JP4346073B2 - Management apparatus, management method, and management program - Google Patents

Description

  The present invention relates to a management apparatus and a management method for managing a multi-agent system configured by combining a plurality of agents modeling an individual that autonomously executes a task imposed on itself using a component assigned to itself And management programs.

  Conventionally, DBR (drum buffer rope) is known as a TOC (constraint theory) method for solving a scheduling problem in a production environment (for example, see Non-Patent Document 1 to Non-Patent Document 3). DBR, for example, explains the progression of the formation. For example, the fastest person and the slowest person are connected with a rope, and the slowest person holds a drum to set the rhythm of the formation. People follow this rhythm. The distance between the earliest person and the slowest person does not exceed the length of the rope. Others can walk faster than the slowest, so you can always close the distance. The rope has room to prevent sudden deadlocks. As a result, everyone in the line is subordinate to the rhythm of the slowest person, walking on average, and not faster than the slowest person.

Here, when the above-described DBR is used in a production system, the drum represents a rhythm set by an agent that models a process that is a restriction of the system, for example, a production pace set in the slowest process of the production system, etc. It is. The rope represents a time that can be relaxed with an agent that is a restriction of the system, and is, for example, a process that leads from the restriction process to the most upstream process of the system. The length of the rope corresponds to the time required for production and the in-process inventory. The buffer represents a time margin set in order to cope with the uncertainty of the system, and is, for example, a stock of work-in-process before a process or a stock before transportation. The size of the buffer represents, for example, the size of a warehouse.
E. M.M. Goldratt and J.M. Cox; The Goal: A Process of Ongoing Improvement, The North River Press (1984) E. Noren, D.C. Smith and J.M. T.A. McThey; The Theory of Constants and Its Implications for Management Accounting, The North River Press (1995) D. The Measurement Nightmare; How the Theory of Constants Can Resolving Conflicting Strategies, Policies and Measurement, St. Lucie Press (2000)

  As described above, in the production management method based on DBR, it is necessary to provide a constraint condition to recognize a constraint process and resolve it. If this constraint is made too strict, the number of agents that become constraints will increase and the production system will stop. On the other hand, if the constraint conditions are relaxed too much, the number of restricted agents will decrease and the production system will move, but the throughput of the production system will decrease.

  In this way, with the conventional DBR-based production management method, it is difficult to determine how much the constraint should be relaxed, and it is difficult to recognize the bottleneck that becomes a system constraint and eliminate the bottleneck. Met.

  An object of the present invention is to provide a management apparatus, a management method, and a management program capable of recognizing a bottleneck that is a limitation of a system more quickly, and eliminating the bottleneck to improve system activity. That is.

The management device according to the present invention includes a creation unit that creates loading information representing a loading state in which a product is loaded on a pallet for each pallet, and a loading of the product on the pallet . As a constraint imposed on the processing , a reception unit that accepts a setting by a user of a mandatory constraint that must be satisfied and a relaxable constraint condition that can be relaxed, the essential constraint condition that is received by the reception unit, and the Restriction condition storage means for storing relaxable constraint conditions; Relaxation means for creating relaxation constraint conditions that relax the relaxable constraint conditions stored in the constraint condition storage means; and the essential constraint conditions and relaxable constraint conditions first group stowage data satisfying, together with the mandatory constraints and relaxed limitations are met prior to mitigate relaxing possible constraint conditions Second group stowage data that does not satisfy the said essential constraints and mitigate stowage data which does not satisfy at least one of the constraints and classifying means for classifying the third group, and the stowage data, the first group a classification information storage means with the product information of the second group and third group is stored as a classification information associates the classified group by said classifying means, are classified into a third group by said classifying means against stowage data to evaluate the achievement of the said essential constraints and the relaxed constraint stowage data should achieved and the value is higher achievement of increases small and the essential constraints for each and Based on the evaluation function defined for each relaxation constraint condition, the evaluation value for each essential constraint condition and each relaxation constraint condition is calculated. The relaxed the evaluation value for each constraint calculates the total evaluation value by adding all the third group calculated the total evaluation value is stored stowage data having the maximum in the classification information storage means and extraction means for extracting an information with restriction product from the stowage information being classified in the mitigation means, the essential constraints and relaxed constraint constraint stowage information extracted by said extraction means The relaxation constraint is further relaxed so as to satisfy.

In the management apparatus according to the present invention, the loading information satisfying the essential constraint condition and the relaxable constraint condition is satisfied in the first group, the essential constraint condition and the relaxation constraint condition are satisfied , and the relaxable constraint condition before relaxation is not satisfied. The loading information is classified into the second group, the loading information that does not satisfy at least one of the essential constraint condition and the relaxation constraint condition is classified into the third group, and the entire system is selected from the loading information classified into the third group. since the extracted constraint stowage information as a constraint, it is possible to recognize more quickly a stowage data as a system constraint. Further, since the relaxation constraint condition is relaxed so that the extracted constraint product information can be classified into either the first group or the second group, the state of the product product information can be improved to an active state. . As a result, it is possible to recognize a bottleneck that is a limitation of the system more quickly, and it is possible to eliminate the bottleneck and improve the activity of the system.

Further, it is possible to accurately and quickly extract the constraint product information from the value of the evaluation function.

Further, in the management device, the mitigation unit may include a product included in the constraint product information and a product included in the other product information between the constraint product information and other product information. Exchanging, causing the constraint loading information to execute an exchange process for satisfying the essential constraint condition and the relaxation constraint condition, and at least the constraint loading information satisfying the essential constraint condition and the relaxation constraint condition and loading the constraint violation When the information is low, the exchange process is established, and the classification means reclassifies each product information after the exchange process into the first to third groups, and the relaxation means When the product information belonging to the third group exists, the relaxation constraint condition is further relaxed. In this case, it is possible to improve the system activity by improving the state of the constraint loading information to the active state.

Further, in the above management device, when the exchange process is not established, the extraction means is a product other than the constrained product information for which the negotiation has not been established from the product information of the third group. extracting with information as a new constraint stowage data, said reducing means, the new constraints the urging new constraint product information the essential constraints and relaxed constraints between the stowage information and other stowage data The replacement process for satisfying the condition is executed again.

In this case, the restriction product with information exchange process is not satisfied, but it is possible to participate in the exchange process as collaborators, since it is not to become the leading role of the exchange process, it is possible to prevent the deadlock of negotiations.

Management method according to the present invention, the computer includes a generating step of generating a stowage data representative of the loading conditions the product has been stowed in the pallet in the stowage of the pallet products for each of the pallet, the computer, the As a constraint imposed on the product loading process in the pallet, a reception step for accepting a user setting of an essential constraint that must be satisfied and a relaxable constraint that can be relaxed, and the computer includes the reception step A storage step for storing the essential constraint condition and the relaxable constraint condition received in the constraint condition storage unit; and a relaxation constraint in which the computer relaxes the relaxable constraint condition stored in the constraint condition storage unit A relaxation step for creating a condition; and The ability satisfying the constraint condition stowage data to the first group, together with the mandatory constraints and relaxed constraint conditions are satisfied, the stowage information which does not satisfy the pre-relaxation possible constraint to relax in the second group, the essential constraints And a classification step for classifying product information that does not satisfy at least one of the relaxation constraint conditions into a third group, and the computer includes the product information , the first group, the second group, and the third group. a classification information storage step stowage data is stored in the classification information storage means as the classification information in association with classified group in the classification step, the computer, third stowage classified into groups in the classification step the information to evaluate the achievement of the said essential constraints and the relaxed constraint stowage data should achieved and the value The greater the degree of achievement, the moreover, the evaluation values for each of the essential constraint conditions and the relaxation constraint conditions were calculated based on the evaluation function defined for each of the essential constraint conditions and the relaxation constraint conditions. A total evaluation value is calculated by adding all the evaluation values for each essential constraint condition and each relaxation constraint condition, and product information that maximizes the calculated total evaluation value is stored in the classification information storage unit . includes an extraction step of extracting as information with restriction product from the third stowage data are classified into groups that are, the relaxation step, the computer, restriction stowage information extracted in said extraction step However, the relaxation constraint condition is further relaxed so as to satisfy the essential constraint condition and the relaxation constraint condition.

The management program according to the present invention includes a creating means for creating, for each pallet, a loading information in which a product in a product pallet is loaded on the pallet, and a product loading on the pallet . As a constraint imposed on the processing , a reception unit that accepts a setting by a user of a mandatory constraint that must be satisfied and a relaxable constraint condition that can be relaxed, the essential constraint condition that is received by the reception unit, and the Restriction condition storage means for storing relaxable constraint conditions; Relaxation means for creating relaxation constraint conditions that relax the relaxable constraint conditions stored in the constraint condition storage means; and the essential constraint conditions and relaxable constraint conditions first group stowage data satisfying, together with the mandatory constraints and relaxed limitations are met prior to mitigate relaxable The stowage data which does not satisfy the about conditions in the second group, and classifying means for classifying the stowage information which does not satisfy at least one of the essential constraints and relaxed constraints third group, and the stowage data, the 1 group, and the classification information storage means with the product information of the second group and third group is stored as a classification information associates the classified group by said classifying means, the third group by the classifying means With respect to the classified loading information , the degree of achievement of the essential constraint condition and the relaxation constraint condition to be achieved by the loading information is evaluated, and the degree of achievement increases as the value decreases, and the essential constraint condition Based on the evaluation function determined for each and each relaxation constraint, the evaluation value for each essential constraint and each relaxation constraint is calculated, and the calculated essential constraint Calculates each and a total evaluation value by adding all the evaluation value for each of the relaxed constraint conditions, the calculated the total evaluation value is stored stowage data having the maximum in the classification information storage means 3 cause the computer to function as an extraction means for extracting an information with restriction product from the stowage data are classified into groups, the relieving means is constrained stowage information extracted by the extraction means the essential constraints and The relaxation constraint condition is further relaxed so as to satisfy the relaxation constraint condition.

  ADVANTAGE OF THE INVENTION According to this invention, while being able to recognize the bottleneck which becomes a restriction | limiting of a system earlier, the said bottleneck can be eliminated and the activity of a system can be improved.

  Hereinafter, a production management system using a production management apparatus according to an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram for explaining a MAS (multi-agent system). The MAS is a system configured by combining a plurality of agents modeling an individual who autonomously executes a task imposed on itself using components assigned to the MAS. In this embodiment, MAS in a series of production, distribution, and sales production systems will be described. However, the present invention is not particularly limited to this, and can be similarly applied to other systems. Further, in the present embodiment, a manufacturing factory, a shipping company, a sales base, a pallet, a truck, and the like correspond to individuals that autonomously execute tasks assigned to themselves using components assigned to themselves.

  The production system shown in FIG. 1 can be divided into an upper agent system of a manufacturing agent, a distribution agent, and a sales agent. The manufacturing agent includes a plurality of manufacturing factories, and the distribution agent includes a plurality of shipping companies. Can be divided into subordinate agent systems composed of a plurality of sales bases. Here, agents having the same components, constraints, and tasks are assumed to be homogeneous agents. For example, in the manufacturing agent of FIG. 1, the factory A, the factory B, and the factory C are made up of processing raw materials, processing machines, and operators. An agent having at least one of a component, a constraint condition, and a task is defined as a foreign agent. For example, the manufacturing agent, the distribution agent, and the sales agent in the higher-level agent system can be different agents because they have different components, constraints, and tasks.

  Homogeneous agents and heterogeneous agents differ in the form of competition and cooperation between agents. That is, for example, heterogeneous agents rarely compete for the same resources, but because they are interdependent, synchronous behavior among agents is important as in centipede competition. In addition, in the case of homogeneous agents, competing resources are almost the same, competitiveness for resources is strong, and dependency is weak. Therefore, cooperation between agents is important when resources are limited. Therefore, in the following description, two embodiments of homogeneous agents and heterogeneous agents will be described.

  First, a production management system between homogeneous agents will be described. FIG. 2 is a diagram illustrating an example of a hardware configuration of the production management apparatus according to the embodiment of the present invention. A production management apparatus 10 shown in FIG. 2 includes a personal computer or the like, and includes an input device 1, a ROM (read only memory) 2, a CPU (central processing unit) 3, a RAM (random access memory) 4, and an external storage device 5. The display device 6 and the recording medium driving device 7 are provided. Each block is connected to an internal bus, and various data and the like are input / output between the blocks via this bus, and various processes are executed under the control of the CPU 3.

  The input device 1 includes a keyboard, a mouse, and the like, and is used by a user to input constraint conditions and the like. The ROM 2 stores a system program and the like in advance. The external storage device 5 is composed of a hard disk drive or the like, and stores a production management program and the like to be described later. The CPU 3 reads a production management program or the like from the external storage device 5 and executes a production management process or the like described later to control the operation of each block. The RAM 4 is used as a work area for the CPU 3. The display device 6 includes a CRT (cathode ray tube), a liquid crystal display device, or the like, and displays various screens under the control of the CPU 3.

  The production management program or the like may be recorded on a computer-readable recording medium 8 composed of a CD-ROM, DVD-ROM, flexible disk, and the like. In this case, a production management program or the like read from the recording medium 8 using the recording medium driving device 7 composed of a CD-ROM drive, a DVD-ROM drive, a flexible disk drive, and the like is installed in the external storage device 5. . When the production management program or the like is stored in another computer or the like connected via a network, the production management program or the like may be downloaded from the computer or the like via the network.

  Next, the main functions of the production management apparatus configured as described above will be described. FIG. 3 is a block diagram showing an example of main functions of the production management apparatus shown in FIG.

  The production management apparatus 10 shown in FIG. 3 is functionally configured to include a control unit 100, a storage unit 200, and an input unit 300. The control unit 100 includes a CPU 3 and the like, and the CPU 3 and the like execute a production management program and the like stored in the external storage device 5, thereby causing a constraint condition setting unit 101, an agent classification unit 102, a constraint agent recognition unit 103, It functions as a constraint agent improvement unit 104, an agent dependent unit 105, a determination unit 106, and a constraint condition relaxation unit 107. The storage unit 200 includes the external storage device 5 and the like, and functions as the constraint condition storage unit 201 and the agent classification information storage unit 202 by the CPU 3 and the like executing a production management program stored in the external storage device 5. To do. The input unit 300 includes the input device 1 and the like, and functions as the constraint condition input reception unit 301 by the CPU 3 and the like executing a production management program and the like stored in the external storage device 5.

  The constraint condition input receiving unit 301 receives a constraint condition input by the user. Specifically, the constraint condition input receiving unit 301 receives a constraint condition desired by the user from the constraint conditions presented in advance. The constraint condition storage unit 201 stores the constraint condition received by the constraint condition input reception unit 301.

  Based on the classification rules stored in advance, the constraint condition setting unit 101 divides the constraint conditions stored in the constraint condition storage unit 201 into essential constraint conditions that must be satisfied and relaxable relaxable constraint conditions. Set. There are many complicated constraints in a real system, and it is difficult to satisfy all the constraints at the same time. These constraints must be satisfied and relaxable and relaxable. There are constraints. Therefore, in order to eliminate the bottleneck of the system, the constraint condition setting unit 101 must satisfy the constraint condition stored in the constraint condition storage unit 201 and the relaxable constraint condition that can be relaxed. The relaxation parameters are provided in the relaxable constraint conditions, and the constraint parameters are relaxed to a certain degree by changing the relaxation parameters according to the bottleneck situation. For example, focus on container efficiency to reduce transportation costs. Each product has two delivery dates: a delivery date and a local delivery date. Generally, products are loaded into containers in units of local delivery dates. If the efficiency is low, the efficiency of the containers can be improved by relaxing the relaxable constraints on local delivery dates and loading them in the delivery delivery date. Instead of providing the constraint condition setting unit 101, the user may set the essential constraint condition and the relaxable constraint condition using the constraint condition input receiving unit 301.

  In order to simplify the complexity of the system, it is assumed that each system is composed of several subsystems (agents), and system goals are assigned to each subsystem. To achieve system goals, each subsystem goal must be achieved. To achieve this goal, each subsystem negotiates with each other. The constraint condition setting unit 101 defines an evaluation function and an establishment rule for evaluating negotiation by each subsystem based on a target and a relaxable constraint condition. The evaluation function represents that the situation of constraint violation or the situation where the target cannot be achieved is severe as the value increases.

  The system is composed of several agents, and each agent competes and uses elements (resources) under the conditions in which it is placed using a heuristic method in order to achieve its goals. Since agents have different capabilities and external environments, the status of each agent changes as a result of competition between agents, and agents that satisfy the constraint condition and agents that do not satisfy the constraint condition occur simultaneously.

  In order to make this situation easier to recognize, the agent classifying unit 102 selects all agents satisfying all the constraint conditions, that is, the essential constraint conditions and the relaxable constraint conditions, based on the constraint conditions set by the constraint condition setting unit 101. In one group, agents that satisfy the essential constraints and relaxation constraints are in the second group, and in the third group, agents that do not satisfy at least one of the mandatory constraints and relaxation constraints are the constraint violations and system bottlenecks. Classify. The agent classification unit 102 stores in the agent classification information storage unit 202, as agent classification information, which group of the first to third groups each agent is classified. The relaxed constraint condition is a relaxed constraint condition relaxed by a constraint condition relaxing unit 107 described later.

  The agent classification information storage unit 202 associates an agent with a group to which the agent belongs among the first group, the second group, and the third group, and stores them as agent classification information.

  The restriction agent recognition unit 103 recognizes a restriction agent that is a restriction on the entire system among the agents classified into the third group by the agent classification unit 102. For example, since the agent that becomes the constraint violation is classified into the third group by the agent classification unit 102, the constraint agent recognition unit 103 determines the target that the agent should achieve for the agent classified in the third group. The value of the evaluation function that evaluates the degree of achievement is obtained, and the agent having the largest evaluation function value is recognized as a constraint agent that prevents the system from reaching the goal. Further, when the negotiation is not established by the restriction agent improvement unit 104 (to be described later), the restriction agent recognizing unit 103 sets other agents other than the restriction agent for which the negotiation is not established among the agents in the third group as a new restriction. Extract as an agent.

  The constraint agent improvement unit 104 causes the constraint agent to negotiate for satisfying the mandatory constraint condition and the relaxation constraint condition between the constraint agent and another agent, and all the agents satisfy the mandatory constraint condition and the relaxation constraint condition. Negotiate if the overall system situation does not deteriorate. At this time, the restriction agent improvement unit 104 preferentially uses the resources available to the restriction agent recognized by the restriction agent recognition unit 103, and improves the restriction state. For example, when the machining process is a constraint agent, the machining speed is improved by increasing the operating time of machines and personnel.

  The agent subordinate unit 105 subordinates agents other than the restriction agent recognized by the restriction agent recognition unit 103 to the restriction agent. Here, the behavior is different between the homogeneous agent and the heterogeneous agent. Homogeneous agents compete for the same resources, so a certain amount of coordination and concession is required. On the other hand, heterogeneous agents do not compete for the same resources, but need to synchronize their actions. If the process that becomes a constraint is not improved, the result will be in stock no matter how much the previous process is advanced, so the overall throughput cannot be improved. Moreover, if cooperation, concessions, and synchronization are performed under no conditions, the progress speed of the entire system becomes slow, and as a result, the throughput cannot be improved. Therefore, it is necessary to set conditions for cooperation, concession and synchronization between agents.

  The determination unit 106 refers to the agent classification information stored in the agent classification information storage unit 202 and determines whether there is an agent belonging to the third group. Further, the determination unit 106 performs processing for classifying the agent by the agent classification unit 102, processing for recognizing the restriction agent by the restriction agent recognition unit 103, processing for improving the restriction agent by the restriction agent improvement unit 104, and restriction by the agent subordinate unit 105 It is determined whether or not the process of making an agent other than the agent subordinate to the constraint agent and the process of relaxing the constraint condition by the constraint condition relaxation unit 107 have been repeated a predetermined number of times. As the predetermined number of times, a number of times that the constraint violation cannot be resolved is set in advance. Furthermore, the determination unit 106 determines whether or not the value of the relaxation parameter has reached the maximum value.

  When the determination unit 106 determines that there is an agent belonging to the third group, the constraint condition relaxing unit 107 further relaxes relaxable constraint conditions that can be relaxed among the constraint conditions stored in the constraint condition storage unit 201. To create relaxed constraints. Depending on the severity of the constraints and the finite nature of the resources, there is a possibility that the constraint agents will not be improved if they are subordinated with conditions. Therefore, it is necessary to provide the system with sufficient flexibility, and the constraint condition that causes deadlock is relaxed, and the active state is changed again in cooperation with other agents.

  For example, a dispatching agent cannot load packages with partially different delivery dates due to constraints. Since the shipping space is limited, it is not possible to increase the number of containers. In such a case, loading is performed so as to increase the filling rate of the container by relaxing the relaxable constraint condition. By this relaxation, the same condition of the load is divided and loaded, making it difficult for the wholesale agent to work, but it is possible to eliminate the deadlock. It should be noted that which relaxable constraint conditions are relaxed and which relaxable constraint conditions are not relaxed are determined by the status of agents belonging to the third group. In addition, the extent to which the relaxable constraint condition is relaxed is determined by the status of the entire system and the mitigation status of the second group.

  In this embodiment, the constraint condition input reception unit 301 and the constraint condition setting unit 101 correspond to an example of a reception unit, the agent classification unit 102 corresponds to an example of a classification unit, and the constraint agent recognition unit 103 serves as an extraction unit. The constraint agent improvement unit 104, the agent dependent unit 105, and the constraint condition relaxation unit 107 correspond to an example, and correspond to an example of a relaxation means.

  Next, production management processing by the production management apparatus 10 of the production management system configured as described above will be described. FIG. 4 is a flowchart for explaining the production management process of the production management apparatus 10 shown in FIG. Note that the production management process shown in FIG. 4 is a process performed by the CPU 3 executing a production management program stored in advance.

  First, in step S <b> 1, the constraint condition setting unit 101 can relax the essential constraint condition that must be satisfied by the constraint condition received by the constraint condition input reception unit 301 and stored in the constraint condition storage unit 201. It is classified and set as a relaxable constraint.

  Next, in step S <b> 2, the agent classification unit 102 assigns agents satisfying the essential constraint condition and the relaxable constraint condition to the first group based on the constraint condition set by the constraint condition setting unit 101, and the essential constraint condition and relaxation. Agents that satisfy the constraint condition are classified into the second group, and agents that do not satisfy at least one of the essential constraint condition and the relaxation constraint condition are classified into the third group. At this time, the agent classification unit 102 stores in the agent classification information storage unit 202 as agent classification information which group of the first group, the second group, and the third group each agent belongs to.

  At the first time, the relaxation constraint that relaxed the relaxable constraint is not created, and the agent that satisfies the mandatory constraint and the relaxable constraint does not satisfy at least one of the mandatory constraint and the relaxable constraint Since there are only agents, each agent is classified into a first group and a third group.

  Next, in step S3, the restricted agent recognition unit 103 reads the agent classification information stored in the agent classification information storage unit 202, and among the agents classified into the third group by the agent classification unit 102, the entire system Recognize the constraint agent that becomes the constraint of.

  Next, in step S <b> 4, the constraint agent improvement unit 104 improves the constraint status of the constraint agent recognized by the constraint agent recognition unit 103. The constraint agent improvement unit 104 causes the constraint agent to negotiate for satisfying the mandatory constraint condition and the relaxation constraint condition between the constraint agent and another agent, and all the agents satisfy the mandatory constraint condition and the relaxation constraint condition. Negotiate if the overall system situation does not deteriorate.

  For example, even if the negotiation by two agents, the restricted agent classified as the third group and the unconstrained agent classified as the first or second group, is executed, and the situation in the entire system of the unconstrained agent deteriorates, If the situation is not worse than that of the other restricted agent, it is assumed that the negotiation is successful. When the negotiation is established, the agent classification unit 102 reclassifies each agent after the negotiation into the first to third groups.

  Next, in step S <b> 5, the agent subordinate unit 105 subordinates other agents other than the restriction agent recognized by the restriction agent recognition unit 103 to the restriction agent.

  Next, in step S <b> 6, the determination unit 106 determines whether the negotiation is established by the constraint agent improvement unit 104. If it is determined that the negotiation has been established (YES in step S6), the process proceeds to step S7. If it is determined that the negotiation has not been established (NO in step S6), the process proceeds to step 8.

  When the negotiation is established, in step S7, the agent classification unit 102 reclassifies each agent after the negotiation into the first to third groups based on the constraint condition set by the constraint condition setting unit 101.

  On the other hand, if the negotiation has not been established, in step S8, the determination unit 106 classifies the constraint agent for which the negotiation has not been established into a check buffer. In this embodiment, a set called a check buffer is prepared in order to prevent a deadlock of negotiation. Even if all the resources present in the constraint agent are used as negotiation elements, if the negotiation is not established, it is determined that the constraint agent cannot be improved at the present time, and is classified into a check buffer. Agents classified in the check buffer are inactive agents and cannot play a leading role in negotiations, but participate in negotiations as cooperators. In this way, by classifying all the constraint agents that failed to negotiate into check buffers, duplicate search can be prevented.

  Next, in step S9, the determination unit 106 determines whether there is an agent belonging to the second group or the third group. If it is determined that there is an agent belonging to the second group or the third group (YES in step S9), the process returns to step S3, and the constraint agent recognition unit 103 recognizes the constraint agent again. If it is determined that there is no agent belonging to the second group or the third group (NO in step S9), the constraint violation has been resolved or an agent other than the agent classified in the first group is classified in the check buffer. Therefore, no more negotiations can be made, and the process proceeds to step S10.

  If there is no agent belonging to the second group or the third group, in step S10, the determination unit 106 determines whether or not the processing from step S2 to step S13 has been repeated a predetermined number of times. If it is determined that the predetermined number of times has been repeated (YES in step S10), the process ends, and if it is determined that the number of times has not been repeated in advance (NO in step S10), The process proceeds to step S11.

  If it has not been repeated a predetermined number of times, in step S11, the determination unit 106 determines whether or not the value of the relaxation parameter has reached the maximum value. Here, if it is determined that the value of the relaxation parameter has reached the maximum value (YES in step S11), the process ends, and if it is determined that the value of the relaxation parameter has not reached the maximum value (NO in step S11). ), The process proceeds to step S12.

  If the value of the relaxation parameter is not the maximum value, in step S12, the determination unit 106 refers to the agent classification information stored in the agent classification information storage unit 202, and the constraint belongs to the third group or the check buffer. Determine whether an agent exists. If it is determined that there is a constraint agent belonging to the third group or check buffer (YES in step S12), the process proceeds to step S13, and it is determined that there is no constraint agent belonging to the third group or check buffer. If so (NO in step S12), it is determined that the system bottleneck has been eliminated, and the process ends.

  If it is determined that there is a constraint agent, in step S13, the constraint condition relaxing unit 107 relaxes the relaxable constraint conditions among the constraint conditions stored in the constraint condition storage unit 201, and stores the constraint condition storage. The constraint condition stored in the unit 201 is reset. Then, in step S2, the agent classification unit 102, based on the constraint condition relaxed by the constraint condition relaxation unit 107, each agent that constitutes the system, a first group that satisfies the essential constraint condition and the relaxable constraint condition, A second group that satisfies the essential constraint condition and the relaxation constraint condition is classified into a second group that does not satisfy at least one of the essential constraint condition and the relaxation constraint condition. It is determined that the processing from step S2 to step S13 has been repeated a predetermined number of times in step S10, the value of the relaxation parameter has reached the maximum value in step S11, or no agent belonging to the third group exists in step S12. Repeat until.

  As described above, an agent satisfying the essential constraint condition and the relaxable constraint condition is satisfied in the first group, and an agent satisfying the essential constraint condition and the relaxation constraint condition is satisfied in the second group, and at least one of the essential constraint condition and the relaxation constraint condition is satisfied. The agents that are not restricted are classified into the third group, and the restriction agents that restrict the entire system are extracted from the agents classified into the third group, so that the agent that becomes the restriction of the system can be recognized more quickly. . Further, since the relaxation constraint condition is relaxed so that the extracted constraint agent can be classified into either the first group or the second group, the state of the agent can be improved to an active state. As a result, it is possible to recognize a bottleneck that is a limitation of the system more quickly, and it is possible to eliminate the bottleneck and improve the activity of the system.

  Further, the constraint agent recognition unit 103 extracts a constraint agent based on an evaluation function that evaluates the degree of achievement of the target that the agent should achieve for the agent classified in the third group. The constraint agent can be extracted accurately and quickly.

  Further, the constraint agent improvement unit 104 causes the constraint agent to perform negotiations for satisfying the essential constraint condition and the relaxation constraint condition between the constraint agent and another agent, and all the agents set the mandatory constraint condition and the relaxation constraint condition. When the conditions are satisfied and the overall system condition does not deteriorate, the negotiation is established. The agent classification unit 102 reclassifies the negotiated agents into the first to third groups, and the constraint relaxation part 107 performs the first classification after the reclassification. When there are agents belonging to the three groups, the relaxation constraint condition is further relaxed, so that the state of the constraint agent can be improved to the active state and the system activity can be improved.

  Furthermore, when the negotiation is not established, the constraint agent recognition unit 103 extracts, as a new constraint agent, agents other than the constraint agent that did not establish the negotiation from the third group agents. The improvement unit 104 re-executes negotiation between the new constraint agent and another agent so that the new constraint agent satisfies the essential constraint condition and the relaxation constraint condition. Therefore, the restriction agent for which the negotiation has not been established can participate in the negotiation as a collaborator, but since it does not become the main player of the negotiation, the deadlock of the negotiation can be prevented.

  Next, the production management process will be described in detail by taking as an example the loading of a load onto a pallet by a homogeneous agent. FIG. 5 is a diagram for specifically explaining the production management process, taking as an example the loading of a load onto a pallet by a homogeneous agent. In the following description, an operator is an operator who loads a load on a pallet. The worker H1 shown in FIG. 5 is a skilled worker who is skilled in loading work on the pallet, and the worker H2 and the worker H3 can be loaded earlier than the worker H1 and earlier than the worker H4. The worker H4 is a beginner.

  Each product (luggage) has a delivery date, and each worker loads a certain amount of product on the pallet. The completion time of the last person is the collection work completion time. Then, the next step, shipment work, that is, the dispatch of trucks, starts. Therefore, an operator who loads a pallet is required to complete the pallet loading operation within a limited time, that is, to synchronize the operation. However, as shown in FIG. 5, since the worker H1 is an expert, the speed of the work is fast and can be completed earlier than a limited time. Therefore, the pallet loaded by the skilled worker H1 can satisfy the essential constraint condition and the relaxable constraint condition, and is classified into the first group B1. On the other hand, since worker H4 is a beginner and the speed of work is slow, it cannot be completed within a limited time. Therefore, the pallet loaded by the worker H4 who is a beginner cannot satisfy at least one of the essential constraint condition and the relaxation constraint condition, and thus is classified into the third group B3. The worker H2 and the worker H3 are slower than the worker H1, can be loaded earlier than the worker H4, and do not affect the whole, so work at a free speed between the worker H1 and the worker H4. be able to. Therefore, the pallet loaded by the worker H2 and the pallet loaded by the worker H3 can satisfy the essential constraint condition and the relaxation constraint condition, and are classified into the second group B2.

  FIG. 6 is a diagram for explaining a pillar created when a product is collected, FIG. 6A is a diagram illustrating an example of a product before being stacked, and FIG. It is a figure which shows an example of the pillar using the product of Fig.6 (a). If the product is large, a pillar with a constant height and a constant weight is created and the pillar is placed in a truck or container. The pillar component is the product, the purpose is to meet the constraints and make the best use of the limited space and weight, and the task is to find and pack a good product to achieve that purpose. is there. Therefore, each pillar is considered to be an autonomous agent. For example, when the products shown in FIG. 6A are stacked, four pillars shown in FIG. 6B can be created.

The constraint condition input receiving unit 301 receives a constraint condition input by the user and stores it in the constraint condition storage unit 201. Here, it is assumed that the constraint conditions received by the constraint condition input receiving unit 301 are the following (A) to (H).
(A) Stack the pillars so that they do not exceed the height of the truck.
(B) The columns are stacked so that the weight does not exceed the maximum transport weight.
(C) Load the same destination on the same pillar.
(D) Load the same delivery date on the same pillar.
(E) Stack so that the heavy load is on the bottom and the light load is on the top.
(F) Load the items so that the large load is on the bottom and the small load is on the top.
(G) Stack the columns with the same height as possible.
(H) Stack the same product as much as possible.

  Next, the constraint condition setting unit 101 must be satisfied by the constraint conditions (A) to (H) received by the constraint condition input reception unit 301 and stored in the constraint condition storage unit 201. Set the constraint condition and the relaxable constraint condition that can be relaxed. Here, the constraint condition setting unit 101 sets the constraint conditions (A), (B), (C), and (D) as essential constraint conditions, and the constraint conditions (E), (F), ( G) and (H) are set as relaxable constraints. Then, the constraint condition setting unit 101 defines relaxation parameters in the constraint conditions (E), (F), and (G) that are relaxable constraint conditions, and limits the degree to which the constraint conditions are relaxed. Furthermore, the constraint condition setting unit 101 formulates constraint conditions and purposes, and defines an evaluation function based on the formulated constraint conditions and purposes.

  Here, an example of formulating the constraint conditions (A) to (H) will be described. The constraint condition (A) which is an essential constraint condition is expressed by the following formula (1), the constraint condition (B) is expressed by the following formula (2), and the constraint condition (C) is expressed by the following formula (3). The constraint condition (D) is expressed by the following equation (4).

∀m _k , mh _k ≦ mh _max (1)
∀m _k , mg _k ≦ mg _max (2)
∀P ^k _i , P ^k _j ∈m _k , i ≠ j, then P ^k _i , P ^k _j ∈ same target (3)
∀P ^k _i , P ^k _j ∈m _k , i ≠ j, then P ^k _i and P ^k _j have the same due date (4)

In the above formula (1), m _k represents the _k-th column, mh _k represents the height of the column, and mh _max represents the maximum height of the column that can be relaxed. In the above formula (2), m _k represents a column, mg _k represents the weight of the column, and mg _max represents the maximum weight of the column that can be relaxed. In the above formulas (3) and (4), P ^k _i and P ^k _j represent the i, j products subordinate to the _{k th} column m _k , and P ^k _i and P ^k _j are different. Indicates a product.

Further, the length ml _k , the width mw _k , the height mh _k and the weight mg _k of the pillar m _k satisfy the conditions of the following expressions (5) to (8).

 The constraint condition (E), which is a relaxable constraint condition, is represented by the following expression (9), the constraint condition (F) is represented by the following expressions (10) and (11), and the constraint condition (G) is It is represented by the following formula (12). The constraint (H) is considered after satisfying the essential constraints when creating the initial solution.

g ^k _{i + 1} ≦ (1-ε _g ) min {g ^k _i , g ^k ₁ } (9)
{1- (4/3) ε _l } max {l ^k _i , l ^k ₁ } ≦ l ^k _{i + 1} ≦ {1 + ε _l } min {l ^k _i , l ^k ₁ } (10)
{1- (4/3) ε _w } max {w ^k _i , w ^k ₁ } ≦ w ^k _{i + 1} ≦ {1 + ε _w } min {w ^k _i , w ^k ₁ } (11)
(1-ε _h ) mh _max <mh _k ≦ mh _max (12)

In the above equations (9) to (12), l ^k _i , w ^k _i and g ^k _i represent the length, width and weight of the product P ^k _i , and ε _l , ε _w , ε _g and ε _h represents a relaxation parameter related to the length, width, weight, and height of an adjacent column of the products arranged above and below.

The relaxation parameters ε _l , ε _w , ε _g , and ε _h are 0 ≦ ε _l ≦ ε _lmax , 0 ≦ ε _w ≦ ε _wmax , 0 ≦ ε _g ≦ ε _gmax and ε ₀ ≦ ε _h ≦ ε _hmax , respectively. . _Where ε _cmax represents the maximum _relaxable length of the product, ε _wmax represents the maximum _relaxable width of the product, ε _gmax represents the maximum _relaxable weight of the product, and ε _hmax Represents the maximum height at which the pillar can be relaxed. The relaxation parameter indicates a state in which the initial value is not relaxed, and is gradually relaxed according to the constraint violation situation. The maximum value of the relaxation parameter is determined according to the user's request.

Next, the evaluation function will be described. The volume ratio E _k ν of the product P ^k _i is expressed by the following equation (13).

In the above equation (13), E _k ν represents the volume ratio, ν ^k _i represents the volume of the product P ^k _i , ml _k represents the length of the column m _k , and mw _k represents the length of the column m _k . It represents the width, and mh _max represents the maximum height of the pillar m _k that can be relaxed. Further, the volume ratio E _k ν is higher as the value is smaller, and E _k ν = 0 is the optimal solution.

The evaluation function E _{k of the} entire system is expressed by the following equation (14).

_{E k = μ 1 × E kl} + μ 2 × E kw + μ 3 × E kg + μ 4 × E kh + μ 5 × E k ν ···· (14)

In the above equation (14), E _k represents an evaluation function of the entire system, E _kl represents an evaluation function related to the column length, E _kw represents an evaluation function related to the column width, and E _kg represents the column An evaluation function related to weight is represented, E _kh represents an evaluation function related to column height, and μ ₁ , μ ₂ , μ ₃ , μ _4, and μ ₅ represent weighting coefficients of the respective evaluation functions. Further, E _kl , E _kw , E _kg and E _kh in the above formula (14) are represented by the following formulas (15) to (18).

In addition, E _kli in the above formula (15) is represented by the following formula (19), E _kwi in the above formula (16) is represented by the following formula (20), and E _kgi in the above formula (17) is represented by the following formula: (21).

 The agent classification unit 102 sets the essential constraints and the relaxable constraint conditions that can be relaxed on the basis of the constraint conditions received by the constraint condition input reception unit 301 for the workers H1 to H4 that are agents. The first group B1 is satisfied, the second group B2 satisfies the essential constraint condition and the relaxation constraint condition, and the third group B3 does not satisfy at least one of the essential constraint condition and the relaxation constraint condition. That is, the agent classification unit 102 classifies workers that satisfy the essential constraint conditions and satisfy all the relaxable constraint conditions into the first group B1. Further, the agent classification unit 102 classifies workers that satisfy the essential constraint conditions and satisfy at least one relaxable constraint condition into the second group B2. Further, the agent classifying unit 102 classifies workers who do not satisfy even one essential constraint condition or all the relaxable constraint conditions into the third group B3. As a result, in the example of FIG. 5, the worker H1 is classified into the first group B1, the worker H2 and the worker H3 are classified into the second group B2, and the worker H4 is classified into the third group B2. It becomes.

  Subsequently, the constraint agent recognizing unit 103 recognizes the agent that becomes a constraint, and the constraint agent improving unit 104 improves the agent that becomes the constraint recognized by the constraint agent recognizing unit 103.

  FIG. 7 is a diagram for explaining the improvement of the constraint agent by taking as an example the loading of a load onto a pallet by a homogeneous agent. FIG. 7A shows each of the items before being classified by the agent classifying unit 102. FIG. 7B is a diagram illustrating agents classified by the agent classifying unit 102, and FIG. 7C is a diagram illustrating each of the agents after relaxing constraints on weight and size. FIG. 7D is a diagram illustrating the agents after further relaxing the constraints on weight and size.

  As shown in FIG. 7A, the agent a1 has two 50 kg loads of the same size stacked thereon, and further a 38 kg load smaller than the 50 kg load is stacked thereon. Agent a2 is loaded with three pieces of 38 kg of the same size. Agent a3 is loaded with three 55 kg loads of the same size. The agent a4 has two 85 kg loads of the same size stacked thereon, and a 20 kg load larger than the 85 kg load is further stacked thereon. Agent a5 has two small 65 kg loads stacked on a 60 kg load. Agent a6 is loaded with a small 48 kg load on a 65 kg load, and further a small 60 kg load is loaded thereon. Agent a7 is loaded with a large 60 kg load on an 85 kg load, and further loaded with a small 38 g load.

When the agents in the state of FIG. 7A are divided into the first group, the second group, and the third group, they are divided as shown in FIG. 7B. The relaxation parameters ε _l , ε _w , ε _g , and ε _h in FIG. 7B are set to initial values ε _l0 , ε _w0 , ε _g0 , and ε _h0 , respectively. In FIG. 7B, since the constraint conditions are severe, agents a1, a2, a3 are classified into the first group B1, and agents a3, a4, a5, a6, a7 are classified into the third group B3. Therefore, the constraint relaxation part 107 classifies as shown in FIG. 7C by relaxing the constraint on weight and size. That is, by setting the relaxation parameters ε _l , ε _w , ε _g , and ε _h to ε _lt , ε _wt , ε _gt , and ε _ht , respectively, the constraints of the above (E) and (F) are relaxed, As shown in FIG. 7C, the agents a4 and a5 are classified into the second group B2. However, since the constraint conditions are still severe, the deadlock state is reached, and the agents a6 and a7 are classified into the third group B3 as shown in FIG. 7C. In this case, the constraint condition is relaxed by using the on-site solution method. The constraint condition alleviating unit 107 relaxes the constraint conditions (E) and (F). For example, the upper package weight may be heavier to 10 kg than the lower package weight, and the length of the lower package Suppose that the width and the width may be 1/10 of the length and width of the upper luggage. Thus, by setting the relaxation parameters ε _l , ε _w , ε _g , and ε _h to be larger than ε _lt , ε _wt , ε _gt , and ε _ht , respectively, the constraints on weight and size are further relaxed. It is possible to exchange the load n1 of the agent a7 and the load n2 of the agent a1 that are in violation of the constraints.

  The agent classification unit 102 classifies the agents based on the relaxed constraints. As a result, as shown in FIG. 7D, the agent a7 that has violated the constraint moves from the third group B3 to the second group B2, and gains. Since the agent a1 does not satisfy all the constraint conditions, the agent a1 moves from the first group B1 to the second group B2, but the system as a whole is improved because there are fewer agents that violate the constraint. In the state of FIG. 7D, when the determination unit 106 determines whether or not there is an agent belonging to the third group B3, the agent a6 belongs to the third group B3. It is determined that the agent to which it belongs exists. Therefore, the constraint condition relaxation unit 107 relaxes the constraint conditions so that the agent a6 moves from the third group B3 to the second group B2.

  FIG. 8 is a diagram illustrating an example of changes in the number of agents belonging to the first to third groups, where the vertical axis represents the number of agents classified into the first to third groups, and the horizontal axis represents simulation. Represents the number of repetitions.

  As shown in FIG. 8, agents C3 classified into a third group of constraint violations that do not satisfy at least one of the essential constraint conditions and the relaxed relaxation constraint conditions that must be satisfied by negotiation and relaxation of the agents. The number of decreases gradually as the number of simulation iterations increases. In addition, the number of agents C2 classified into the second group satisfying the essential constraint condition and the relaxation constraint condition gradually increases as the number of times of repeating the simulation increases, and the agent classified into the first group satisfying the essential constraint condition. The number of agents C1 classified as C1 is also gradually increasing.

  From the results shown in FIG. 8, it is possible to automatically improve the constraint violation, and to reduce the number of agents classified into the third group and increase the number of agents classified into the first and second groups. it can. Therefore, the constraint relaxation technique in the present embodiment not only relaxes constraint violations, but also increases the chances of avoiding or escaping the deadlock state by changing the search environment due to relaxation. The competition and concessions between agents can also be seen from the vibration status of each curve.

  FIG. 9 is a diagram illustrating an example of the number of agents classified into the check buffer. The vertical axis represents the number of agents classified into the check buffer, and the horizontal axis represents the number of simulation iterations. Yes.

  The number of agents classified in the check buffer shown in FIG. 9 is the number of times that the constraint condition has been relaxed. When agents classified in the check buffer are classified into the second group or the third group, negotiation between agents is not established unless the search situation changes. The reason is that the constraint condition is too severe, and it is necessary to relax the constraint condition. By relaxing the restriction conditions, the negotiation conditions of each agent change, and it becomes possible to establish negotiations between the agents again. Further, useless search can be avoided by the check buffer and the value of the relaxation parameter, and the process can be automatically terminated without reaching a preset number of repetitions. This means that the present technique can be applied to a strict search time given by a real problem.

  In this way, the multi-agent system is composed of homogeneous agents having the same components, constraints, and tasks, and the constraint agent improvement unit 104 uses the drum buffer rope to restrict between the constraint agent and other agents. Let the agent negotiate to meet the constraints. Therefore, it is possible to recognize a bottleneck that is a limitation of a multi-agent system composed of homogeneous agents more quickly, and to eliminate the bottleneck and improve the system activity.

  Next, a production management system between heterogeneous agents will be described. FIG. 10 is a diagram for explaining a production management system between heterogeneous agents in a physical distribution system. The logistics system 400 shown in FIG. 10 includes a packing / collecting process 410, a shipping process 420, a delivery process 430, and the like. Each process has different components (resources), constraint conditions, targets, and evaluation criteria, and is a heterogeneous agent. is there. The order information reception 440 receives order information related to shipping or export instructions, and the automatic simulation 450 simulates all or part of packing, pallet, vehicle loading, and route calculation.

  The packing / collecting process 410 includes a picking operation 411 for picking up a specified product from the inventory and a pallet loading operation 412 for stacking the product on the pallet. Create The shipping process 420 includes an actual vehicle matching 421 for selecting a vehicle (truck) on which the pallet is loaded and a vehicle loading operation 422 for loading the pallet on the vehicle. The pallet created in the packing / pickup process 410 is subjected to the constraint condition. Pack on trucks that meet, and have high volume and weight, and low cost. The delivery process 430 includes a vehicle information registration 431 for registering vehicle information such as a load amount of a vehicle to be delivered and a vehicle / route / time decision 432 for determining a delivery vehicle, a delivery route, and a delivery time. The loaded vehicle determines the optimal cost and route while taking delivery time, operation time and lead time into consideration, and delivers it to the destination 490.

  In addition, when there is a change in an order, a route, or the like in the order information reception 440, or when there is a change in a vehicle situation or weather in the actual vehicle matching 421, the changed information is changed and changed in the plan change 460. The received order information is reflected in the automatic simulation 450 and the picking work 411. In addition, after the pallet loading work 412 is completed, the pallet loading result is corrected in the pallet loading result correction 470, and after the vehicle loading work 422 is finished, the vehicle loading result is corrected, and each result is automatically This is reflected in the simulation 402.

  Heterogeneous agents have different components and do not compete for the same resources, but are interdependent as a chain of MASs. Each agent must achieve the goal in order to achieve the goal of the entire system. Therefore, each agent must cooperate, not just consider their own interests. Therefore, agents that satisfy the essential constraints that must be satisfied and the mitigable constraints that can be relaxed according to the completion time of dispatch and delivery work according to the operating time, lead time, delivery date, route, etc. The situation of the third group to which the agent satisfying the relaxed constraint condition is classified into the second group, the agent that does not satisfy at least one of the mandatory constraint condition and the relaxed constraint condition is classified into the third group, and the agent that is in violation of the constraint belongs The second group serving as a buffer is managed while tracking.

  The configuration of the production management device between the different agents is the same as the configuration of the production management device 10 between the homogeneous agents shown in FIG.

First, the constraint condition input receiving unit 301 receives a constraint condition input by a user and stores it in the constraint condition storage unit 201. Here, it is assumed that the constraint conditions received by the constraint condition input receiving unit 301 are the following (A) to (F).
(A) Be sure to observe the closing time and delivery date (arrival time) of each operation.
(B) Make sure to keep the operating hours of each warehouse and distribution center.
(C) Even if an external element changes, the start time and completion time of each planned work are secured as much as possible.
(D) Increase the volume ratio of the track as much as possible.
(E) Select a truck with the lowest possible cost.
(F) Make delivery time and waiting time as short as possible.

  Next, the constraint condition setting unit 101 sets the constraint conditions (A) and (B) as essential constraint conditions that must be satisfied, and the constraint conditions (C), (D), and (E). And (F) are set to relaxable constraints that can be relaxed. Then, the constraint condition setting unit 101 sets a relaxation parameter related to the start time and completion time of each work, a relaxation parameter related to the truck volume ratio, a relaxation parameter related to the truck cost, and a relaxation parameter related to the delivery time.

  The initial values of the relaxation parameters relating to the start time and completion time of each work are the earliest start time for each work and the completion time that can be completed earliest. The maximum value of the relaxation parameter relating to the start time and completion time of each work is the latest start time of each work that can be started and the latest completion time that can be completed. That is, the start time and completion time of each work can be relaxed to a time that does not violate the essential constraints (A) and (B).

  The maximum value of the relaxation parameter related to the truck volume ratio is set according to the minimum volume ratio specified in advance, but the initial value (minimum value) is set according to the amount of packages to be delivered, for example.

  The maximum value of the relaxation parameter relating to the cost of the truck is determined by the experience of the expert and is the cost that can be allowed at a minimum. Further, the minimum value of the relaxation parameter relating to the cost of the truck is determined in consideration of the constraint conditions (A) and (B) and the situation of constraint violation.

  The delivery time is related to the delivery date and the cost. The longer the delivery time, the higher the cost and the more difficult it is to meet the delivery date. The maximum value (maximum allowable time) of the relaxation parameter related to the delivery time is determined based on the experience of the skilled person. The optimum time is determined in consideration of the constraint conditions (A) and (B) and the situation of constraint violation. To do.

Constraint condition setting unit 101, with a weight and on the target above constraints (C) ~ (F), defined by the equation (22) below the evaluation function E _t to evaluate the negotiations by between each agent To do.

E _t = EC _t + ES _t + ED _t + EP _t (22)

In the above equation (22), E _t is an evaluation function for evaluating the entire system, EC _t is an evaluation function for evaluating a pickup process including a relaxation parameter, and ES _t is a shipment including a relaxation parameter. An evaluation function for evaluating a process, ED _t is an evaluation function for evaluating a delivery process including a relaxation parameter, and EP _t is an evaluation function for evaluating a plan change including a relaxation parameter.

The values of these evaluation functions EC _t , ES _t , ED _t , EP _t vary depending on the nature of the data, limited work time, relaxation of constraints, and negotiation between agents. If the constraint conditions are not strict and the data has sufficient time, a simulation result that does not require much relaxation of the constraint conditions can be obtained. In addition, if pallets and trucks having higher volume ratios can be easily obtained, the number of trucks can be reduced, and the cost of the entire system can be reduced. The three factors of volume ratio, cost and time are dependent on each other. If there is an allowance for delivery, a good solution (a truck with a high volume ratio and a low overall cost) can be obtained, and the cost required for the delivery process can be reduced.

  In the physical distribution system shown in FIG. 10, the order of each process is fixed. Therefore, a buffer of a certain size is provided for each process, and the size of the buffer is adjusted by a relaxation parameter. The size of the buffer is related to the start time and completion time of the corresponding process. Recognizing that the slowest process in the system is the limiting process, the process sets the pace of the entire system. Other processes adjust the start time and completion time according to the pace set by the restrictive process. For example, a precondition that the shipping process (vehicle loading operation) in FIG. 10 can be started is that the packaging / collecting process (picking operation and pallet loading operation) has been completed. If this condition is not satisfied, the vehicle loading operation cannot be performed. In this case, it is considered that there is a problem in the packing / collecting process. However, if the completion time cannot be observed due to the late arrival of the vehicle, extended working hours, lack of workers, or low work level of workers, it is considered a problem in the shipping process. It is done. Therefore, the agent classifying unit 102 assigns agents that satisfy the essential constraint condition and the relaxable constraint condition to the first group and agents that satisfy the mandatory constraint condition and the relaxation constraint condition to the second group according to the start time and the completion time including the relaxation parameter. Agents that do not satisfy at least one of the essential constraint condition and the relaxation constraint condition are classified into a third group.

  FIG. 11 is a diagram for explaining determination of a delivery route (route). FIG. 11 shows an example in which packages are delivered from warehouse A, warehouse B, and warehouse C to base A. Note that the collection work at each warehouse is 2 hours, and the shipping work is 2 hours. In addition, in the case of additional loading via a plurality of warehouses, the collection work is performed between the shipment and the transportation time in the previous warehouse without considering the collection time from the second passing warehouse. Furthermore, a truck with a maximum loading weight of 4 tons costs 70,000 yen per hour, a truck with a maximum loading weight of 10 tons costs 10,000 yen per hour, and a truck with a maximum loading weight of 13 tons It will cost 13,000 yen per hour. Assume that the quantity of cargo shipped from each warehouse is equivalent to one 4-ton truck.

  When delivering from each warehouse to the base A, a plurality of routes can be considered. For example, the first route for delivering directly from the warehouse A, the warehouse B, and the warehouse C to the base A, the direct delivery from the warehouse C to the base A, and the second delivery from the warehouse B to the base A via the warehouse A Route, delivery from warehouse A directly to base A, delivery from warehouse B to warehouse A via warehouse C, and delivery from warehouse C to warehouse A via warehouse B and warehouse A A fourth route can be considered.

  When the first route is used, the route 501 that directly delivers from the warehouse A to the base A has a lead time of 8 hours, and is delivered in a total of 12 hours including the collection time and the shipping time. The route 502 for direct delivery from the warehouse B to the base A has a lead time of 10 hours and is delivered in a total of 14 hours including the collection time and the shipping time. The route 503 for direct delivery from the warehouse C to the base A has a lead time of 9 hours and is delivered in a total of 13 hours including the collection time and the shipping time. When a 4 ton truck is used for delivery from each warehouse to site A, the delivery from all warehouses is completed in 14 hours, and the total fare is 1890,000 yen.

  When the second route is used, the route 503 for direct delivery from the warehouse C to the base A has a lead time of 9 hours and is delivered in a total of 13 hours including the collection time and the shipping time. The route 504 for delivery from the warehouse B to the warehouse A has a lead time of 1 hour, and is loaded from the warehouse B to the warehouse A and delivered to the base A in a total of 15 hours. When a 4 ton truck is used for delivery from warehouse C to site A, the fare is 63,000 yen, and it is loaded from warehouse B to warehouse A and delivered to site A using a 10 ton truck. In this case, the fare is 90,000 yen. When the second route is used, delivery from all warehouses is completed in 15 hours, and the total fare is 153,000 yen.

  When the third route is used, the route 501 that delivers directly from the warehouse A to the base A has a lead time of 8 hours and is delivered in a total of 12 hours including the collection time and the shipping time. The route 505 for delivery from the warehouse B to the warehouse C has a lead time of 2 hours, and is loaded from the warehouse B to the warehouse C and delivered to the base A in a total of 17 hours. When a 4 ton truck is used for delivery from warehouse A to site A, the fare is 56,000 yen, and it is loaded from warehouse B to warehouse C and delivered to site A using a 10 ton truck. In this case, the fare is 110,000 yen. When the third route is used, the delivery from all warehouses is completed in 17 hours, and the total fare is 166,000 yen.

  In the case of using the fourth route, the route 506 for delivering from the warehouse C to the warehouse B has a lead time of 2 hours. Is done. When the cargo is loaded from warehouse C to warehouse B and warehouse A and delivered to base A using a 13-ton truck, the fare is 14.3 million yen. When the fourth route is used, the delivery from all warehouses is completed in 19 hours, and the fare is 14.3 million yen.

  In accordance with the planned order, closing time, lead time, working time and delivery date at each warehouse or base, etc., each process is based on the constraints obtained by the upper process and the initial unrelieved constraints of the lower process. A simulation is performed to relax the own constraints and create a solution that does not violate the constraints while negotiating between agents. Let this simulation result be the initial solution of the system. However, since the constraint relaxation of this process affects the upper process, it is necessary to improve the solution of the upper process again. For example, in the shipping process, on top of the solution obtained in the pickup process, based on the unrelaxed constraints of the delivery process, relax your own constraints and select a track that does not violate the constraints while negotiating between the tracks To do. However, when there is a truck with a low volume ratio, the constraint condition that has been dispatched for each base is relaxed so that it is dispatched according to the direction of delivery (delivery to a plurality of bases). If the reason why the truck volume ratio is not satisfied is due to the low pallet volume ratio, the pallet should be stacked according to the direction of distribution (delivered to multiple bases). To relax. However, in this case, it is not necessary to re-simulate so that pallets with a high volume ratio are stacked. In actual work, the system automatically determines and changes within the minimum range. Note that the user may specify a range to be simulated.

  Each warehouse determines the number of workers based on the simulation results. However, if the experience of the workers is different, an unpredictable problem may occur, and therefore the planned start time and completion time of each process may not be satisfied. Therefore, a certain time margin (time buffer) is set between the processes. The agent classifying unit 102 arrives at the first group satisfying the mandatory constraint condition and the relaxable constraint condition for the set of jobs that arrived before the completion time of the current work from the completion time of the current work to the start time of the next work. A set of jobs is classified into a second group that satisfies the mandatory constraint condition and the relaxation constraint condition, and a set of jobs that arrived at the next work start time is classified into a third group that does not satisfy at least one of the mandatory constraint condition and the relaxation constraint condition. .

  If the work of the upper process is not completed at the start time of the current process, the restriction agent recognition unit 103 recognizes that a bottleneck has occurred in the upper process and recognizes the upper process as a restriction agent (restriction process). To do. For example, when the work of the shipping process is not completed at the start time of the delivery process, the restriction agent recognition unit 103 recognizes that a bottleneck has occurred in the shipping process and recognizes the shipping process as a restriction agent.

  The constrained process affects the overall system throughput and must be improved. Therefore, the constraint agent improvement unit 104 makes maximum use of resources that can be preferentially used in the constraint process, and improves the constraint status. The agent subordinate unit 105 subordinates the work speed of the non-constrained processes other than the constrained process to the work speed of the constrained process. For example, if the dispatch of trucks is not yet completed at the start time of the delivery process, the shipping process speeds up the work by adding overtime and adding workers. Other processes change the work plan according to the situation of the constraint process.

  Next, determine negotiation criteria for coordination, concessions and synchronization between processes. There are three possible negotiating criteria: satisfying relaxed constraints, improving the constraint process, and ensuring that the overall system situation is not compromised. For example, when the completion time of the shipping process is exceeded by 1 hour, if there is a margin of 1 hour in the start time of the shipping process, this time can be transferred to the shipping process, thereby eliminating constraint violations in the shipping process. In addition, since the number of workers in the shipping process is insufficient, if it can be predicted that the work will not be completed by the start time of the delivery process, the constraint violation is resolved by handing over the buffer of the upper pickup process to the shipping process. However, when the pickup process also exists in the third group, this transfer movement cannot be performed.

  Then, the constraint relaxation part 107 cannot resolve the constraint violation even if negotiations by cooperation and concession of each process are performed, and if the relaxation parameter of each relaxable constraint does not reach the maximum value, the constraint relaxation is performed. ease. For example, it becomes easier to work by setting a low relaxation parameter for the volume ratio of the pallet or truck. In addition, the relaxation parameters relating to the start time and completion time are changed, and the size of the time buffer of each process is adjusted. In other words, the constraint violation situation can be resolved by setting the time buffer of the process that becomes a constraint longer and setting the time buffer of the process that completes work the shortest. Furthermore, by relaxing the parameters related to the volume ratio of the truck and changing the truck scheduled to be loaded in another warehouse to the direct delivery, it is possible to eliminate the situation of the constraint violation.

  As described above, at least a part of the multi-agent system is configured by heterogeneous agents having at least one of components, constraint conditions, and tasks, and the constraint agent improvement unit 104 maintains a dependency relationship between adjacent heterogeneous agents. While using the drum buffer rope, the constraint agent and the other agents are allowed to negotiate to satisfy the constraint condition. Therefore, it is possible to recognize a bottleneck that is a restriction in a part of a multi-agent system composed of heterogeneous agents more quickly, and it is possible to eliminate the bottleneck and improve the system activity.

  In the present embodiment, the user may specify the simulation time and range of the system using the input device 1. The determination unit 106 sets the number of times to repeat the simulation according to the designated time and range. Further, the user may set the maximum value and the minimum value of the relaxation parameter using the input device 1. In addition, when the maximum value and the minimum value of the relaxation parameter set by the user exceed a preset threshold value, the display device 6 may present an error or danger to the user. Further, by accepting the input of the work speed of the worker by the input device 1, the start time and the completion time of the actual work are predicted, and the processes are classified into the first to third groups according to the prediction result. May be. Furthermore, when a constraint violation occurs by accepting an input of an actual start time and completion time in each process, the display device 6 may present it to the worker.

  In this case, a plurality of relaxation parameters are set and the worker can adjust the relaxation parameters according to the actual work situation. Simulation results can be presented.

It is a figure for demonstrating MAS (multi agent system). It is a figure which shows an example of the hardware constitutions of the production management apparatus by one embodiment of this invention. It is a block diagram which shows an example of the main functions of the production management apparatus shown in FIG. It is a flowchart for demonstrating the production management process of the production management apparatus shown in FIG. It is a figure for demonstrating a production management process concretely for the example of the loading to the pallet of the load by the homogeneous agent. It is a figure for demonstrating the pillar produced when collecting a product. It is a figure for demonstrating improvement of a restriction | limiting agent for the example of the loading to the pallet of the load by a homogeneous agent. It is a figure which shows an example of the change of the number of agents which belong to the 1st thru | or 3rd group. It is a figure which shows an example of the number of agents classified into the check buffer. It is a figure for demonstrating the production management system in the heterogeneous agent of a distribution system. It is a figure for demonstrating regarding the determination of a delivery route (path | route).

DESCRIPTION OF SYMBOLS 100 Control part 101 Restriction condition setting part 102 Agent classification part 103 Restriction agent recognition part 104 Restriction agent improvement part 105 Agent subordinate part 106 Discriminating part 107 Restriction condition relaxation part 200 Storage part 201 Restriction condition storage part 202 Agent classification information storage part 300 Input Part 301 Constraint input accepting part

Claims (5)

Creating means for creating, for each pallet, the loading information indicating the loading state in which the product in the loading of the product on the pallet is loaded on the pallet;
As a constraint imposed on the product loading process in the pallet, an accepting unit that accepts a setting by a user of an essential constraint that must be satisfied and a relaxable constraint that can be relaxed;
Constraint storage means for storing the essential constraint condition and the relaxable constraint condition received by the reception means;
A relaxation means for creating a relaxation constraint that relaxes the relaxable constraint stored in the constraint storage;
The loading information satisfying the essential constraint condition and the relaxable constraint condition is set to the first group, and the loading information satisfying the essential constraint condition and the relaxation constraint condition and not satisfying the relaxable constraint condition before the relaxation is set to the second group. Classifying means for classifying product information that does not satisfy at least one of the essential constraint condition and the relaxation constraint condition into a third group;
Said stowage data, the first group, and the classification information storage means with the product information of the second group and third group is stored as a classification information associates the classified group by said classifying means,
With respect to the loading information classified into the third group by the classification means, the achievement degree of the essential constraint condition and the relaxation constraint condition to be achieved by the loading information is evaluated, and the achievement degree becomes smaller as the value becomes smaller. And an evaluation value for each essential constraint condition and each relaxation constraint condition is calculated on the basis of the evaluation function determined for each essential constraint condition and each relaxation constraint condition, and for each calculated essential constraint condition And the total evaluation value is calculated by adding all the evaluation values for each of the relaxation constraint conditions, and the loading information that maximizes the calculated total evaluation value is stored in the classification information storage unit . Extracting means for extracting as constrained product information from product information classified into groups,
The relaxation device further relaxes a relaxation constraint condition so that the constraint product information extracted by the extraction unit satisfies the essential constraint condition and the relaxation constraint condition. The relaxation means is configured to exchange the product included in the constraint product information and the product included in the other product information between the constraint product information and other product information. Executes an exchange process for satisfying the essential constraint condition and the relaxation constraint condition, and at least the constraint loading information satisfies the essential constraint condition and the relaxation constraint condition and the constraint violation loading information is reduced. Establishing the exchange process ,
The classification means reclassifies each product information after the exchange processing into the first to third groups,
The management device according to claim 1, wherein the relaxation unit further relaxes the relaxation constraint condition when there is product information belonging to the third group after reclassification. If the exchange process is not established, the extraction means adds new constrained product information other than the constrained product information for which the negotiation has not been established from the third group product information . Extracted as information ,
The relaxation means that to execute the replacement process of the new constraints stowage information for the required constraints and relaxed constraint condition is satisfied between the new constraint stowage information and other stowage data again The management apparatus according to claim 2, wherein: A step of creating a loading information for each pallet indicating a loading state in which the product is loaded on the pallet in the loading of the product onto the pallet;
As a constraint condition imposed on the product loading process in the pallet by the computer, a reception step for accepting a setting by a user of an essential constraint condition that must be satisfied and a relaxable constraint condition that can be relaxed;
A storage step in which the computer stores the essential constraint condition and the relaxable constraint condition received in the reception step in a constraint condition storage unit;
A relaxation step in which the computer creates a relaxation constraint that relaxes the relaxable constraint stored in the constraint storage;
The computer stores the loading information satisfying the essential constraint condition and the relaxable constraint condition in the first group, the loading information satisfying the essential constraint condition and the relaxation constraint condition and not satisfying the relaxable constraint condition before relaxation. Classifying the product information into the second group and the product information that does not satisfy at least one of the essential constraint condition and the relaxation constraint condition into the third group;
The computer, and the stowage data, the first group, the classification information storage means as the classification information attached the product information in association with groups classified in the classification step of the second and third groups A classification information storing step to store ;
The computer evaluates the achievement degree of the essential constraint condition and the relaxation constraint condition that the product information should achieve for the product information classified into the third group in the classification step, and the value is small. As the degree of achievement increases, and based on the evaluation function defined for each essential constraint and each relaxation constraint, the evaluation value for each essential constraint and each relaxation constraint is calculated and calculated. A total evaluation value is calculated by adding all the evaluation values for each essential constraint condition and for each relaxation constraint condition, and product information that maximizes the calculated total evaluation value is stored in the classification information storage unit. Extracting from the product information classified into the third group as constraint product information ,
The management method, wherein the relaxation step further relaxes a relaxation constraint condition so that the constraint product information extracted in the extraction step satisfies the essential constraint condition and the relaxation constraint condition. Creating means for creating, for each pallet, the loading information indicating the loading state in which the product in the loading of the product on the pallet is loaded on the pallet;
As a constraint imposed on the product loading process in the pallet, an accepting unit that accepts a setting by a user of an essential constraint that must be satisfied and a relaxable constraint that can be relaxed;
Constraint storage means for storing the essential constraint condition and the relaxable constraint condition received by the reception means;
A relaxation means for creating a relaxation constraint that relaxes the relaxable constraint stored in the constraint storage;
The loading information satisfying the essential constraint condition and the relaxable constraint condition is set to the first group, and the loading information satisfying the essential constraint condition and the relaxation constraint condition and not satisfying the relaxable constraint condition before the relaxation is set to the second group. Classifying means for classifying product information that does not satisfy at least one of the essential constraint condition and the relaxation constraint condition into a third group;
Said stowage data, the first group, and the classification information storage means with the product information of the second group and third group is stored as a classification information associates the classified group by said classifying means,
With respect to the loading information classified into the third group by the classification means, the achievement degree of the essential constraint condition and the relaxation constraint condition to be achieved by the loading information is evaluated, and the achievement degree becomes smaller as the value becomes smaller. And an evaluation value for each essential constraint condition and each relaxation constraint condition is calculated on the basis of the evaluation function determined for each essential constraint condition and each relaxation constraint condition, and for each calculated essential constraint condition And the total evaluation value is calculated by adding all the evaluation values for each of the relaxation constraint conditions, and the loading information that maximizes the calculated total evaluation value is stored in the classification information storage unit . Let the computer function as an extraction means to extract as constrained product information from the product information classified into groups,
The relaxation means further relaxes a relaxation constraint condition so that the constraint product information extracted by the extraction means satisfies the essential constraint condition and the relaxation constraint condition.

Images