CN107992999B - Multi-subject scheduling method oriented to personalized production environment - Google Patents

Abstract

The invention provides a multi-main-body scheduling method facing to an individualized production environment, wherein a resource party and each user are made to be a scheduling main body with an autonomous behavior decision capability, and an initial scheduling scheme is randomly generated by the resource main body and submitted to each user main body to start multi-main-body scheduling; then the resource autonomous body and the user autonomous body carry out scheduling negotiation through an alternative proposal, and when the user autonomous body and the resource autonomous body fail to reach the negotiation, the user autonomous body with conflict starts the alternative proposal to carry out scheduling negotiation; if the negotiation of the resource autonomous body and the user autonomous body is completely achieved, outputting a multi-body scheduling scheme with consistent negotiation, otherwise, adopting a preset arbitration rule to eliminate the rest conflicts and generating a final multi-body scheduling scheme. The method gives full play to the autonomy of resources and each user, can fully meet the target requirements of each main body, and improves the group satisfaction degree of result scheduling.

Description

Multi-subject scheduling method oriented to personalized production environment

Technical Field

The invention relates to the field of production system performance optimization, in particular to a multi-agent scheduling method oriented to an individualized production environment.

Background

The document "A hybrid algorithm based on a new design structure evaluation method for job shop scheduling, computers & Industrial Engineering,2015,88: 417-42" discloses a production system performance optimization method based on a hybrid scheduling algorithm. The method comprises the steps of firstly, carrying out global optimization on the scheduled maximum completion time by using a particle swarm algorithm, and then carrying out local optimization by using a variable neighborhood search algorithm so as to minimize the scheduled maximum completion time. The scheduling scheme obtained by the method has smaller maximum completion time, so that the production efficiency of manufacturing enterprises can be improved, the utilization rate of production resources is improved, and the production cost is reduced.

In the large-scale production mode, a manufacturing enterprise generally combines and processes the demands of different users without distinguishing, and then optimizes and schedules the demands with the related targets (such as maximum completion time) of a resource side so as to reduce the manufacturing cost of the enterprise and improve the production efficiency. However, with increasingly global market competition and rapid development of manufacturing informatization, the traditional enterprise operation mode is rapidly shifted to personalized customized production from a large-scale production mode. Under the personalized production environment, each user has respective customized task and corresponding differentiated target, and at the moment, if a resource-side-dominant scheduling control method is still adopted, the differentiated target requirements of the users are difficult to meet, which affects the group satisfaction of the users and reduces the industry competitiveness of manufacturing enterprises.

In order to solve the scheduling control problem in the personalized production environment, the optimization target to be scheduled is shifted from the original target mainly considering resource correlation to the target considering resource and differentiation targets of each user; the main body of the scheduling decision is changed from a single main body of the resource side into a multi-main body formed by the resource side and multiple users together. Therefore, a novel scheduling mode, namely multi-agent scheduling facing to the personalized production environment is generated.

Disclosure of Invention

The invention provides a multi-agent scheduling method oriented to an individualized production environment, aiming at the problem that the existing scheduling method mainly based on a resource party is difficult to meet the differentiated target requirements of multiple users in the individualized production environment. The method comprises the steps that a resource party and each user are enabled to be a scheduling subject (namely a resource subject and a user subject) with an autonomous behavior decision capability, and an initial scheduling scheme is randomly generated by the resource subject and submitted to each user subject to start multi-subject scheduling; then the resource autonomous body and the user autonomous body carry out scheduling negotiation through alternative proposal (the user autonomous body receives the proposal of the resource autonomous body, then the user autonomous body modifies the proposal and feeds back the modified proposal to the resource autonomous body, and then the resource autonomous body generates a new proposal and submits the new proposal to each user autonomous body), and when the user autonomous body and the resource autonomous body fail to reach negotiation, the user autonomous body with conflict starts the alternative proposal to carry out scheduling negotiation; if the negotiation of the resource autonomous body and the user autonomous body is completely achieved, outputting a multi-body scheduling scheme with consistent negotiation, otherwise, adopting a preset arbitration rule to eliminate the rest conflicts and generating a final multi-body scheduling scheme. The method gives full play to the autonomy of resources and each user, can fully meet the target requirements of each main body, and improves the group satisfaction degree of result scheduling.

The technical scheme of the invention is as follows:

the multi-agent scheduling method facing the personalized production environment is characterized in that: the method comprises the following steps:

step 1, initializing multi-agent scheduling:

(a) each user autonomous entity establishes communication with a resource autonomous entity to determine a set of autonomous entities { A } that participate in multi-entity scheduling₀,A₁,A₂,...,A_i,...,A_nIn which A is₀Represents a resource autonomous body, A_iN represents the respective user's own subject;

(b) resource autonomous body A₀With each user's own body A_i1, 2.. n to determine a multi-master scheduled process task set T ═ T ·₁,T₂,...,T_i,...,T_nWhere T is_iFor users to self-body A_iThe processing task of (2);

(c) respective main body A_iN sets a respective maximum scheduling objective function O, 0,1,2_iI is 0,1, 2.., n, and initializes the respective negotiation policy S according to equation (1)_i；

In the formula (I), the compound is shown in the specification,

is a self body A_iAt negotiated time t_iThe objective function value of time, β i, is the subject A_iThe negotiation policy adjustment factor of (2) is,

is a self body A_iThe value of the ideal objective function of (c),

is a self body A_iThe minimum value of the objective function that is acceptable,

is a self body A_iA negotiation deadline of (2);

(d) each user's own agent A_i1, 2.. n initializes its own acceptance condition C by equation (2)_i,i＝1,2,...,n；

In the formula, pi_i,jIs a self body A_iReceived from the self-body A_jScheduling proposal of (1), O_i(π_i,j) Is a self body A_iIn scheduling proposal pi_i,jThe objective function value of;

(e) respective main body A_iN, initializing respective negotiation conditions N according to equation (3)_i,i＝0,1,2,...,n；

Step 2, starting multi-agent scheduling:

resource autonomous body A₀Initializing a negotiation constraint set as empty R { }, then taking a multi-main-body scheduling processing task set T as a scheduling object, and generating a scheduling scheme pi in a mode shown by a formula (4)₀And submits the scheduling scheme as an initial proposal to an individual user main body A_i1,2, n to start multi-subject scheduling;

step 3, starting the negotiation information on-line learning of the negotiation object:

when the negotiation object is autonomous A_jAt the time of starting from the body A_jAfter submitting a new proposal, updating a historical proposal information set, and then optimizing a negotiation strategy model shown in formula (5) by using a random gradient descent method by taking the historical proposal and corresponding proposal time as input so as to learn the autonomous agent A_jNegotiation strategy S_j(ii) a From subject A obtained by learning_jApproximate negotiation strategy of

Obtain a subject A_jEstimate of the value of the ideal objective function

Estimate of minimum acceptable objective function value

And an estimate of the negotiation deadline

In the formula (I), the compound is shown in the specification,

is a self body A_jAt negotiated time t_jAn estimate of the objective function value of the proposal at time,

is a self body A_jAn estimate of the negotiated policy adjustment factor of a_j,b_j,c_jIs a constant parameter;

step 4, the user self-body and the resource self-body negotiate and schedule: the user self-body and the resource self-body finally obtain a multi-body scheduling scheme with multi-body negotiation consistency through continuous alternate proposal and modification proposal:

step 4.1, making a user self-body autonomous behavior decision:

(a) when the user is self-owned A_iReceiving a message from a resource autonomous agent A₀Scheduling proposal pi_i,0Thereafter, the resource self-agent A is learned through step 2₀Negotiation information of

Synthetic resource autonomous body A₀Negotiation information X of₀And user-autonomous body A_iNegotiation information of

User-owned entity A_iThe formula (6) is adopted to modify the negotiation strategy adjustment factor, and then the formula (1) is adopted to adjust the self negotiation strategy S_i

β_i＝β_i+g_i(X₀,X_i) (6)

In the formula, g_i(.) is the user's own principal A_iAdjusting a function by adopting a negotiation strategy;

(b) user-owned entity A_iUsing the adjusted negotiation strategy S_iUpdating itself at the currently negotiated time t_iObjective function value of time proposal

Adjusting the self-acceptance condition C_iAnd negotiation condition N_iAnd controlling the negotiation process according to the following steps: if C is present_iNotify resource autonomous A if 1₀Accepting the current proposal and completing the negotiation, otherwise calculating N_i(ii) a If N is present_iIf 0, the resource self-body A is informed₀Refusing current proposal and abandoning negotiation, otherwise user self-body A_iTurning to step (c) and continuing to the resource self-body A₀Proposing a proposal;

(c) strip satisfying negotiation constraint set RUnder the article, the user is self-body A_iAccording to formula (7), with resources from subject A₀Is proposed pi_i,0Based on the obtained target value, local search algorithm is adopted to obtain the target value satisfying itself

And then propose the scheme as a feedback scheme pi_0,iSubmission to resource autonomous agent A₀；

Step 4.2, making a resource autonomous behavior decision:

(a) the resource autonomous body waits for and receives feedback information of all the user autonomous bodies which do not finish the negotiation: if the user is from principal A_iReceiving the proposal, the resource is from the subject A₀Recording user self-body A_iCompleting negotiation and updating a negotiation constraint set R; if the user is from principal A_iRefusing proposal, then the resource is from subject A₀Recognition and user-autonomous body A_iConflicting user-autonomous agents A_jThen, turning to step 5; if the user is from principal A_iFeedback proposal pi_0,iThen go to step 2 to learn the user's own body A_iNegotiation information of

(b) Resource autonomous body A₀Judging whether the multi-subject scheduling is completed: calculating the judgment condition C according to the formula (8)₀If C is present₀If the result is 1, the multi-agent scheduling is finished and the step 6 is carried out, otherwise, the step (c) is carried out to continue the negotiation;

in the formula, B_iFor users to self-body A_iNegotiation status flag of (1): b is_i0 stands for abort negotiation, B_iWith 1 standing for continued negotiation, B_i2 stands forCompleting the negotiation;

(c) synthetic resource autonomous body A₀Self negotiation information X₀And negotiation information { X ] of each user-owned entity that does not complete negotiation₁,X₂,.., the resource is from principal A₀The formula (9) is adopted to modify the adjustment factor of the negotiation strategy, and the formula (1) is substituted to update the self negotiation strategy S₀；

β₀＝β₀+g₀(X₀,X₁,X₂,...) (9)

In the formula, g₀(.) is a self-owned agent A₀Adjusting a function by adopting a negotiation strategy;

(d) resource autonomous body A₀Using the adjusted negotiation strategy S₀Updating itself at the currently negotiated time t₀Objective function value of time proposal

Adjusting self negotiation condition N according to formula (3)₀To control the negotiation process: if N is present_iIf 0, then the resource is from principal A₀Abandoning to continue the multi-subject scheduling and turning to step 6, otherwise, the resource self-subject A meets the condition of the negotiation constraint set R₀Amending the proposal and making a new scheduling scheme pi₀Satisfy the current objective function value of the self

Then scheduling scheme pi₀Submitting the data to all the user autonomous bodies which do not finish the negotiation;

step 5, the user self-body A_iUser-autonomous body A with scheduling conflict_jAnd (3) negotiation scheduling:

step 5.1, user-independent agent A_jJudging whether to accept the user self-body A_iThe negotiation request of (2); if the negotiation is accepted, go to step 5.2, otherwise the user is self-owned A_jRefusing and user self-body A_iThe scheduling negotiation is transferred to step 5.3;

step 5.2, user's own agent A_iWith the user's own body A_jScheduling negotiation:

(a) user-owned entity A_jAnd (3) autonomous behavior decision making: user-owned entity A_jReceiving from a user self-body A_iScheduling proposal pi_j,iThereafter, the user-independent entity A is learned through step 2_iNegotiation information of

And updating the negotiation strategy S of the self_jAdjusting the self-acceptance condition C_jAnd negotiation condition N_j(ii) a If the proposal is continued, the user self-body A is used under the condition that the negotiation constraint set R is satisfied_iIs proposed pi_j,iOn the basis of which the user is self-owned A_jObtaining target value satisfying itself by local search algorithm

And then propose the scheme as a feedback scheme pi_i,jSubmitted to the user's autonomous body A_iOtherwise, the user is self-owned A_jFinish negotiation and go to step 5.3;

(b) user-owned entity A_iAnd (3) autonomous behavior decision making: user-owned entity A_iReceiving from a user self-body A_jFeedback scheduling proposal pi_i,jThereafter, the user-independent entity A is learned through step 2_jNegotiation information of

And updating the negotiation strategy S of the self_iAdjusting the self-acceptance condition C_iAnd negotiation condition N_i(ii) a If the proposal is continued, the user self-body A is used under the condition that the negotiation constraint set R is satisfied_jFeedback proposal of_i,jOn the basis of which the user is self-owned A_iObtaining target value satisfying itself by local search algorithm

And then the scheme is taken as a new proposal pi_j,iSubmitted to the user's autonomous body A_jOtherwise, the user is self-owned A_iFinish negotiation and go to step 5.3;

step 5.3, turning to step (a) of step 4.2, feeding back a negotiation result to the resource autonomous body: if the user is from principal A_iAnd a user-autonomous body A_jIf the negotiation is reached, then the agreed scheduling scheme pi will be reached_i,jAs a user self-body A_iAnd a user-autonomous body A_jIs submitted to the resource autonomous body A₀(ii) a Otherwise the user is self-owned A_iNotification of resource autonomous body A₀Abandoning negotiation and accepting the final multi-subject scheduling scheme;

step 6, if the multi-agent scheduling is finished, outputting a scheduling scheme pi with the negotiation consistency between the resource autonomous agent and all the user autonomous agents_0,1,...,m(ii) a Otherwise, scheduling the processing tasks with the negotiation not being consistent except the negotiation constraint set R by adopting a preset arbitration rule to generate a final multi-main-body scheduling scheme pi_0,1,...,mAnd output.

Advantageous effects

The invention has the beneficial effects that: compared with the traditional scheduling which takes resources as the only scheduling main body, the method ensures that the resources and all users have the capability of autonomous scheduling decision at the same time, so that the scheduling decision is converted from the traditional scheduling which takes the resources as a single main body into the multi-main scheduling which takes the resources and all users to participate together, thereby adapting to the differentiated target requirements of the resources and all users in the personalized production environment; users and resources can fully meet self targets through negotiation scheduling based on alternative proposals, and meanwhile, all main bodies can fully know negotiation objects through online learning, so that self negotiation strategies are more reasonably adjusted, and scheduling decisions which are beneficial to meeting self targets and achieving negotiation consistency are made; when the users and the resources can not be directly agreed, the space for conflict resolution can be fully excavated through negotiation scheduling between the users so as to improve a multi-agent scheduling scheme; if each subject finally fails to reach the agreed multi-subject scheme, the arbitration decision will generate the final scheme according to the established rule, and since it is uncertain whether the scheme can meet the target of each subject, the subjects will be motivated to actively participate in the negotiation and rationally make the scheduling decision, thereby reaching the agreed multi-subject scheduling scheme as much as possible. In a word, the multi-subject scheduling provided by the invention can exert the autonomy of resources and each user, fully meet the target requirements of each subject and improve the group satisfaction degree of result scheduling.

Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.

Drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

FIG. 1 is a block diagram of the multi-agent scheduling method for personalized production environment of the present invention.

Fig. 2 is a flow chart of inter-autonomous body negotiation scheduling of the method of the present invention.

FIG. 3 is a flow chart of the autonomous behavioral decision of the method of the present invention.

Detailed Description

The following detailed description of embodiments of the invention is intended to be illustrative, and not to be construed as limiting the invention.

Referring to fig. 1-3, the specific steps of the multi-agent scheduling method for the personalized production environment of the present invention are as follows:

step 1, initializing multi-agent scheduling:

(a) each user autonomous entity establishes communication with a resource autonomous entity to determine a set of autonomous entities { A } that participate in multi-entity scheduling₀,A₁,A₂,...,A_i,...,A_nIn which A is₀Represents a resource autonomous body, A_iN represents the respective user's own subject;

(b) resource autonomous body A₀With each user's own body A_i1,2, n to determine a multi-master scheduled process task set T ═ T · of multi-master scheduling₁,T₂,...,T₁,...,T_nWhere T is_iFor users to self-body A_iThe processing task of (2);

(c) resource autonomous body A₀And each user's own body A_iN sets a respective maximum scheduling objective function O_iN and initializing the respective negotiation strategy S according to equation (1)_i；

In the formula (I), the compound is shown in the specification,

is a self body A_iAt negotiated time t_iObjective function value of time, beta_iIs a self body A_iThe negotiation policy adjustment factor of (2) is,

is a self body A_iThe value of the ideal objective function of (c),

is a self body A_iThe minimum value of the objective function that is acceptable,

is a self body A_iA negotiation deadline of (2);

(d) each user's own agent A_i1, 2.. n initializes its own acceptance condition C by equation (2)_i,i＝1,2,...,n；

In the formula, pi_i,jIs a self body A_iReceived from the self-body A_jScheduling proposal of (1), O_i(π_i,j) Is a self body A_iIn scheduling proposal pi_i,jThe objective function value of;

(e) resource autonomous body A₀And each user's own body A_i1, 2.. N, N initialize respective negotiation conditions N with equation (3)_i,i＝0,1,2,...,n。

Step 2, starting multi-agent scheduling: resource autonomous body A₀Initializing a negotiation constraint set as empty R { }, then taking a multi-main-body scheduling processing task set T as a scheduling object, and generating a scheduling scheme pi in a mode shown by a formula (4)₀And submits the scheduling scheme as an initial proposal to an individual user main body A_i1,2, n to start multi-subject scheduling. The method comprises the following specific steps:

step 3, starting the negotiation information on-line learning of the negotiation object: when the negotiation object is autonomous A_jAt the time of starting from the body A_jAfter submitting a new proposal, updating a historical proposal information set, and then optimizing a negotiation strategy model shown in formula (5) by using a random gradient descent method by taking the historical proposal and corresponding proposal time as input so as to learn the autonomous agent A_jNegotiation strategy S_j(ii) a From subject A obtained by learning_jApproximate negotiation strategy of

Obtain a subject A_jEstimate of the value of the ideal objective function

Estimate of minimum acceptable objective function value

And an estimate of the negotiation deadline

The method comprises the following specific steps:

in the formula (I), the compound is shown in the specification,

is a self body A_jAt negotiated time t_jAn estimate of the objective function value of the proposal at time,

is a self body A_jAn estimate of the negotiated policy adjustment factor of a_j,b_j,c_jAre constant parameters.

(a) Receiving information from a negotiated object A_jP (th) proposal information of

Wherein the content of the first and second substances,

for negotiating object A_jSubmit the p' th proposal

Negotiated time of day;

(b) updating the history proposal information set of the negotiation object:

(c) with set of objective functions { O₁,O₂,....,O_q,...,O_mEvaluate the negotiated object A for each objective function in_jIs proposed

And records the value of each objective function { O }_1,p,O_2,p,...,O_q,p,...,O_m,pIn which O is_q,pQ is 1, 2.. times.m is the objective function O_qIn the negotiation of object A_jIs proposed

A value of;

(d) updating the target value data set of the history proposal of the negotiation object:

(e) order to

And

optimizing the negotiation strategy model shown in the formula (5) by adopting a random gradient descent method so as to learn the negotiation strategy of a negotiation opponent; then, the data are recorded in each objective function O_q1,2,.. m, after learning

Least square error of

And parameters of the corresponding negotiation strategy model

Is estimated value of

(f) Order negotiation object A_jObject function O of_jIs composed of

Corresponding objective function O_θθ ∈ {1, 2., m }, and let negotiation object A_jIs approximately negotiated the strategy of

Then connected vertically

And

finding a negotiated object A_jEstimate of the value of the ideal objective function

Estimation of minimum acceptable target value

And an estimate of the negotiation deadline

Step 4, the user self-body and the resource self-body negotiate and schedule: and finally obtaining a multi-subject scheduling scheme with multi-subject negotiation consistency by the user subject and the resource subject through continuous alternative proposal and modification proposal.

Step 4.1, making a user self-body autonomous behavior decision:

(a) when the user is self-owned A_iReceiving a message from a resource autonomous agent A₀Scheduling proposal pi_i,0Thereafter, the resource self-agent A is learned through step 2₀Negotiation information of

Synthetic resource autonomous body A₀Negotiation information X of₀And user-autonomous body A_iNegotiation information of

User-owned entity A_iThe formula (6) is adopted to modify the negotiation strategy adjustment factor, and then the formula (1) is adopted to adjust the self negotiation strategy S_i. The method comprises the following specific steps:

β_i＝β_i+g_i(X₀,X_i) (6)

in the formula, g_i(.) is the user's own principal A_iThe negotiation strategy employed adjusts the function.

When the user is self-owned A_iReceiving from the resource autonomous body A₀Scheduling proposal pi_i,0Thereafter, the resource self-body A is recorded₀Currently negotiated time t₀；

Resources are transferred from subject A₀Has negotiated time t of₀And scheduling proposal pi_i,0Submit step 2 learning and obtain resource autonomous body A₀Negotiation information X of₀: ideal objective function value

Minimum acceptable target value

And negotiating a deadline

User-owned entity A_iAccording to the negotiation information X of the resource autonomous body obtained by learning₀And user-autonomous body A_iSelf negotiation information X_iThe user self-body A is reasoned and modified on line through the formula (6)_iNegotiation policy adjustment factor of (2):

user-owned entity A_iAdjusting the modified negotiation strategy by a factor beta_iSubstituting formula (1), adjusting its own negotiation strategy S_i。

(b) User-owned entity A_iUsing the adjusted negotiation strategy S_iUpdating itself at the currently negotiated time t_iObjective function value of time proposal

Adjusting the self-acceptance condition C_iAnd negotiation condition N_iAnd controlling the negotiation process according to the following steps: if C is present_iNotify resource autonomous A if 1₀Accepting the current proposal and completing the negotiation, otherwise calculating N_i(ii) a If N is present_iIf 0, the resource self-body A is informed₀Refusing current proposal and abandoning negotiation, otherwise user self-body A_iTurning to step (c) and continuing to the resource self-body A₀A proposal is made.

(c) Under the condition of meeting the negotiation constraint set R, the user self-body A_iAccording to formula (7), with resources from subject A₀Is proposed pi_i,0Based on the obtained target value, local search algorithm is adopted to obtain the target value satisfying itself

And then propose the scheme as a feedback scheme pi_0,iSubmission to resource autonomous agent A₀。

Step 4.2, making a resource autonomous behavior decision:

(a) the resource autonomous body waits for and receives feedback information of all the user autonomous bodies which do not finish the negotiation: if the user is from principal A_iReceiving the proposal, the resource is from the subject A₀Recording user self-body A_iCompleting negotiation and updating a negotiation constraint set R; if the user is from principal A_iRefusing proposal, then the resource is from subject A₀Recognition and user-autonomous body A_iConflicting user-autonomous agents A_iThen, turning to step 5; if the user is from principal A_iFeedback proposal pi_0,iThen go to step 2 to learn the user's own body A_iNegotiation information of

The method comprises the following specific steps:

resource autonomous body A₀Waiting for all outstanding negotiated user autonomous A_iThe feedback information of (2): if the user is from principal A_iThe feedback is the negotiation decision information, thenTurning to the step 2; if the feedback is the scheduling proposal, turning to step 3; if the user is from principal A_iReceiving the current proposal, the resource is from the main body A₀Recording user self-body A_iThe negotiation is completed, and the user is self-owned A_iAdding the processing tasks of the other users into a negotiation constraint set R in the time interval of the processing tasks, wherein the processing tasks of the other users in the time interval in the negotiation constraint set R cannot be occupied; if the user is from principal A_iDisagreement with the current proposal and quit with the resource self-body A₀Negotiation of (2), then the resource is autonomous from principal A₀Will identify and user from principal A_iUser-autonomous body A with conflict on scheduling scheme_jAnd go to step 5.

Resource autonomous body A₀Receiving from a user self-body A_iIs given by a feedback scheme pi_0,iAnd records the user self-body A_iHas negotiated time t of_iAnd turning to step 2 to learn the user's own body A_iNegotiation information X of_i: ideal value of objective function

Minimum acceptable target value

And negotiating a deadline

Resource autonomous body A₀Judging whether all the user self-bodies which do not finish the negotiation have submitted a feedback scheme: if both are submitted, go to step (b); otherwise, continuously waiting and receiving the feedback information of the user self-body.

(b) Resource autonomous body A₀Judging whether the multi-subject scheduling is completed: calculating the judgment condition C according to the formula (8)₀If C is present₀If 1, the multi-master scheduling is completed and goes to step 6, otherwise go to step (c) to continue negotiation.

In the formula, B_iFor users to self-body A_iNegotiation status flag of (1): b is_i0 stands for abort negotiation, B_iWith 1 standing for continued negotiation, B_iAnd 2 represents the completion of the negotiation.

(c) Synthetic resource autonomous body A₀Self negotiation information X₀And negotiation information { X ] of each user-owned entity that does not complete negotiation₁,X₂,.., the resource is from principal A₀The formula (9) is adopted to modify the adjustment factor of the negotiation strategy, and the formula (1) is substituted to update the self negotiation strategy S₀. The method comprises the following specific steps:

β₀＝β₀+g₀(X₀,X₁,X₂,...) (9)

in the formula, g₀(.) is a self-owned agent A₀The negotiation strategy employed adjusts the function.

Resource autonomous body A₀Subject A of each user obtained by learning_iNegotiation information X of_i: ideal value of objective function

Minimum acceptable target value

And negotiating a deadline

And resource-autonomous body A₀Self negotiation information X₀The resource self-body A is reasoned and modified on line through the formula (9)₀Negotiation policy adjustment factor of (2):

resource autonomous body A₀Adjusting the modified negotiation strategy by a factor beta₀Substituting formula (1), adjusting its own negotiation strategy S₀。

(d) Resources are derived fromMain body A₀Using the adjusted negotiation strategy S₀Updating itself at the currently negotiated time t₀Objective function value of time proposal

Adjusting self negotiation condition N according to formula (3)₀To control the negotiation process: if N is present_iIf 0, then the resource is from principal A₀Abandoning to continue the multi-subject scheduling and turning to step 6, otherwise, the resource self-subject A meets the condition of the negotiation constraint set R₀Amending the proposal and making a new scheduling scheme pi₀Satisfy the current objective function value of the self

Then scheduling scheme pi₀Submitted to all the user-autonomous bodies that do not complete the negotiation.

Step 5, the user self-body A_iUser-autonomous body A with scheduling conflict_jAnd (3) negotiation scheduling:

step 5.1, user-independent agent A_jJudging whether to accept the user self-body A_iThe negotiation request of (2); if the negotiation is accepted, go to step 5.2, otherwise the user is self-owned A_jRefusing and user self-body A_iThe scheduling negotiation is transferred to step 5.3;

step 5.2, user's own agent A_iWith the user's own body A_jScheduling negotiation:

(a) user-owned entity A_jAnd (3) autonomous behavior decision making: user-owned entity A_jReceiving from a user self-body A_iScheduling proposal pi_j,iThereafter, the user-independent entity A is learned through step 2_iNegotiation information of

And updating the negotiation strategy S of the self_jAdjusting the self-acceptance condition C_jAnd negotiation condition N_j(ii) a If the proposal is continued, the user self-body A is used under the condition that the negotiation constraint set R is satisfied_iIs proposed pi_j,iOn the basis of which the user is self-owned A_jBy local searchingThe cable algorithm obtains a target value satisfying itself

And then propose the scheme as a feedback scheme pi_i,jSubmitted to the user's autonomous body A_iOtherwise, the user is self-owned A_jFinish negotiation and go to step 5.3; the method comprises the following specific steps:

when the user is self-owned A_jReceiving from user self-body A_iScheduling proposal pi_j,iThen, recording the user self-body A_iCurrently negotiated time t_i；

The user is self-owned A_iHas negotiated time t of_iAnd scheduling proposal pi_iSubmitting step 2, learning and obtaining user self-body A_iNegotiation information X of_i: ideal objective function value

Minimum acceptable target value

And negotiating a deadline

User-owned entity A_jUser-independent body A obtained according to learning_iNegotiation information X of_iAnd user-autonomous body A_jSelf negotiation information X_jOn-line reasoning and modifying user-autonomous body A_iNegotiation policy adjustment factor of (2):

user-owned entity A_jAdjusting the modified negotiation strategy by a factor beta_jSubstituting formula (1), adjusting its own negotiation strategy S_j. User-owned entity A_jUsing the adjusted negotiation strategy S_jUpdating itself at the currently negotiated time t_jProposed at the time of dayValue of objective function

Adjusting the self-acceptance condition C_jAnd negotiation condition N_jAnd controlling the negotiation process according to the following steps: if C is present_jNotify the user of the user's own A if 1_iAccepting the current proposal and completing the negotiation, otherwise calculating N_j(ii) a If N is present_jIf 0, the user is notified of the agent A_iRefusing current proposal and abandoning negotiation, otherwise user self-body A_jContinue to the user self-body A_iA proposal is made.

In the case of meeting the negotiation constraint set R, the user self-body A_iIs proposed pi_j,iOn the basis of which the user is self-owned A_iObtaining a current target value satisfying the local search algorithm

And then propose the scheme as a feedback scheme pi_i,jSubmitted to the user's autonomous body A_i。

(b) User-owned entity A_iAnd (3) autonomous behavior decision making: user-owned entity A_iReceiving from a user self-body A_jFeedback scheduling proposal pi_i,jThereafter, the user-independent entity A is learned through step 2_jNegotiation information of

And updating the negotiation strategy S of the self_iAdjusting the self-acceptance condition C_iAnd negotiation condition N_i(ii) a If the proposal is continued, the user self-body A is used under the condition that the negotiation constraint set R is satisfied_jFeedback proposal of_i,jOn the basis of which the user is self-owned A_iObtaining target value satisfying itself by local search algorithm

And then the scheme is taken as a new proposal pi_j,iSubmitted to the user's autonomous body A_jOtherwise, the user is self-owned A_iThe negotiation is ended and step 5 is passed.3; the method comprises the following specific steps:

when the user is self-owned A_iReceiving from user self-body A_jFeedback scheduling proposal pi_i.jThen, recording the user self-body A_jCurrently negotiated time t_j；

The user is self-owned A_jHas negotiated time t of_jAnd scheduling proposal pi_jSubmitting step 2, learning and obtaining user self-body A_jNegotiation information X of_j: ideal objective function value

Minimum acceptable target value

And negotiating a deadline

User-owned entity A_iUser-independent body A obtained according to learning_jNegotiation information X of_jAnd user-autonomous body A_iSelf negotiation information X_iOn-line reasoning and modifying user-autonomous body A_iNegotiation policy adjustment factor of (2):

user-owned entity A_iAdjusting the modified negotiation strategy by a factor beta_iSubstituting formula (1), adjusting its own negotiation strategy S_i. User-owned entity A_iUsing the adjusted negotiation strategy S_iUpdating itself at the currently negotiated time t_iObjective function value of time proposal

Adjusting the self-acceptance condition C_iAnd negotiation condition N_iAnd controlling the negotiation process according to the following steps: if C is present_iNotify the user of the user's own A if 1_jAccept the current proposal and completeNegotiate, otherwise calculate N_i(ii) a If N is present_iIf 0, the user is notified of the agent A_jRefusing current proposal and abandoning negotiation, otherwise user self-body A_iContinue to the user self-body A_jA proposal is made.

In the case of meeting the negotiation constraint set R, the user self-body A_jIs proposed pi_i,jOn the basis of which the user is self-owned A_iObtaining target value satisfying itself by local search algorithm

And then propose the scheme as a feedback scheme pi_j,iSubmitted to the user's autonomous body A_j。

Step 5.3, turning to step (a) of step 4.2, feeding back a negotiation result to the resource autonomous body: if the user is from principal A_iAnd a user-autonomous body A_jIf the negotiation is reached, then the agreed scheduling scheme pi will be reached_i,jAs a user self-body A_iAnd a user-autonomous body A_jIs submitted to the resource autonomous body A₀(ii) a Otherwise the user is self-owned A_iNotification of resource autonomous body A₀The negotiation is abandoned and the final multi-master scheduling scheme is accepted.

Step 6, if the multi-agent scheduling is finished, outputting a scheduling scheme pi with the negotiation consistency between the resource autonomous agent and all the user autonomous agents_0,1,...,m(ii) a Otherwise, scheduling the processing tasks which are not in accordance with the negotiation except the negotiation constraint set R by adopting a preset arbitration rule to generate a final multi-main-body scheduling scheme pi_0,1,...,mAnd output.

Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made in the above embodiments by those of ordinary skill in the art without departing from the principle and spirit of the present invention.

Claims (1)

1. A multi-subject scheduling method facing to personalized production environment is characterized in that: the method comprises the following steps:

step 1, initializing multi-agent scheduling:

(a) each user autonomous entity establishes communication with a resource autonomous entity to determine a set of autonomous entities { A } that participate in multi-entity scheduling₀,A₁,A₂,...,A_i,...,A_nIn which A is₀Represents a resource autonomous body, A_iN represents the respective user's own subject;

(b) resource autonomous body A₀With each user's own body A_i1, 2.. n to determine a multi-master scheduled process task set T ═ T ·₁,T₂,...,T_i,...,T_nWhere T is_iFor users to self-body A_iThe processing task of (2);

(c) respective main body A_iN sets a respective maximum scheduling objective function O, 0,1,2_iI is 0,1, 2.., n, and initializes the respective negotiation policy S according to equation (1)_i；

In the formula (I), the compound is shown in the specification,

is a self body A_iAt negotiated time t_iObjective function value of time, beta_iIs a self body A_iThe negotiation policy adjustment factor of (2) is,

is a self body A_iThe value of the ideal objective function of (c),

is a self body A_iThe minimum value of the objective function that is acceptable,

is a self body A_iA negotiation deadline of (2);

(d) each user's own agent A_i1, 2.. n initializes its own acceptance condition C by equation (2)_i,i＝1,2,...,n；

In the formula, pi_i,jIs a self body A_iReceived from the self-body A_jScheduling proposal of (1), O_i(π_i,j) Is a self body A_iIn scheduling proposal pi_i,jThe objective function value of;

(e) respective main body A_iN, initializing respective negotiation conditions N according to equation (3)_i,i＝0,1,2,...,n；

Step 2, starting multi-agent scheduling:

resource autonomous body A₀Initializing a negotiation constraint set as empty R { }, then taking a multi-main-body scheduling processing task set T as a scheduling object, and generating a scheduling scheme pi in a mode shown by a formula (4)₀And submits the scheduling scheme as an initial proposal to an individual user main body A_i1,2, n to start multi-subject scheduling;

step 3, starting the negotiation information on-line learning of the negotiation object:

when the negotiation object is autonomous A_jAt the time of starting from the body A_jAfter submitting a new proposal, updating a historical proposal information set, and then optimizing a negotiation strategy model shown in formula (5) by using a random gradient descent method by taking the historical proposal and corresponding proposal time as input so as to learn the autonomous agent A_jNegotiation strategy S_j(ii) a From subject A obtained by learning_jApproximate negotiation strategy of

Obtain a subject A_jEstimate of the value of the ideal objective function

Estimate of minimum acceptable objective function value

And an estimate of the negotiation deadline

In the formula (I), the compound is shown in the specification,

is a self body A_jAt negotiated time t_jAn estimate of the objective function value of the proposal at time,

is a self body A_jAn estimate of the negotiated policy adjustment factor of a_j,b_j,c_jIs a constant parameter;

step 4, the user self-body and the resource self-body negotiate and schedule: the user self-body and the resource self-body finally obtain a multi-body scheduling scheme with multi-body negotiation consistency through continuous alternate proposal and modification proposal:

step 4.1, making a user self-body autonomous behavior decision:

(a) when the user is self-owned A_iReceiving a message from a resource autonomous agent A₀Scheduling proposal pi_i,0Thereafter, the resource self-agent A is learned through step 2₀Negotiation information of

Synthetic resource autonomous body A₀Negotiation information X of₀And user-autonomous body A_iNegotiation information of

User-owned entity A_iThe formula (6) is adopted to modify the negotiation strategy adjustment factor, and then the formula (1) is adopted to adjust the self negotiation strategy S_i

β_i＝β_i+g_i(X₀,X_i) (6)

In the formula, g_i(.) is the user's own principal A_iAdjusting a function by adopting a negotiation strategy;

(b) user-owned entity A_iUsing the adjusted negotiation strategy S_iUpdating itself at the currently negotiated time t_iObjective function value of time proposal

Adjusting the self-acceptance condition C_iAnd negotiation condition N_iAnd controlling the negotiation process according to the following steps: if C is present_iNotify resource autonomous A if 1₀Accepting the current proposal and completing the negotiation, otherwise calculating N_i(ii) a If N is present_iIf 0, the resource self-body A is informed₀Refusing current proposal and abandoning negotiation, otherwise user self-body A_iTurning to step (c) and continuing to the resource self-body A₀Proposing a proposal;

(c) under the condition of meeting the negotiation constraint set R, the user self-body A_iAccording to formula (7), with resources from subject A₀Is proposed pi_i,0Based on the obtained target value, local search algorithm is adopted to obtain the target value satisfying itself

And then propose the scheme as a feedback scheme pi_0,iSubmission to resource autonomous agent A₀；

Step 4.2, making a resource autonomous behavior decision:

(a) the resource autonomous body waits for and receives feedback information of all the user autonomous bodies which do not finish the negotiation: if the user is from principal A_iReceiving the proposal, the resource is from the subject A₀Recording user self-body A_iCompleting negotiation and updating a negotiation constraint set R; if the user is from principal A_iRefusing proposal, then the resource is from subject A₀Recognition and user-autonomous body A_iConflicting user-autonomous agents A_jThen, turning to step 5; if the user is from principal A_iFeedback proposal pi_0,iThen go to step 2 to learn the user's own body A_iNegotiation information of

(b) Resource autonomous body A₀Judging whether the multi-subject scheduling is completed: calculating the judgment condition C according to the formula (8)₀If C is present₀If the result is 1, the multi-agent scheduling is finished and the step 6 is carried out, otherwise, the step (c) is carried out to continue the negotiation;

in the formula, B_iFor users to self-body A_iNegotiation status flag of (1): b is_i0 stands for abort negotiation, B_iWith 1 standing for continued negotiation, B_i2 represents the completion of the negotiation;

(c) synthetic resource autonomous body A₀Self negotiation information X₀And negotiation information { X ] of each user-owned entity that does not complete negotiation₁,X₂,.., the resource is from principal A₀The formula (9) is adopted to modify the adjustment factor of the negotiation strategy, and the formula (1) is substituted to update the self negotiation strategy S₀；

β₀＝β₀+g₀(X₀,X₁,X₂,...) (9)

In the formula, g₀(.) is a self-owned agent A₀Adjusting a function by adopting a negotiation strategy;

(d) resource autonomous body A₀Using the adjusted negotiation strategy S₀Updating itself at the currently negotiated time t₀Objective function value of time proposal

Adjusting self negotiation condition N according to formula (3)₀To control the negotiation process: if N is present_iIf 0, then the resource is from principal A₀Abandoning to continue the multi-subject scheduling and turning to step 6, otherwise, the resource self-subject A meets the condition of the negotiation constraint set R₀Amending the proposal and making a new scheduling scheme pi₀Satisfy the current objective function value of the self

Then scheduling scheme pi₀Submitting the data to all the user autonomous bodies which do not finish the negotiation;

step 5, the user self-body A_iUser-autonomous body A with scheduling conflict_jAnd (3) negotiation scheduling:

step 5.1, user-independent agent A_jJudging whether to accept the user self-body A_iThe negotiation request of (2); if the negotiation is accepted, go to step 5.2, otherwise the user is self-owned A_jRefusing and user self-body A_iThe scheduling negotiation is transferred to step 5.3;

step 5.2, user's own agent A_iWith the user's own body A_jScheduling negotiation:

(a) user-owned entity A_jAnd (3) autonomous behavior decision making: user-owned entity A_jReceiving from a user self-body A_iScheduling proposal pi_j,iThereafter, the user-independent entity A is learned through step 2_iNegotiation information of

And updating the negotiation strategy S of the self_jAdjusting the self-acceptance condition C_jAnd negotiation condition N_j(ii) a If the proposal is continued, the user self-body A is used under the condition that the negotiation constraint set R is satisfied_iIs proposed pi_j,iOn the basis of which the user is self-owned A_jObtaining target value satisfying itself by local search algorithm

And then propose the scheme as a feedback scheme pi_i,jSubmitted to the user's autonomous body A_iOtherwise, the user is self-owned A_jFinish negotiation and go to step 5.3;

(b) user-owned entity A_iAnd (3) autonomous behavior decision making: user-owned entity A_iReceiving from a user self-body A_jFeedback scheduling proposal pi_i,jThereafter, the user-independent entity A is learned through step 2_jNegotiation information of

And updating the negotiation strategy S of the self_iAdjusting the self-acceptance condition C_iAnd negotiation condition N_i(ii) a If the proposal is continued, the user self-body A is used under the condition that the negotiation constraint set R is satisfied_jFeedback proposal of_i,jOn the basis of which the user is self-owned A_iObtaining target value satisfying itself by local search algorithm

And then the scheme is taken as a new proposal pi_j,iSubmitted to the user's autonomous body A_jOtherwise, the user is self-owned A_iFinish negotiation and go to step 5.3;

step 5.3, turning to step (a) of step 4.2, feeding back a negotiation result to the resource autonomous body: if the user is from principal A_iAnd a user-autonomous body A_jIf the negotiation is reached, then the agreed scheduling scheme pi will be reached_i,jAs a user self-body A_iAnd a user-autonomous body A_jIs submitted to the resource autonomous body A₀(ii) a Otherwise the user is self-owned A_iNotification of resource autonomous body A₀Abandoning negotiation and accepting the final multi-subject scheduling scheme;

step 6, if the multi-agent scheduling is finished, outputting a scheduling scheme pi with the negotiation consistency between the resource autonomous agent and all the user autonomous agents_0,1,...,m(ii) a Otherwise, scheduling the processing tasks with the negotiation not being consistent except the negotiation constraint set R by adopting a preset arbitration rule to generate a final multi-main-body scheduling scheme pi_0,1,...,mAnd output.

Images