JPH10111804A - Device and method for coordinative inference - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to adopt a proper negotiation strategy for tackling a complicated problem, easily settle negotiations, and improve the efficiency of problem solution by autonomously changing the negotiation strategy of each problem solving means into a proper strategy according to a negotiation history, an internal state, and an external state. SOLUTION: Each problem solving means is equipped with a communication part 21 which exchanges messages with other problem solving means, an inference part 22 which solves a given scheduling problem, an inference result storage part 23 which stores a generated schedule, and a state storage part 24 which stores a negotiation history, a current internal state, and a current external state. A strategy selection part 26 selects a negotiation strategy according to information stored in a data base 25 stored with negotiation strategies and the state storage part 24 and indicates it to the inference part 22. Then negotiations are settled by making a change into a proper negotiation strategy with a given problem to improve the efficiency of problem solution of the coordinate inference decision eventually.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a cooperative inference apparatus and a cooperative inference method in which a plurality of agents cooperate to solve a problem.

[0002]

2. Description of the Related Art With the recent advance in computer network technology, a distributed problem solving method for performing processing using a plurality of inference units has been proposed. In this distributed problem solving method, each inference unit independently sets a goal for problem solving and infers a possible processing procedure. In this case, the result of your inference is
Inconsistent results with the inference results of other inference units, or the problem cannot be solved only by the results of one's own inference.Therefore, exchange information with other inference units to correct the processing procedure, May be configured to perform consistent processing. This information exchange procedure is called negotiation. To date, negotiation protocols such as a contract net protocol and an integrated negotiation protocol have been proposed.

The contract net protocol simulates the market principle and defines a procedure for exchanging information for task assignment. Here, tasks that occur in the course of problem solving and cannot be solved by the inference unit alone or tasks that are more efficient when requesting another inference unit to solve the problem are referred to as tasks. When the inference unit (manager) holding the task announces the task, the other inference units (contractors) infer whether or not to make a contract to execute the task according to each situation, and determine that the contract should be made. If inferred, bid.

[0004] When an inference unit announces a task, it can be a manager. Bidding is performed by attaching the cost of the bidding inference unit to the task. This cost is, for example, the current load or the time required for processing the task. The inference unit that has made the announcement compares the conditions for each bid and wins the task for the inference unit that has presented the optimal conditions. For example, when several units cooperate to schedule a job, a job that cannot be arranged in its own timetable may be transferred as a task to another inference unit.

For the contract net protocol, RGSm
ith's paper, `` The Contract Net Protocol: High-Lebel
Communication and Control in a distributed proble
m Solver (IEEE transactions on Computers, Vol.29, N
o.12, pp.1104-1113, 1980) ”.

[0006] On the other hand, the integrated negotiation protocol works for each inference unit to maximize the utility of its own action plan, thereby establishing a consensus on race conditions and possible coordination between the inference units. It defines the method of finding.

Here, each inference unit creates a plan based on its own convenience, and obtains a joint plan obtained by combining the plans. Further, the cost is calculated for each plan in the joint plan, and a set of plans (negotiation set) in which the utility after subtracting the cost is the most positive for each inference unit is obtained.

[0008] Competition / compromise / cooperation situation is defined by this negotiation set, and each unit adopts the most beneficial conclusion plan as a plan. In other words, “cooperation” refers to a case where both parties have a welcomed conclusion. "Compromise" is when there is a sub-optimal, but more current, solution. Also,
"Competition" refers to the case where no negotiations are concluded due to no conclusions. In this case, by selecting an established choice, etc.,
Advance negotiations.

[0009] Under such a premise, the inference unit presents each of the inferred conclusions from the one with the highest utility, and agrees when the agreement proposed by the other party has a greater utility than its own.

[0010] For the contract negotiation protocol, see G. Zlotki.
n and JSRosenschein's paper `` Cooperation and Co
nflict Resolution via Negotiation Among Autonomous
Agents in Non-Cooperative Domain (IEEE Transaction
s on Systems, Man, and Cybernetics, Vol. 21, No. 6,
pp.1317-1324, 1991) ”.

[0011]

In the above-mentioned prior art, it is expected that consistency between units is realized by making various proposals between units. The inference was proceeding on the assumption that the latest inference state established in the negotiations was an implicit premise.

However, depending on the result of the negotiations with other units, there are actually a plurality of assumptions that may occur, or the existing negotiation contents that have been assumed for the inference may be overturned. In such a case, each unit had to proceed with the inference while taking into account a plurality of possibilities or redo the inference whose assumption was reversed. As a result, it has been difficult to improve the efficiency of the inference.

Particularly, in such conventional inference, when various options in the middle of the inference process are considered to be branches of a tree diagram, a depth-first search is performed. In addition, every time a branch close to the root is overturned, a great deal of inference is required.

It is also conceivable to make inferences based on the negotiation contents slightly earlier than the latest, but it is difficult to set a standard for judging at what point the negotiation contents are appropriate. is there. Furthermore, when the inference is at a standstill, it is difficult to distinguish between a mere factual proposal or process and a definitive agreement. Not clear and difficult to process.

The present invention has been proposed to solve the above-mentioned conventional problems, and an object of the present invention is to provide a cooperative inference apparatus and a cooperative inference method capable of performing inference based on efficient negotiation. is there.

[0016]

According to a first aspect of the present invention, there is provided a cooperative inference apparatus in which a plurality of problem solving means solve a predetermined problem by performing cooperative inference. First storage means for storing a first rule for obtaining the inference result of the above, second storage means for storing a second rule for negotiating with another problem solving means, Third storage means for storing predetermined information obtained in the course of negotiations with another problem solving means, and selection for selecting a predetermined negotiation strategy based on the predetermined information and the second rule A cooperative inference apparatus comprising: means; and inference means for performing a predetermined inference based on the first rule and the predetermined negotiation strategy.

According to a second aspect of the present invention, there is provided a cooperative inference apparatus in which a plurality of problem solving means solve a predetermined problem by performing cooperative inference, wherein one problem solving means is another problem solving means. Negotiation is made between a first communication unit for communicating with the other, a first storage unit storing a first rule for obtaining a predetermined inference result, and the other problem solving unit. A second rule for storing a second rule, a third memory for storing predetermined information obtained in the course of negotiations with the other problem solving means, Selecting means for selecting a predetermined negotiation strategy based on the second rule; and inference means for performing a predetermined inference based on the first rule and, if necessary, the predetermined negotiation strategy, The other problem solving means is the one problem solving device. Second for communication with the stages
Based on the content obtained from the second communication means and whether or not a predetermined target can be achieved, and sending the inference result to the second communication means. A cooperative inference apparatus, comprising:

Further, the invention according to claim 3 is characterized in that the third storage means stores, as predetermined information, the number of times a request for processing to another problem solving means has been rejected. It is a type inference device.

According to a fourth aspect of the present invention, the second storage means stores information between the problem processing of the one problem solving means itself and the problem processing sent from the other problem solving means. A cooperative inference apparatus that stores the second rule having a priority and a margin for a request value when the one problem solving unit requests the other problem solving unit to solve a problem.

On the other hand, the invention described in claim 5 is described from the viewpoint of a method. In a cooperative inference method in which a plurality of problem solving means solve a predetermined problem by performing cooperative inference, one of the inventions comprises: The problem solving means includes a first storage means for storing a first rule for obtaining a predetermined inference result and a second storage means for storing a second rule for negotiating with another problem solving means. Storage means, and a third storage means for storing predetermined information obtained in the course of negotiations with the other problem solving means, based on the predetermined information and the second rule, A cooperative inference method comprising: selecting a predetermined negotiation strategy; and performing an inference step of performing a predetermined inference based on the first rule and the predetermined negotiation strategy.

According to a sixth aspect of the present invention, there is provided a cooperative inference method in which a plurality of problem solving means solve a predetermined problem by performing a cooperative inference.
A first communication unit for communicating with another problem solving unit, a first storage unit for storing a first rule for obtaining a predetermined inference result, and the other problem solving unit. A second storage unit for storing a second rule for negotiating between the first and second storage units, and a third storage unit for storing predetermined information obtained in a process of the negotiation with the other problem solving unit. The other problem solving means comprises: a second communication means for communicating with the one problem solving means; and a predetermined target based on the content obtained from the second communication means. Second inference means for inferring whether or not achievement can be achieved, and sending the inference result to the second communication means, wherein the one problem solving means comprises the predetermined information and the second information. A selection step of selecting a predetermined negotiation strategy based on the rule of And a first inference step of performing a predetermined inference based on the predetermined negotiation strategy, and a case in which the first inference step infers that a predetermined negotiation is necessary for another problem solving means. Performing negotiations with the other problem solving means via the first communication means, wherein the other problem solving means responds to the negotiation from the one problem solving means by Infer whether or not to negotiate the negotiation using a second inference means,
Sending the inference result to the one problem solving means; and performing the step of storing the inference result from the other problem solving means in the third storage means. It is a cooperative inference method characterized by

Further, the invention according to claim 7 is characterized in that the third storage means stores, as predetermined information, the number of times a request for processing to another problem solving means has been rejected. This is a type inference method.

In the invention described in claim 8, the second storage means stores the problem between the problem processing of the one problem solving means itself and the problem processing sent from the other problem solving means. Storing the second rule having a priority of processing and a margin with respect to a request value when the one problem solving means requests the other problem solving means to solve the problem. It is an inference method.

According to a ninth aspect of the present invention, there is provided a storage medium storing a program for executing a cooperative inference for solving a predetermined problem by a plurality of problem solving means performing a cooperative inference. A predetermined rule for negotiating with the problem solving means, and means for selecting a predetermined negotiation strategy based on predetermined information obtained in the course of negotiation with the other problem solving means. A program for executing a predetermined inference based on a predetermined rule for obtaining a predetermined inference result and a predetermined negotiation strategy.

According to a tenth aspect of the present invention, there is provided a storage medium storing a program for executing a cooperative inference for solving a predetermined problem by a plurality of problem solving means performing cooperative inference. Based on the second rule for negotiating with the problem solving means and the predetermined information obtained in the course of the negotiation with the other problem solving means, a predetermined negotiation is provided to one of the problem solving means. Means for selecting a strategy; means for causing the one problem solving means to perform a first inference based on a first rule for obtaining a predetermined inference result and the predetermined negotiation strategy; The inference step of
Means for causing the one problem solving means to negotiate with the other problem solving means when it is inferred that predetermined negotiation is necessary for the other problem solving means; In response to the negotiation from the solving means, the other problem solving means infers whether or not to establish the negotiation, and means for sending the inference result to the one problem solving means, And means for causing the one problem solving means to obtain the inference result of (1), and a storage medium storing a program for executing the means.

[0026]

Embodiments of the present invention will be described below with reference to the drawings. In the present embodiment, roughly,
Each problem solving means has two viewpoints, that is, a priority (role) between the problem solving means, and a margin (degree of request) in a transaction between the own problem solving means and another problem solving means which is a negotiating partner. ), And autonomously selects and changes the negotiation strategy according to the negotiation history between itself and the negotiation partner, and the current internal state and external state at the time of negotiation.

In this embodiment, a scheduling system for drafting a production plan and a cooperative inference method realized on this system will be described. The scheduling system, based on the pre-input resource information,
This is a system for allocating a given job to resources. Here, the resource information is information on various resources used for processing the job, and is, for example, data such as “there are two milling machines and one drilling machine”, and “two workers can be secured”. Also, a job is a manufacturing process to be processed such as “first process for manufacturing the product A: shaving the surface (man-hours 2 hours)”, “second process: making a hole with a drilling machine (man hours 1 hour)”. And so on. The scheduling system determines which process is to be manufactured using which resource from when to while considering the open state of the resource, and solves the problem for the purpose of finally allocating all jobs to the resource.

By the way, a group of resources is called a shop, but a production plan is not completed for each job, and it is necessary to go through a plurality of shops in a certain order depending on the job. As described above, in the production plan, there are various constraint relationships such as an order relationship between shops and jobs. For this reason, it is necessary to plan the time process while considering the delivery date of the shop responsible for all processes,
Scheduling is performed between the inference units that are in charge of a plurality of shops by using an intersystem. In addition,
In the above example, a “milling machine”, a “drilling machine”, etc. are shops.

In such a scheduling system, for example, the following scheduling information is provided for manufacturing the product A. The same applies to the products B and C. First step: Lathe man-hour: 2h Earliest start time: 0 Latest end time: 20 Second step: Milling machine man-hour: 4h Earliest start time: 20 Latest end time: 60 Third step: Drilling machine Man-hour: 2h Earliest start time: 60 Latest end time: 80 In the present embodiment, based on such scheduling information, schedulers (corresponding to problem solving means) 1a, 1b, 1c corresponding to each shop are required in each work plan table. Time (man-hours). However, each scheduler must perform arrangement within the given earliest start time and latest end time. Also, the work in the second step must be scheduled after the scheduled end time of the work in the first step. Further, the scheduled work end time in the final process cannot be shifted after the latest end time (delivery date).

The present embodiment will be described on the assumption of the above-described scheduling problem. FIG. 1 is a diagram illustrating a conceptual configuration of a cooperative inference apparatus according to the present embodiment.
As shown in the figure, the present embodiment has the problem solving means 1a,
1b, 1c, and the respective problem solving means 1a, 1b, 1c are connected to each other using the communication path 2. In such a problem solving means, each processor may be independently configured, or a single processor may be configured to logically operate a plurality of processes in parallel.

In the present embodiment, the problem solving means 1a, 1b, 1c are schedulers which are responsible for arranging jobs on a lathe, milling machine and drilling machine, respectively.
FIG. 2 shows problem solving means 1a, 1b,
It is a figure which shows the structure of 1c. As shown in the figure, each problem solving means 1a, 1b, 1c is composed of a communication unit 21 for exchanging messages with other problem solving means, an inference unit 22 for solving a given scheduling problem. Inference result storage unit 23 for storing schedules, state storage unit 2 for storing negotiation history, current internal state, and external state
4. A strategy selection unit 26 that selects a negotiation strategy based on the information stored in the strategy database 25 storing a plurality of negotiation strategies and the state storage unit 24 and instructs the inference unit 22.
Consists of The inference unit 22 includes an inference rule 221 and an inference engine 222.

The strategy selector 26 has a strategy rule for solving the above-mentioned scheduling problem. Hereinafter, in the present embodiment, the order between the problem solving means used to assign a processing priority between a problem occurring in the own problem solving means and a problem requested from another problem solving means. The viewpoint is called "role". In addition, the viewpoint of the size (degree) of a request in a transaction between the own problem solving means and another problem solving means to be a negotiation partner is referred to as a “transaction”.

In the present embodiment, the "roles" are classified into three, namely, upper / equal / lower, and each problem solving means assumes one of the "roles". The order of the problem solving means related to the processing priority is set in the order of higher order> equal order> lower order. In this case, the problem solving means positioned as equal gives priority to the problem generated by its own problem solving means over the problem requested by the lower problem solving means, but the problem requested by the higher problem solving means is It shall take precedence over the problems that have arisen in its own problem solving means.

In addition, "transactions" are classified into two, that is, margin / minimum, and each problem solving means performs any one of "transactions" at the time of negotiation. The size of the request in the transaction between its own problem solving means and the negotiation partner is set as margin> minimum.
In this case, the problem solving means for conducting the marginal transaction requests the negotiating partner a request obtained by adding a margin to the required minimum requirement. Conversely, when a request is received from a negotiating partner, if the request is processed, it is determined whether or not the processing can be performed in consideration of whether or not it is possible to secure its own margin. Rule 1: When negotiations are started (initial state) then “Role” = equal, “Trading” = Margin Rule 2: In negotiations, receive a rejection response from the negotiating partner in response to a request issued by the user, and When a rejection response is received from the negotiating partner even if the request has been re-issued again (when two or more rejecting responses are received from one negotiating partner) then "role" = higher, "transaction" = minimum rule 3: In the case of if negotiation, if the request from the negotiating partner is rejected, and the request issued by the negotiating partner is rejected again (two or more times When a rejection is answered) then "role" = lower, "transaction" = minimum In this strategy selection, in the initial state, each scheduler negotiates as a "role" on an equal footing, and "transaction"
In both cases where a request is issued and a request is received, it is determined whether or not the request can be created or accepted.

However, when receiving two or more rejected replies from one negotiation partner, the "role" is ranked higher, and the negotiation is advanced so that the user's own arrangement is prioritized. In addition, as to “transactions”, whether a request is issued or received, a request is made or the request is accepted or not, and whether or not the request is accepted is determined.

Conversely, if one or more replies are rejected to one negotiation partner, the "role" will be lower, and negotiations will proceed with priority given to the arrangement of the other partners. Also,
Regarding the “transaction”, it is determined whether the request can be created or the request can be accepted within the minimum range when the request is issued or received.

Next, a negotiation strategy in the strategy selecting section 26 will be described. FIG. 3 is a diagram showing a state in which each negotiation strategy is classified. “Role” = equivalent: When a partition change request (shifting the earliest start time or latest end time indicating the partition time between each process forward or backward) is received, it is interpreted whether the request can be interpreted and executed. Judge and answer the result. Even if it is not feasible, there is no new request to make it possible.

When a job placement failure has occurred in its own problem solving means, a request for enabling the given job to be arranged is created, and the request is transmitted to the relevant problem solving means.

If the response to the request issued by the user is rejected, a different partition change request is made or the same partition change request is made again after a while. “Role” = higher: When a partition change request is received, a rejection reply is made to the request. Requests are not accepted.

When a job placement failure has occurred in its own problem solving means, a request is made to allow the given job to be arranged, and the request is transmitted to the relevant problem solving means.

When the response to the request issued by the user is rejected, a different partition change request is made or the same partition change request is made again after a while. “Role” = Lower order: When a partition change request is received, the contents of the request are interpreted and it is determined whether the request can be executed. If it is determined that the request can be processed, acceptance is answered to the request.
If it is determined that the request cannot be processed, the user himself / herself issues a new request to enable the request.

When a failure in job placement has occurred in the problem solving means, a request for enabling a given job to be placed is created, and the request is transmitted to the corresponding scheduler. If the response to the request issued by the user is rejected, a different partition change request is made or the same partition change request is made again after a while. "Transaction" = margin: When a partition change request is received and it is determined whether or not execution is possible, it is determined that execution is not possible if a margin of n hours cannot be taken by itself after changing the partition time. .

When a partition change request is created, the request is created after shifting the partition time by at least the necessary amount, and adding a margin for n hours. “Transaction” = minimum: When a partition change request is received and it is determined whether or not it can be executed, if it is possible to change the partition time by the requested amount, even if there is no margin on the part of the user, Judge as executable.

When a partition change request is created, the request is created without adding any margin after shifting the partition time by at least the necessary time. In this embodiment, all the problem solving means have the same strategy selecting unit and negotiation strategy. However, the present invention is not limited to this embodiment, and each problem solving means may have its own strategy selecting unit and negotiation strategy.

The cooperative inference apparatus configured as described above solves the scheduling problem by operating as follows. Here, the manufacturing procedure of the product A and the product B is shown for each of the problem solving means 1a, 1b, and 1c in charge of the lathe, the milling machine, and the drilling machine.
It is assumed that the following scheduling information is input. "Product A" 1st process: Lathe man-hour: 3 Earliest start time: 0 Latest end time: 5 Second process: Milling machine man-hour: 3 Earliest start time: 5 Latest end time: 10 "Product B" 1st process: Lathe man-hours: 2 Earliest start time: 0 Latest end time: 5 Second step: Milling machine man-hours: 3 Earliest start time: 5 Latest end time: 10 Third step: Drilling machine Man-hours: 2 Earliest start time: 10 Latest end Time: 15 The problem solving means 1a, 1b, 1c secures the required number of hours (man-hours) in each work plan table based on the input scheduling information of the products A and B.

For example, the problem solving means 1b includes the inference unit 22
, The job (man-hours 3) for the product A is 5 to 10
Between the two. Here, it is assumed that the jobs are arranged at the front and arranged between 5 and 8. Similarly, the other problem solving means also arranges a given job using each inference unit and creates a work plan table as shown in FIG.

Each of the problem solving means 1a, 1b, and 1c creates a work plan table. If there is a collision of job arrangements during the creation process (within the dotted line in the figure), the problem solving means is replaced with another one. Negotiate a solution to the problem.

FIG. 5 is a diagram for explaining the processing operation of the problem solving means at the time of negotiation. As shown in the figure,
When it is determined that negotiation is necessary in one problem solving means (STEP 501), the problem solving means creates a partition change request (STEP 502). Next, the problem solving means selects one of the other problem solving means as the negotiation partner, and transmits the created partition change request to the selected other problem solving means (STEP 503).

On the other hand, the selected other problem solving means determines whether or not the received partition change request is executable (STEP 504).
Is returned as a message if it is impossible (STEP 505).

One problem solving means receives the returned request result (STEP 506), and determines again whether negotiation is necessary (STEP 501). One problem solving means thereafter repeats the above-described processing as much as possible.

The negotiation process as described above will be described with a specific example. That is, in the work plan table shown in FIG. 4, the problem solving means 1a and the problem solving means 1c can arrange all the given jobs, but the problem solving means 1b simultaneously arranges the job arrangement of the product A and the product B. You cannot do it. Therefore, the problem solving means 1b negotiates with another problem solving means in order to enable this job arrangement.

FIGS. 6 and 7 are diagrams for explaining the progress of the negotiation between the respective problem solving means. First, the problem solving means 1b asks the problem solving means 1a for a partition change request. Specifically, in the initial state, the strategy selecting unit 26 of the problem solving means 1b sets “role” = equal, “transaction”
A negotiation strategy corresponding to == room is acquired from the strategy database 25 and instructed to the inference unit 22. Therefore, after arranging the job of the product B in 7 to 10, the earliest start time is changed to 4 in order to enable the job arrangement of the product A, and 1 is added as a margin to set the earliest start time to 3. A partition change request is created and transmitted to the problem solving means 1a via the communication unit 21 (1R in FIG. 7).

In the initial state, the strategy selecting unit 26 of the problem solving means 1 a also acquires a negotiation strategy corresponding to “role” = equal and “transaction” = room from the strategy database 25 and instructs the inference unit 22. The inference unit 22 determines whether the latest end time of the job arrangement of the product A can be set to 3.
Infers. In the problem solving means 1a, since the job arrangement of the product A is arranged between 0 and 3, it is determined that it is possible. However, if the latest end time of the job arrangement of the product A is set to 3, the earliest start time 0 to the latest end time 3
, The man-hours 3 are assigned to, and it is not possible to take a sufficient amount of time. Therefore, problem solving means 1
a transmits the rejection of this request to the problem solving means 1b via the communication unit (FIG. 6A and FIG. 7A).
1).

Next, the problem solving means 1b asks the problem solving means 1c for a partition change request. Problem solving means 1b
Uses the same negotiation strategy as before, places the jobs of product A at 5 to 8 and then changes the latest end time to 11 to allow the job of product B to be placed. One is added as a minute, and the latest end time 12 is transmitted as a partition change request (R2 in FIG. 7).

In the initial state, the problem solving means 1c also adopts a negotiation strategy of “role” = equal and “transaction” = room. Therefore, the earliest start time of the job B of the product B is set to 1
It is inferred whether it is possible to make 2 and whether it can take a margin. By arranging the job (man-hours 2) of the product B between 12 and 15, the problem solving means 1c can be arranged including a margin. Therefore, the problem solving means 1c
Accepts this request, and a negotiation is established between the problem solving means 1b and the problem solving means 1c (A in FIG. 6B and FIG. 7).
2). As a result, the job arrangement of the product A and the product B is completed (FIG. 6C).

However, depending on a given scheduling problem, “role” = equal, “transaction” = room,
Negotiation may not be successful even if a negotiation strategy equivalent to is used. In such a case, it will be necessary to change to a more appropriate negotiation strategy.

Here, it is assumed that the following scheduling information representing the manufacturing procedure of the product C and the product D is input to each of the problem solving means 1a, 1b, and 1c in charge of the lathe, the milling machine, and the drilling machine. I do. "Product C" 1st process: Lathe man-hour: 3 Earliest start time: 0 Latest end time: 5 Second process: Milling machine man-hour: 3 Earliest start time: 5 Latest end time: 10 "Product D" 1st process: Lathe man-hours: 6 Earliest start time: 0 Latest end time: 5 Second step: Milling machine man-hours: 2 Earliest start time: 5 Latest end time: 10 Third step: Drilling machine Man-hours: 2 Earliest start time: 10 Latest end Time: 15 The problem solving means 1a, 1b, 1c secures the required number of hours (man-hours) in each work plan table based on the input scheduling information of the products C and D.

FIG. 8 shows each problem solving means 1a, 1b, 1c.
Is a diagram showing an example of a work plan table in which a given job is arranged by the inference unit 22 of FIG. In FIG.
b and the problem solving means 1c can arrange all the given jobs, but the problem solving means 1a cannot perform the job arrangement of the product C and the product D at the same time. ing. Therefore, the problem solving means 1a negotiates with another problem solving means to enable this job arrangement.

FIG. 9 and FIG. 10 are diagrams for explaining the progress of the negotiation between the problem solving means. First, the problem solving means 1a asks the problem solving means 1b for a partition change request (R1 in FIG. 10). That is, in the initial state, the strategy selecting unit 26 of the problem solving unit 1a
A negotiation strategy corresponding to “= equal,“ transaction ”= room” is acquired from the strategy database and instructed to the inference unit 22. Therefore, after arranging the job of the product C in 0 to 3, the latest end time is changed to 9 in order to enable the job arrangement of the product D, 1 is added as a margin, and the latest end time is set to 10 And sends the request to the problem solving means 1b using the communication unit 21.

In the initial state, the strategy selecting unit 26 of the problem solving unit 1b also acquires a negotiation strategy corresponding to “role” = equal and “transaction” = room from the strategy database, and instructs the inference unit 22. Therefore, it is determined using the inference unit 22 whether the earliest start time of the job arrangement of the product D can be set to 10 and whether or not a margin can be obtained. In the problem solving means 1b, since the latest end time of the job arrangement of the product D is set to 10, it is determined that it is impossible. Therefore, the problem solving means 1b transmits the rejection of this request to the problem solving means 1a using the communication section 21 (FIG. 9A and A1 in FIG. 10).

Next, the problem solving means 1a again sends a partition change request to the problem solving means 1b (R in FIG. 10).
2). That is, the problem solving means 1a uses the same negotiation strategy as above, arranges the jobs of the product D at 0 to 6, and then sets the latest end time to 9 to enable the job arrangement of the product C. Then, 1 is added as a margin, and the latest end time 12 is transmitted as a partition change request. The problem solving means 1b employs the same negotiation strategy as before. Therefore,
It is determined whether the earliest start time of the job arrangement of the product C can be set to 10 and whether or not a margin can be obtained.
In the problem solving means 1b, since the latest end time of the job arrangement of the product C is set to 10, it is determined that it is impossible. Therefore, the problem solving means 1b transmits the rejection of this request to the problem solving means 1a using the communication section 21 (FIG. 9 (b) and A2 in FIG. 10).

When the negotiation strategy is fixed as in the conventional negotiation model, the problem solving means 1a determines at this point that the partition cannot be changed, and the schedule of the product C or the product D becomes impossible. However, the problem solving means 1a according to the present invention requires the strategy selecting unit 2
Use 6 to make a change to a more appropriate negotiation strategy.

That is, the status storage unit 24 of the problem solving means 1a counts that both of the previous two partition change requests have been rejected. Strategy selector 2 of problem solving means 1a
6 inputs the information from the state storage unit 24, acquires a negotiation strategy corresponding to “role” = higher-order, and “transaction” = minimum as a more suitable negotiation strategy from the strategy database, and instructs the inference unit 22. . After that, the problem solving means 1a again asks the problem solving means 1b for a partition change request. Therefore, after arranging the job of the product C in 0 to 3, the product D
A partition change request is created so that the latest end time is changed to 9 as the minimum change required to enable the job arrangement, and transmitted to the problem solving means 1b via the communication unit 21 (R3 in FIG. 10). ).

The problem solving means 1b counts that the state storage unit 24 has rejected the partition change request twice in the past. The strategy selecting unit 26 of the problem solving means 1b inputs the information from the state storage unit 24 and acquires a negotiation strategy corresponding to “role” = lower order and “transaction” = minimum as a more appropriate negotiation strategy from the strategy database. Then, it instructs the inference unit 22. Therefore, the inference unit 22 determines whether the earliest start time of the job arrangement of the product D can be set to 9. In the problem solving means 1b, since the latest end time of the job arrangement (man-hours 2) of the product D is set to 10, it is determined that it is impossible. Therefore, the problem solving means 1b
Submits a new partition change request in order to accept this request. Therefore, the job of the product D is started at the earliest start time 9
A request to change the partition is made to change the latest end time to 11 in order to arrange it in accordance with the request, and the request is transmitted to the problem solving means 1c (R4 in FIG. 10).

The problem solving means 1c employs a negotiation strategy of “role” = equal and “transaction” = room in the initial state. Therefore, it is determined whether the earliest start time of the job arrangement of the product D can be set to 11 and whether or not a margin can be obtained. By arranging the job (man-hours 2) of the product D between 11 and 15, the problem solving means 1c can be arranged including a margin. Therefore, the problem solving means 1c
Accepts this request and responds to the problem solving means 1b that the request has been accepted. As a result, negotiation is established between the problem solving means 1b and the problem solving means 1c (FIG. 9 (d) and FIG. 10).
A3). Further, the problem solving means 1b can accept the request from the problem solving means 1a, and replies to that effect to the problem solving means 1a. Thus, the negotiation between the problem solving means 1a and the problem solving means 1b is also established (FIG. 9 (e) and A4 in FIG. 10). As a result, the job arrangement of the products C and D is completed (FIG. 9).
(F)).

As described above, by changing to a negotiation strategy suitable for a given problem, negotiation can be established,
As a result, the problem solving efficiency of the cooperative inference apparatus can be improved.

The present invention is not limited to the above embodiment. The field and contents of the inference problem to which the present invention is applied,
The number of problem solving means to cooperate and the method of classifying negotiation strategies are arbitrary. For example, the present invention may be applied to a failure diagnosis or plant control, five or more problem solving means may be provided, or a negotiation strategy may be classified based on the viewpoint of a property inherent to the problem solving means.

[0068]

As described in detail above, according to the present invention,
The negotiation strategy of each problem solving means is autonomously changed to an appropriate strategy according to the negotiation history, the internal state, and the external state. That way, when dealing with complex issues,
An appropriate negotiation strategy can be adopted, and negotiations can be easily completed. Further, as a result, it is possible to improve the problem solving efficiency of the cooperative inference apparatus.

[Brief description of the drawings]

FIG. 1 is a diagram showing a conceptual configuration of a cooperative inference apparatus according to an embodiment.

FIG. 2 shows problem solving means 1a and 1 according to the embodiment.
The figure which shows the structure of b, 1c.

FIG. 3 is a diagram showing a state in which each negotiation strategy is classified.

FIG. 4 is a diagram showing an example of a work plan table for products A and B according to the embodiment of the present invention.

FIG. 5 is a diagram for explaining the processing operation of the problem solving means at the time of negotiation.

FIG. 6 is a diagram for explaining the progress of negotiation between the problem solving means.

FIG. 7 is a diagram for explaining the progress of negotiation between the problem solving means.

FIG. 8 is a diagram showing an example of a work plan table for products C and D according to the embodiment of the present invention.

FIG. 9 is a diagram for explaining the progress of negotiation between the problem solving means.

FIG. 10 is a diagram for explaining the progress of negotiation between the problem solving means.

[Explanation of symbols]

 1a, 1b, 1c ... Problem solving means 2 ... Communication channel 21 ... Communication unit 22 ... Inference unit 221 ... Inference rule 222 ... Inference engine 23 ... Inference result storage unit 24 ... State storage unit 25 ... Strategy database 26 ... Strategy selection unit

Claims (10)

[Claims] 1. A cooperative inference apparatus in which a plurality of problem solving means solves a predetermined problem by performing cooperative inference, wherein one problem solving means stores a first rule for obtaining a predetermined inference result. A second storing means for storing a second rule for negotiating between the first storing means and another problem solving means; and a second storing means for storing a second rule for negotiating between the first problem solving means and the other problem solving means. Third storage means for storing predetermined information to be obtained, selection means for selecting a predetermined negotiation strategy based on the predetermined information and the second rule, and the first rule and the predetermined negotiation strategy. And a inference means for performing a predetermined inference based on the inference device. 2. A cooperative inference apparatus in which a plurality of problem solving means solves a predetermined problem by performing cooperative inference, wherein one problem solving means is a first for communicating with another problem solving means. Communication means, a first storage means for storing a first rule for obtaining a predetermined inference result, and a second storage means for storing a second rule for negotiating with the other problem solving means. A second storage unit, a third storage unit that stores predetermined information obtained in a process of negotiations with the other problem solving unit, and a predetermined storage unit based on the predetermined information and the second rule. Selecting means for selecting a negotiation strategy, and inference means for performing a predetermined inference based on the first rule and, if necessary, the predetermined negotiation strategy, wherein the other problem solving means comprises: To communicate with the Based on the content obtained from the second communication means, inferring whether or not a predetermined goal can be achieved, and transmitting the inference result to the second communication means. 2. A cooperative inference apparatus comprising: (2) inference means. 3. The cooperative inference apparatus according to claim 1, wherein said third storage means stores, as predetermined information, the number of times a request for processing to another problem solving means has been rejected. . 4. The second storage means includes: a priority between the problem processing of the one problem solving means itself and the problem processing sent from the other problem solving means; and the first problem solving means. 4. The cooperative inference apparatus according to claim 1, wherein the second rule stores a second rule having a margin for a request value when requesting the other problem solving means to solve the problem. 5. A cooperative inference method in which a plurality of problem solving means solve a predetermined problem by cooperative inference, wherein one problem solving means stores a first rule for obtaining a predetermined inference result. A second storing means for storing a second rule for negotiating between the first storing means and another problem solving means; and a second storing means for storing a second rule for negotiating between the first problem solving means and the other problem solving means. Selecting a predetermined negotiation strategy based on the predetermined information and the second rule; and a step of selecting a predetermined negotiation strategy based on the predetermined information and the second rule. A cooperative inference method, comprising: an inference step of performing a predetermined inference based on a strategy. 6. A cooperative inference method in which a plurality of problem solving means solve a predetermined problem by cooperative inference, wherein one problem solving means is a first for communicating with another problem solving means. Communication means, a first storage means for storing a first rule for obtaining a predetermined inference result, and a second storage means for storing a second rule for negotiating with the other problem solving means. A second storage means, and a third storage means for storing predetermined information obtained in the course of negotiations with the other problem solving means, wherein the other problem solving means comprises: Second communication means for communicating with the means, and inferring whether or not a predetermined goal can be achieved based on the content obtained from the second communication means; And second inference means for sending to the second communication means, One problem solving means includes: a selecting step of selecting a predetermined negotiation strategy based on the predetermined information and the second rule; and a predetermined inference based on the first rule and the predetermined negotiation strategy. A first inference step of performing the following, and when it is inferred by the first inference step that a predetermined negotiation is necessary for another problem solving means, the first inference step is performed via the first communication means. Performing a negotiation with another problem solving means. The other problem solving means establishes the negotiation with the second inference means with respect to the negotiation from the one problem solving means. Performing a step of inferring whether or not to do so and sending the inference result to the one problem solving means; the one problem solving means stores the inference result from the other problem solving means in the third storage Step of storing in means Cooperative reasoning method and performing. 7. The cooperative inference method according to claim 1, wherein said third storage means stores, as predetermined information, the number of times a request for processing to another problem solving means has been rejected. . 8. The second storage means includes: a priority of a problem process between a problem process of the one problem solving device itself and a problem process sent from the other problem solving device; 4. A cooperative inference method according to claim 1, wherein said second rule having a margin for a request value when said problem solving means requests said other problem solving means to solve a problem is stored. . 9. A storage medium storing a program for executing a cooperative inference for solving a predetermined problem by cooperative inference by a plurality of problem solving means, and negotiating with another problem solving means. Means for selecting a predetermined negotiation strategy on the basis of predetermined rules for negotiation and predetermined information obtained in the course of negotiations with the other problem solving means; and predetermined means for obtaining a predetermined inference result. A program for executing a predetermined inference based on the above rule and the predetermined negotiation strategy. 10. A storage medium storing a program for executing cooperative inference for solving a predetermined problem by cooperative inference by a plurality of problem solving means, and negotiating with other problem solving means. Means for causing one of the problem solving means to select a predetermined negotiation strategy, based on the second rule for determining and the predetermined information obtained in the course of the negotiation between the other problem solving means; Means for causing the one problem solving means to make a first inference on the basis of a first rule for obtaining a result and the predetermined negotiation strategy; If it is inferred that certain negotiations are necessary for the means,
Means for causing the one problem solving means to negotiate with the other problem solving means; and determining whether the other problem solving means should negotiate the negotiation from the one problem solving means. To infer
A storage medium storing a program for executing means for sending the inference result to the one problem solving means, and means for causing the one problem solving means to acquire the inference result from the other problem solving means.