JPS63291128A - Production system - Google Patents

Abstract

PURPOSE:To realize a high efficient production system by dispersing a large quantity of collating operation necessary for preparation of a competitive set into respective slave processors and executing it serially and in parallel. CONSTITUTION:The initial values of all rules and all working memories are given to a master processor, and all rules are equally assigned through a common bus to all slave processors. A master processor simultaneously gives the contents of the working memories through the common bus to all slave processors and the slave processor prepares the competitive set between the contents and the rules assigned to itself. The slave processor receives the information of the preceding competitive cancellation from itself through a ring bus. The master processor executes the rules based on the information of the result of the competitive cancellation from the preceding slave processor. Thus, the load of the system is dispersed and the inference is efficiently executed.

Description

[Detailed description of the invention] [Industrial application field] The present invention is an addition to a production system.

[Conventional technology]

A production system is a method for making computers process knowledge information.

Configuration of the production system in the case of conventional technology,
And its operation will be explained based on FIG.

The numbers in the figure schematically indicate the entire production system controlled by the control functions. Also, 22 is a working memory function that stores system control information, etc., and 23 is generally used to say "If... then execute...".
It represents the rule memory function expressed in the form. Further, 24 represents a competitive set generation function that creates a set of executable rules (hereinafter this set will be referred to as a competitive set) from the data in the working memory and the rules in the rule memory. 25 represents a conflict resolution function that selects the only rule to be executed from the conflict set.

26 represents a rule execution function that actually executes the selected rule, 27 represents a man-machine interface function, and 24, 25, 26, and 27 are collectively called an inference engine.

The system operates through what is commonly referred to as a "recognize-execute cycle." More specifically, this is a repetition of the following four actions.

1) The individual contents in the working memory (hereinafter referred to as WME) are compared with the rules, and a set of executable rule-WME pairs (hereinafter referred to as conflict set) is determined.

).

2) Rules and W that should be actually executed from the 1 gold set
Select a set of MEs. (Hereinafter, this operation will be referred to as conflict resolution.) 3) Execute the execution part of the rule selected in 2).

4) Return to 1).

This repetition is repeated until the contention set in 1) becomes empty or until there is an explicit stop command in 3).

Of these repetitions, the part that has the greatest impact on system efficiency is 1), which creates a competitive set. If no measures are taken, more than 80-90% of the total execution time will be consumed for this purpose.

The reason why this part consumes such a large amount of time is that it takes time to match the conditional clause of the rule with the WME.

Now, if we simply perform this matching, the total time T required for this is the number of rules, r1 the average number of condition parts in a rule, Cl the number of WMEs, and the average matching time for one ~vME and one conditional clause. If t is t, it is given by 'r=rxcxwxt.

Several speed-up techniques are used to shorten this time T.

For example, the RETE algorithm focuses on redundancy and similarity of conditional clauses and WMEs to reduce the number of matching operations.

However, as the system becomes more complex and larger, the effectiveness of these algorithms decreases in practical terms.

[Problem that the invention seeks to solve]

The problem that the present invention aims to solve is that in conventional production systems, a large amount of time is consumed in creating conflict sets, which reduces the execution efficiency of the production system. It operates more efficiently, is adaptable to systems of various sizes, is easily expandable, and is more compatible with microprocessor systems.

[Means for solving problems]

The present invention provides: A) At least a) Rule memory function b) Working memory function C) Conflict set generation function d) Conflict resolution function e) At least one set of output functions r) At least two sets of input functions g) Above a) to f B) and at least h) a rule memory function i) a working memory function j) a rule execution function k) at least one set of inputs; Function ■) At least two sets of output functions m) includes in its configuration a master processor with an operation control function that controls the functions h) to 1) above, and C) and at least one of the slave processor elements of A). and at least one set of output functions of the master processor element of B) are connected to a common bus; and D) and at least one set of input/output functions of the slave processor element of A); At least one set of input/output functions of the master processor of the i
This is a production system that is characterized by being connected to

〔Example〕

Figures 1 and 2 schematically show an embodiment of the present invention.
The present invention will be described in detail with reference to FIGS.

FIG. 1 is a conceptual diagram of a production system according to the present invention. In the figure, number 1 indicates a group of slave processors, 2 indicates a common bus, 3 indicates a master processor, and 4 indicates a ring bus.

FIG. 2 is a conceptual diagram of the functions of the slave processor. In the figure, number 5 is an input function section, 6 is an output function section, 7 is a rule memory function section, 8 is a working memory function section, 9 is a conflict set generation section, 10 is a conflict resolution function section, 11 is an input function section, 12 indicates the range of control functions of the slave processors.

FIG. 3 is a conceptual diagram of the functions of the master processor. In the figure, number 13 is the output function section, 14 is also the output function section,
15 is an input function section, 16 is a working memory function section, 1
7 is a rule memory function section, 18 is a rule execution function section, 1
9 indicates the output to the common bus, and 20 indicates the range of control functions of the mask processor.

Based on these figures, the operation of a typical production system according to an embodiment of the present invention will be explained in detail, step by step.

1.1f Give initial values of all rules and all arcing memories to the master processor.

2. The master processor allocates all its rules equally to all slave processors via a common bus. Here, being equal does not mean simply assigning the same number of rules to each, but also considering the complexity of the rules and assigning them so that the load is even. .

3. The mask processor provides the contents of the working memory to all slave processors simultaneously via a common bus.

4. The slave processor creates a conflict set between the given working memory contents and the rules assigned to it. What should be noted here is that the load of creating this conflict set does not depend on the size of the entire system, but only on the number of rules assigned to the slave processor. In other words, a huge system can be as efficient as a small system by increasing or decreasing the number of slave processors according to the size of the system.

5. Receive contention resolution information of the slave processor immediately before itself via the slave processor beam/gang bus.

6. The slave processor adds 5 to the contention set created in 4.
Conflict resolution is performed by adding the information received in .

7. The slave processor outputs the result of 6 to the ring bus.

If this slave processor is the one before the master processor on the ring bus,
The conflict resolution results output here are the conflict resolution results for all rules.

8. The master processor executes its rules based on the conflict resolution result information from the previous slave processor.

9. Based on the information in 8, the master processor rearranges the rules of the slave processor group, if necessary.

10. The master processor outputs working memory update information to the common bus based on the update information of the working memory based on the result of the execution of step 8.

11. The slave processor group is based on the information in 10,
Updates own conflict set if necessary.

12. Return to 6.

As exemplified above, in the production system according to the present invention, the rules and contents of the working memory are not statically arranged as in the conventional production system. Rather, by dynamically circulating them within the system, the system load is distributed and inference is executed efficiently.

〔Effect of the invention〕

As explained above, according to the present invention, the large amount of collation operations required to create a conflict set, which was a problem in conventional production systems, is distributed to each slave processor, and it can be performed serially or in parallel. executed. Therefore, a highly efficient production system can be realized.

Further, since the functions of each slave processor are all equivalent, they can be easily manufactured in large quantities using a microprocessor or the like.

Furthermore, the number of slave processors essentially has no effect on the operation of this system. Therefore, depending on the size of the target system, the capacity of the system can be increased very easily by increasing the number of slave processors. Furthermore, the capabilities of a completed system can be easily extended using the same method.

As described above, the proxy system according to the present invention is much more efficient, more expandable, and more compatible with hardware than conventional systems.

[Brief explanation of drawings]

FIG. 1 is a conceptual diagram of a production system according to the present invention. In the figure, number 1 indicates a group of slave processors, 2 common buses, 3 a master processor, and 4 a ring bus. FIG. 2 is a conceptual diagram of the functions of the slave processor. In the figure, number 5 is the input function section, 6 is the output function section, 7 is the rule memory function section, 8 is the working memory function section, 9 is the conflict set generation section, 10 is the conflict resolution section, and 11 is the conflict resolution section. Reference numeral 12 indicates an input function section, and 12 indicates the range of control functions of the slave processor. FIG. 3 is a conceptual diagram of the functions of the master processor. In the figure, number 13 indicates the output function section, 14 also indicates the output function section,
15 is an input function section, 16 is a working memory function section, 17 is a rule memory function section, 18 is a rule execution function section, 19 is an output to a common bus, and 20 is a range of control functions of the master processor. show. FIG. 4 is a conceptual diagram of the functions of a conventional production system. The number 21 in the figure indicates the range controlled by the control function, 22 the working memory function section, 23 the rule memory function section, 24 the conflict set generation function section, 25 the conflict resolution function section, 2G 27 indicates a rule execution function section, and 27 indicates a man-machine interface function section. Applicant Seiko Epson Co., Ltd. Agent Patent Attorney Mogami 1 and 1 other person 2' G3 Diagram Roasted Cow Diagram

Claims (1)

[Scope of Claims] A) At least a) rule memory function, b) working memory function, c) conflict set generation function, d) conflict resolution function, e) at least one set of output functions, f) at least two sets of input functions. , g) includes in its configuration a plurality of slave processor elements having an operation control function for controlling the functions a) to f) above; B) and at least h) a rule memory function; i) a working memory function; j) a rule execution function, k) at least one set of input functions, l) at least two sets of output functions, m) an operation control function for controlling the functions h) to l) above; C) and at least one set of input functions of the slave processor element of A) and at least one set of output functions of the master processor of B) are connected to a common bus; D) and A production system characterized in that at least one set of input/output functions of the slave processor element of (B) and at least one set of input/output functions of the master processor of (B) are connected to each other in series in a ring shape.