table, test assignment t _jThe testing scheme collection be defined as

K wherein _jBe test assignment t _jThe testing scheme sum, j=1,2 ..., N;

And the temporal constraint relation table between test assignment, as for test assignment collection Strong_Set1, there is preceding item constraint 2, back item constraint 13, i.e. test assignment t ₁₃Will be at test assignment t ₂Just can carry out after finishing.

Step 2: dispatching method initialization.

The dispatching method initialization comprises definite objective function, dispatching method variable and parameter initialization, calculating weight vector neighborhood indexed set and produces steps such as initial solution.

Step 2.1: determine objective function.

Two objective functions are distributed as test duration C in multi-task scheduling platform of the present invention ₁With each step average load C ₂Objective function is described below:

After the test assignment example imports, obtain test assignment table T={t _l, t ₂,, t _j..., t _N, instrument resource table R={r ₁, r ₂,, r _i..., r _M, the temporal constraint relation table between testing scheme table and test assignment.

Suppose

With

Represent respectively at test assignment t _jAt instrument resource r _iZero-time when carrying out, deadline and elapsed time have

In automatic test dispatching, there is the common cooperation that often needs a plurality of test resources of finishing of a test assignment.Therefore, can be with a judgment matrix The demand relation of expression instrument resource and test assignment.Judgment matrix is defined as:

In general, test assignment t _jHaving a plurality of optional testing schemes selects.t _jThe testing scheme collection be defined as

With

Expression test assignment t _jSelect testing scheme

The test elapsed time, r _iThe expression testing scheme

In resource.

Resource constraint between the testing scheme collection can be expressed as:

Wherein

With

Represent task t respectively _jK Scheme Choice and task

K ^*Individual Scheme Choice.

Therefore can define task t _jSelection scheme

The time deadline be

r _iThe expression testing scheme

In resource.So objective function 1:

Test duration $C_1=f_1\left(x\right)=\underset{1\&le;j\&le;N}{\underset{1\&le;k\&le;k_j}{\max }}{C}_j^k$

Wherein, N is the test assignment sum, k _jBe test assignment t _jThe testing scheme sum.

Use symbol d to represent parallel step number.The initial value of d is set to 1.After all arranging test resource for all tasks, if D=d+1 then.Calculate parallel total step number, so objective function 2 is as follows:

Each goes on foot average load $C_{}$ ${}_2=f_2\left(x\right)=\frac{1}{d}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{P}_j^i{Q}_j^i$

Suppose Expression test assignment t _jSelect testing scheme

The test elapsed time, for simplify calculating, in calculating target function 2, get

Judgment matrix

The demand relation of expression resource and task, N is the test assignment sum, M is the instrument resource sum.

Step 2.2, variable and parameter initialization.

The set that note is preserved noninferior solution is EP, and initial

Initialization reference point optimal solution set z=(z ₁, z ₂) ^T, z ₁=min{f ₁(x), x ∈ Ω }, z ₂=min{f ₂(x), x ∈ Ω }; Wherein Ω represents solution space.Be z _iBe the theoretical minimum value of each objective function in field of definition Ω, wherein z is temporary transient optimum solution, changes along with evolutionary process.

Initialized parameter comprises: iterations M'; Population size N' also claims the weight vector number; The neighborhood variation range, maximum of T _HighAnd minimum value T _LowCrossover probability CR and variation Probability p.Each parameter can change according to the variation that test assignment is dispatched example, and the neighborhood variation range changes according to the population size variation.Retraining 20 tasks, 8 resource testing examples with nothing is example, and the dispatching method parameter arranges as shown in table 6:

Table 6: the dispatching method basic parameter arranges

Iterations	The population size	Neighborhood maximums	Neighborhood minimum	Crossover probability	The variation probability
						200	200	30	10	0.5	0.1

The neighborhood size changes along with evolutionary process in this dispatching method.Evolving, neighborhood is more greatly to guarantee that individuality intersects fully in earlier stage, and the later stage neighborhood of evolving is less to prevent the individuality degeneration.Neighborhood changes the control curve as shown in Figure 3:

Among the figure, horizontal ordinate is iterations, and ordinate is the neighborhood size.The control curve is second-degree parabola, and curve slope when evolving end (being that iterations is 200) is 0.It is fast in the variation in early stage of evolving to have embodied neighborhood, and the later stage changes slow, and pace of change is 0 characteristics when evolving end.Suppose that each is T for the neighborhood size, corresponding iterations is ger, and the neighborhood variation range is 30 to 10.T=floor (0.0005* (ger-200) is then arranged among the figure ²+ 10), floor represents to round downwards.

Step 2.3: calculate weight vector neighborhood indexed set.

Calculate and i T the weight vector neighborhood indexed set that weight vector is nearest.Wherein indexed set is designated as B (i)={ i ₁,, i _T, note λ ⁱBe i weighted value in the equally distributed N' weight vector, i ∈ [1, N'],

Be λ ⁱT nearest weighted value, distance is determined by its Euclidean distance between two weight vector, N' is for based on decomposing and the number of the subproblem of the multiple goal test assignment dispatching method of optimum solution follow-up strategy is the population size, and T is the neighborhood size for the quantity of the nearest weight vector of the whenever single weight vector of distance.

Suppose that population number is 50, then the weight vector collection is: λ ¹={ 1/50,49/50}, λ ²={ 2/50,48/50} ..., λ ⁵⁰={ 1,0}.

With weight vector λ ¹⁰={ 10/50,40/50} is example, is 10 as if neighborhood size in this iteration, then Dui Ying weight vector field indexed set B (10)=(5,6,7,8,9,11,12,13,14,15).

Step 2.4: produce initial solution.

Produce initial population at random and be designated as x ¹,, x ^N', and to make the solution of each individual corresponding objective function be f _i(x ^l), i ∈ [1,2] wherein, l ∈ [1, N'].Comprise N decision variable in each individuality, N is current test assignment sum, and the span of decision variable is 0～1 in the embodiment of the invention, and the execution time of j the test assignment that the more big expression of the value of j decision variable is corresponding is more forward.

Be example with test assignment collection None_Set1, population size N' is set to 50.Suppose in the initial population that body is x one by one ¹=0.125,0.325,0.245,0.865,0.742, The number of data equals the concentrated task sum of test assignment collection in the individuality.Data represent the decision variable of test assignment respectively, in order to determine the priority execution sequence of test assignment.For the test assignment collection that has temporal constraint, the task sequencing that then obtains need be checked with temporal constraint, if do not satisfy temporal constraint, then decision variable need adjust to meet temporal constraint.Then six corresponding decision variables of test assignment are followed successively by 0.125,0.325,0.245,0.865,0.742,0.631.It is t that the sequential of definite each test assignment that sorts can get six test assignment sequential relationships ₄＞t ₅＞t ₆＞t ₂＞t ₃＞t ₁Utilizing formula k=[x _j* 10] mod k _j+ 1 instrument resource that obtains corresponding test assignment is selected, and wherein k is selected testing scheme ordinal number, x _jBe the decision variable of test assignment correspondence, k _jBe the optional testing scheme sum of test assignment.With task t ₂Be example, k=[0.325 * 10] mod2+1=2, then the instrument resource of selected testing scheme correspondence is $w_2^2=\{r_1,r_5\}.$

Step 3: the individual intersection.

Note in t generation an individuality be The individuality that produces after intersecting

As follows:

$x_{t+1}^i=\left\{\begin{array}{cc}{x}_t^i+F_1\&times;(x_t^i-x_t^{i1})+{\mathrm{<figure-callout\; id="2"\; label="F"\; filenames="HDA00003533436800042.png"\; state="\{\{state\}\}">F</figure-callout>}}_2\&times;(x_t^i-x_t^{i2})& \mathrm{rand}\left(1\right)<\mathrm{CR}\\ x_t^i& \mathrm{rand}\left(1\right)\&GreaterEqual;\mathrm{CR}\end{array}\right.$

Wherein,

With

Be the weight vector of random choose among the weight vector B (i), constant F ₁With F ₂Generally all be set to 1, rand (1) and for variation range be a decimal at random of 0 to 1.Be example with 8 resources of 6 tasks, neighborhood size T is set to 10.Individual Corresponding weight vector index value is 10, the individuality of random choose in the field of correspondence $x_t^{i1}=\left\{\mathrm{0.130,0.646,0.345,0.318,0.354,0.693}\right\}$ With ${\mathrm{<figure-callout\; id="2"\; label="x\; t\; i"\; filenames="HDA00003533436800042.png"\; state="\{\{state\}\}">x</figure-callout>}}_{ti}^2=\left\{\mathrm{0.153,0.682,0.315,0.363,0.387,0.623}\right\},$ Its corresponding weight vector index value is 7,12.Rand (1)=0.7＞CR is so after interlace operation, obtain according to crossing formula

$x_{t+1}^i=\left\{\mathrm{0.077,0.670,0.465,0.093,0.444,0.826}\right\}.$

Step 4: individual variation, Gaussian mutation is adopted in variation.

For t i the individuality in generation

Each decision variable in the individuality

All adopt following Gaussian mutation operator to obtain:

$x_{\left(j\right)}^{i*}=\left\{\begin{array}{cc}\mathrm{normal}(x_{\left(j\right)}^i,\mathrm{\&sigma;})& \mathrm{rand}\left(1\right)<p\\ x_{\left(j\right)}^i& \mathrm{rand}\left(1\right)\&GreaterEqual;p\end{array}\right.,j=\mathrm{1,2},...,N$

Wherein,

Expression x ⁱNew individuality after the variation,

Be the number of a Normal Distribution, wherein

Be average, σ is variance, and σ is set to 1/20 of decision variable variation range in natural number coding.Rand (1) is a decimal at random of 0 to 1 for variation range, and p is traditionally arranged to be 0.05.

Be x before the individual variation ⁱ={ 0.047,0.580,0.432,0.076,0.444,0.826}, the decision variable variation range is to be 0.05 by 0 to 1, σ in this example.The new individuality of gained is after the Gaussian mutation

x ^i*＝{0.063,0.629,0.483,0.120,0.431,0.842}。

Step 5: upgrade neighborhood and reference point.

Upgrade neighborhood: establishing y' is the solution that obtains behind i the individual variation; For j ∈ B (i), the definition reference point is z _iThe fitness value of j subproblem be

M=2 is the objective function dimension, wherein

Be j weight vector value in equally distributed i the weight vector neighborhood indexed set.

For minimization problem, the more little Xie Yueyou that then represents of the fitness value that solution obtains is to j ∈ B (i), if fitness value F (y')≤F (x ^j), the individual x of new population more then ^j=y', F (x ^j)=F (y'), otherwise, keep the population individuality constant.

Upgrade z: to any i=1,, if m is z _i＜f _i(y'), assignment z then _i=f _i(y'), z _iBe any optimum solution.

Step 6: preserve noninferior solution in set EP.

Dispatching method is preserved all noninferior solutions that obtain (non-domination solution) after each is finished for evolutionary process, and itself and the noninferior solution of preserving are before merged.Solution after merging is concentrated inferior solution is removed, and keeps noninferior solution to EP.As shown in Figure 4, suppose that dispatching method finally obtains 4 and separates A ₁, A ₂, A ₃And A ₄, corresponding target function value is distributed as A ₁(10,20), A ₂(15,15), A ₃(20,10), A ₄(13,22).Separate in the corresponding objective function point, for A for 4 as can be seen ₁, A ₂And A ₃, the target function value of neither one solution can be all littler than it on two dimensions.And for A ₄, A ₁Two dimensions on value all littler than it.So for the minimum value optimization problem, separate A ₁, A ₂And A ₃Be noninferior solution, separate A ₄Be inferior solution.A is separated in deletion ₄, the solution under final the preservation is A ₁, A ₂And A ₃

Step 7: scheduling result decision-making.

After the dispatching method evolutionary process is finished, by analytical hierarchy process a plurality of noninferior solutions of final reservation are carried out final decision.Analytical hierarchy process is at first set up the hierarchical structure of problem.Top is general objective, and the criterion of the required consideration of problem analysis, sub-criterion are represented in the middle layer, and the bottom is expressed as realizes the alternative various schemes of target, measure etc., is also referred to as solution layer.The hierarchical structure of the Auto-Test System task scheduling problem analytical hierarchy process that the present invention sets up as shown in Figure 5.Wherein, rule layer C is divided into test duration C ₁With each step average load C ₂, solution layer A is made of the noninferior solution that dispatching method obtains.

To compare the structure judgment matrix in twos with each element of one deck importance about a certain criterion in the last layer.In traditional analytical hierarchy process, generally all construct judgment matrix by the mode of marking, relatively Chang Yong marking scale is 1～9 scale, and is as shown in table 7:

The definition of table 7 judgment matrix scale

Scale	Implication
		1	No less important
3	The former is important slightly
		5	The former is obviously important
7	The former is important strongly
		9	The former is extremely important
2,4,6,8	The intermediate value of above-mentioned adjacent judgement
		Reciprocal	The latter is than the former importance scale

Because the influence of marking scale human factor is very big, can produce the inaccurate even conflict of judgement meeting more for a long time in alternatives, cause being difficult to satisfy Consistency Check in Judgement Matrix.Therefore the present invention has adopted a kind of change yardstick judgment matrix generating operator to generate judgment matrix automatically.It can be with the decision information quantification, makes the uncertainty of analytical hierarchy process in can the control decision process, easily by consistency check, need not to consider that scale selects problem.Suppose each dimension for objective function, wherein f _MaxAnd f _MinBe respectively maximum and the minimum value of objective function on this dimension.a _IjBe the element in the judgment matrix, shown option A _iWith respect to option A _jSignificance level, f _iAnd f _jIt is option A _iAnd option A _jThe target function value of a certain dimension.Two objective functions are arranged among the present invention, and the objective function dimension is 2, the element a of each dimension correspondence _IjDetermine with following formula.

Wherein round represents the result of calculation fraction part is rounded up.

Each dimension objective function is all calculated a judgment matrix.After obtaining judgment matrix, at first judgment matrix is carried out consistency check.Utilize the eigenvalue of maximum λ of judgment matrix _MaxCalculate coincident indicator CI, CI=(λ _Max-n)/(n-1).N is the exponent number of judgment matrix, is that 2, the three layers of n of objective function number are the number of the noninferior solution of preservation for second layer n of the present invention.Calculate Consistency Ratio CRT then, CRT=CI/RI.Wherein RI is the mean random coincident indicator, and 1 to 15 rank mean random coincident indicator table is as shown in table 8.

Table 8:1 rank～15 rank mean random coincident indicator

Exponent number	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
																RI	0	0	0.52	0.89	1.12	1.26	1.36	1.41	1.46	1.49	1.52	1.54	1.56	1.58	1.59

If judgment matrix is then thought by consistency check in CRT≤0.10, otherwise need to revise judgment matrix.

After consistency check is passed through, be compared key element for the relative weighting of this criterion by judgment matrix calculating.If eigenvalue of maximum λ _MaxThe characteristic of correspondence vector is u=(u ₁, u ₂..., u _n) ^TWith the u normalized, namely for i=1,2 ..., n asks Obtain w=(w ₁, w ₂..., w _n) ^TBe the relative ordering weight vectors of element under the single criterion.

After obtaining the relative ordering weight vectors of element under the single criterion, to single each layer of calculating element to the synthetic weight of the aims of systems line ordering of going forward side by side.Suppose (k-1) laminated one-tenth weight vector W ^(k-1)And the relative ordering weight vectors w of element under the single criterion of k layer ^kKnown, the laminated one-tenth weight vector of k W then ^k=w ^kW ^(k-1), k＞1 obtains each element of solution layer about the ordering weight of general objective.Second layer synthetic weight vector W among the present invention ²Directly calculated by judgment matrix, the 3rd layer is passed through formula W later on ^k=w ^kW ^(k-1), k＞1 iteration obtains.

The one group of noninferior solution that utilizes a certain dispatching method to obtain with nothing constraint 20 tasks 8 resource testing system task scheduling problem examples is [(39,14.357), (34,19.546), (45,14.286), (52,13.667), (38,15.539), (32,20.000), (33,19.636), (35,18.182), (59,13.533), (65,12.813)].Hypothetical target function 1 is that 5(is that target 1 is obviously important for the significance level of objective function 2), then obtain the middle layer about top judgment matrix, as shown in table 9:

Table 9 judgment matrix Goal-C

Goal	C 1	C 2
			C 1	1	5
C 2	1/5	1

Can obtain eigenwert is λ ₁=2, λ ₂=0, λ _Max=λ ₁=2.CI=0, CRT=0＜0.1 is by consistency check.Correspondence can obtain second layer synthetic weight vector W ²=w ²=(0.833,0.167) ^T

Become yardstick judgment matrix generating operator by judgment matrix and generate judgment matrix automatically, the judgment matrix that obtains respectively with respect to objective function 1 and objective function 2 is shown in table 10 and table 11.

Table 10 judgment matrix C ₁-A

C 1	A 1	A 2	A 3	A 4	A 5	A 6	A 7	A 8	A 9	A 10
											A 1	1	1/2	2	4	1	1/3	1/2	1/2	6	7
A 2	2	1	4	5	2	1	1	1	7	9
											A 3	1/2	1/4	1	3	1/3	1/4	1/4	1/3	4	6
A 4	1/4	1/5	1/3	1	1/4	1/6	1/6	1/5	3	4
											A 5	1	1/2	3	4	1	1/2	1/2	1/2	6	8
A 6	3	1	4	6	2	1	1	2	8	9
											A 7	2	1	4	6	2	1	1	1	7	9
A 8	2	1	3	5	2	1/2	1	1	7	8
											A 9	1/6	1/7	1/4	1/3	1/6	1/8	1/7	1/7	1	2
A 10	1/7	1/9	1/6	1/4	1/8	1/9	1/9	1/8	1/2	1

Table 11 judgment matrix C ₂-A

C 2	A 1	A 2	A 3	A 4	A 5	A 6	A 7	A 8	A 9	A 10
											A 1	1	7	1	1/2	2	7	7	5	1/2	1/3
A 2	1/7	1	1/7	1/8	1/5	2	1	1/3	1/8	1/8
											A 3	1	7	1	1/2	2	7	7	5	1/2	1/3
A 4	2	8	2	1	3	8	8	6	1	1/2
											A 5	1/2	5	1/2	1/3	1	6	6	4	1/3	1/4
A 6	1/7	1/2	1/7	1/8	1/6	1	1	1/3	1/8	1/9
											A 7	1/7	1	1/7	1/8	1/6	1	1	1/3	1/8	1/9
A 8	1/5	3	5	1/6	1/4	3	3	1	1/6	1/7
											A 9	2	8	2	1	3	8	8	6	1	1/2
A 10	3	8	3	2	4	9	9	7	2	1

For judgment matrix C ₁-A can get λ _Max=10.338, CI=0.038, CRT=0.025＜0.10 is by consistency check.Can get the 3rd layer of ordering weight vectors relatively:

$w_{C1}^3={\left(\mathrm{0.093,0.167,0.056,0.033,0.103,0.195,0.170,0.150,0.019,0.014}\right)}^T.$

For judgment matrix C ₂-A can get λ _Max=10.404, CI=0.045, CRT=0.030＜0.10 is by consistency check.Can get the 3rd layer of ordering weight vectors relatively:

${\mathrm{<figure-callout\; id="2"\; label="w\; C"\; filenames="HDA00003533436800042.png"\; state="\{\{state\}\}">w</figure-callout>}}_C^2={\left(\mathrm{0.115,0.020,0.115,0.174,0.078,0.017,0.018,0.034,0.174,0.256}\right)}^T.$

Therefore can take all factors into consideration the ordering weight vectors that two dimensions of objective function obtain is w ³=(0.093,0.167,0.056,0.033,0.103,0.195,0.170,0.150,0.019,0.014;

。

0.115,0.020,0.115,0.174,0.078,0.017,0.018,0.034,0.174,0.256) ^T

Final synthetic weight is:

W ³＝w ³W ²＝(0.097,0.142,0.066,0.056,0.099,0.166,0.144,0.131,0.045,0.054) ^T。

Ordering to all noninferior solutions is 6,3,7,8,5,1,2,4,10,9.

Find the maximal value in the bottom synthetic weight vector, the corresponding scheme of this value is exactly optimal case, and satisfactory solution just is so the satisfactory solution of correspondence is A among the embodiment ₆(32,20.000).

Step 8: net result output.

After the scheduling result decision-making, final scheduling result is exported.Output information comprises the corresponding objective function point diagram of optimum solution and corresponding scheduling Gantt chart thereof.Retraining 20 tasks, 8 resources with nothing is example, and net result is exported as shown in Figure 6.

Wherein Fig. 6 left side has shown two target function values of the final optimum solution correspondence of determining, the testing tool resource that the Gantt chart of right side optimum solution correspondence provides the scheduling of each test assignment and taken.

Claims (3)

1. the change neighborhood multiple goal test assignment dispatching method based on decomposition is characterized in that, comprises the steps:

Step 1: obtain the information of selected test assignment collection, comprising:

Test assignment table T={ t ₁, t _2,, t _j, t _N, t _jBe j test assignment, N is the test assignment sum;

Instrument resource table R={r ₁, r ₂,, r _i..., r _M, r _iBe i instrument, M is the instrument sum;

Testing scheme collection table, test assignment t _jThe testing scheme collection be defined as

K wherein _jBe test assignment t _jThe testing scheme sum, j=1,2 ..., N;

And the temporal constraint relation table between test assignment;

Step 2: initialization specifically comprises step 2.1～2.4;

Step 2.1: determine objective function, objective function comprises test duration f ₁(x) and each step average load f ₂(x);

Test duration

Wherein,

Expression test assignment t _jSelection scheme

The time deadline; X ∈ Ω, Ω are field of definition;

If d represents parallel step number, initial value is 1, after all test assignments are all arranged instrument resource, when the instrument resource of two test assignments of every existence has repetition, upgrades d=d+1, calculates total parallel step number, determines respectively to go on foot average load then:

Wherein, Expression test assignment t _jAt instrument r _iElapsed time when carrying out; Expression test assignment t _jWith instrument r _iDemand relation, as test assignment t _jNeed take instrument r _iSituation under,

Otherwise

Step 2.2: variable and parameter initialization;

The set that note is preserved noninferior solution is EP, and initial

Initialization reference point optimal solution set z=(z ₁, z ₂) ^T, z ₁=min{f ₁(x), x ∈ Ω }, z ₂=min{f ₂(x), x ∈ Ω }; Wherein Ω represents solution space;

Initialized parameter comprises: maximum iteration time M'; Population size N'; Neighborhood variation range, maximal value are T _High, minimum value is T _LowCrossover probability CR and variation Probability p; When iterations was ger, the field size in corresponding generation was

T=floor (((T _High-T _Low)/(M') ²) * (ger-M') ²)+T _Low, floor represents to round downwards;

Step 2.3: calculate weight vector neighborhood indexed set; Find and i T the weight vector that weight vector is nearest, defining i weight vector neighborhood indexed set is B (i)={ i ₁,, i _T, i ∈ [1, N'];

Step 2.4: produce initial solution; Produce initial population at random, and determine the value of the objective function of each individual correspondence; Comprise N decision variable in each individuality, wherein the execution time of j test assignment of the more big expression correspondence of the value of j decision variable is more forward;

Step 3, the individual intersection; I the individuality of t in generation The individuality that produces after intersecting

For:

$x_{t+1}^i=\left\{\begin{array}{cc}{x}_t^i+F_1\&times;(x_t^i-x_t^{i1})+F_2\&times;(x_t^i-x_t^{i2})& \mathrm{rand}\left(1\right)<\mathrm{CR}\\ x_t^i& \mathrm{rand}\left(1\right)\&GreaterEqual;\mathrm{CR}\end{array}\right.$

Wherein, With

Be the weight vector of random choose among the weight vector B (i), constant F ₁With F ₂All be set to 1, rand (1) and for variation range be 0 to 1 decimal at random;

Step 4: individual variation;

If i the individuality in t generation is

Decision variable

Variation obtains

For:

$x_{\left(j\right)}^{i*}=\left\{\begin{array}{cc}\mathrm{normal}(x_{\left(j\right)}^i,\mathrm{\&sigma;})& \mathrm{rand}\left(1\right)<p\\ x_{\left(j\right)}^i& \mathrm{rand}\left(1\right)\&GreaterEqual;p\end{array}\right.,j=\mathrm{1,2},...,N$

Wherein, Be that an obedience average is

Variance is the number of the normal distribution of σ, and σ is set to 1/20 of decision variable variation range; Expression x ⁱNew individuality after the variation;

Step 5: upgrade neighborhood and reference point;

Upgrade neighborhood: for j ∈ B (i), the definition reference point is z _iThe fitness of j subproblem be

Wherein Represent i j weight vector value in the weight vector neighborhood indexed set; If individual y' is the solution that obtains behind i the individual variation, to j ∈ B (i), if fitness value F (y')≤F (x ^j), the individual x of new population more then ^j=y', F (x ^j)=F (y'), otherwise, keep the individual x of population ^jConstant;

Upgrade z: to any i=1,2, if z _i＜f _i(y'), assignment z then _i=f _i(y');

Step 6: preserve noninferior solution; After each is finished for evolution, all noninferior solutions that obtain are saved among the set EP, and the inferior solution that will gather among the EP is removed;

If two are separated and target function value is respectively: A ₁(f ₁(A ₁), f ₂(A ₁)), A ₂(f ₁(A ₂), f ₂(A ₂)); If f ₁(A ₁) 〉=f ₁(A ₂) and f ₂(A ₁) 〉=f ₂(A ₂), then separate A ₁Be inferior solution, A is separated in deletion ₁

Step 7: carry out final decision by the noninferior solution among the analytical hierarchy process pair set EP;

Step 8: optimal case is exported.

2. according to claim 1 based on the change neighborhood multiple goal test assignment dispatching method that decomposes, it is characterized in that described step 7 specific implementation method is: set up the tri-layer structure: the top final decision result of being, the middle layer is dimension n ₂=2 objective function, the bottom is the noninferior solution among the set EP, the noninferior solution number is n ₁

At first, each noninferior solution of bottom importance about the objective function of each dimension of middle layer is compared in twos, the structure judgment matrix, matrix is n ₁* n ₁Matrix; For a certain dimension, the capable j column element of i a in the judgment matrix _IjA is separated in expression _iWith respect to separating A _jSignificance level, f _iAnd f _jBe to separate A _iConciliate A _jAt the target function value of corresponding dimension, f _MaxAnd f _MinBe respectively maximum and the minimum value of objective function on the corresponding dimension; Element a _IjValue be:

Wherein, round represents the result of calculation fraction part is rounded up;

Then, judgment matrix is carried out consistency check, specifically: calculate coincident indicator CI=(λ earlier _Max-n)/(n-1), λ _MaxBe the eigenvalue of maximum of judgment matrix, n is the exponent number of judgment matrix; Calculate Consistency Ratio CRT=CI/RI again, RI is the mean random coincident indicator; If CRT≤0.10, then judgment matrix is by consistency check, otherwise needs the correction judgment matrix;

After consistency check is passed through, to the eigenvalue of maximum characteristic of correspondence vector u=(u of judgment matrix ₁, u ₂..., u _n) ^T, carry out normalized, obtain the relative ordering weight vectors w=(w of element under the corresponding single criterion ₁, w ₂..., w _n) ^T, Corresponding two objective functions, weight vectors w obtains sorting ³

Equally, according to the importance of objective function to the final decision result, set up the middle layer for top judgment matrix, matrix is n ₂* n ₂Matrix; Obtain the synthetic weight vector W in middle layer ², finally obtain bottom synthetic weight vector W ³=w ³W ²Find the maximal value in the bottom synthetic weight vector, the corresponding scheme of this value is exactly optimal case.

3. one kind based on claim 1 or 2 described multiple goal test assignment dispatching platforms, it is characterized in that, comprise as lower module: user management module, test assignment scheduling example interpolation and maintenance module, database module, test case import module, dispatching method module, dispatching method expansion module, scheduling result decision-making module and scheduling result and show output module;

User management module is used for information and the password of leading subscriber, is the entrance of test assignment dispatching platform; Test assignment scheduling example interpolation and maintenance module add for five kinds of information tables to test and safeguard, and store data in the database module; Described five kinds of information tables are: the temporal constraint relation table between test assignment collection table, test assignment table, instrument resource table, test assignment and testing scheme collection table; Test case imports the test assignment collection that module is selected according to the user, obtains the needed information table of scheduling from database module, and outputs to the dispatching method module; Integrated more than one dispatching method in the dispatching method module, the user chooses a kind of dispatching method from the dispatching method module, and the parameter of selected dispatching method set, dispatching method module operation test assignment collection, with all noninferior solution input schedulings that obtain decision-making module as a result, be integrated with in the dispatching method module based on the change neighborhood multiple goal test assignment dispatching method that decomposes; The scheduling result decision-making module carries out final decision by analytical hierarchy process to noninferior solution, and the output optimum solution shows output module to scheduling result; Scheduling result shows that output module is shown to the user with the objective function point diagram of optimum solution and corresponding scheduling Gantt chart; The dispatching method expansion module is used for new dispatching method is added the dispatching method module.

Images