JPH04332041A - Rule base type control device - Google Patents

Abstract

PURPOSE:To rapidly execute face adding/deleting operation by storing a competitive set element, and at the time of adding a fact, calculating only a change in the competitive set element and executing storing and retrieval based upon the application of identifiers(IDs) and their attributions. CONSTITUTION:A rule base 10 stores the set of rules. A working memory 12 stores the set of fact data, i.e., fact data, working memory fact IDs, competitive set element IDs, and attributes. A competitive set memory 14 stores the set of competitive set elements. A rule collating data management part 16 has an indicator for indicating a rule in the rule base 10, an indicator for specifying an area to which a new face is to be stored in the working memory 12 and an indicator for operating an area storing an already registered face in the memory 12. The management part 16 executes collation/retrieval by means of these indicators to register/delete respective facts.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a rule-based control device, and more particularly to the management of adding and deleting facts to a working memory and the management of adding and deleting conflicting set elements to a conflicting set memory.

0002

[Prior Art] The production system has IF to TH
Production rules written in EN~ format (hereinafter referred to as
It is a system configured using rules (simply called rules). The "IF~" part of the rule is called the condition part, and the "THEN~" part is called the execution part, and adding or deleting a rule corresponds to changing the inference program.

[0003] Also, forward inference
asoning) is a method of repeatedly making inferences and drawing conclusions based on given facts. That is, a given fact is input into the condition part of the rule and a conclusion is drawn. Treat that conclusion as a newly added fact, enter it again into the condition section of the rule, and draw the next conclusion. By repeating such inference operations, the rules are verified sequentially from the beginning. In other words, it is a method of starting from certain facts and selecting a specific hypothesis as a conclusion.

Production System A rule-based control device that performs forward reasoning is a production system that performs forward reasoning, and includes a rule base for storing rules and a working memory as a memory for registering facts. Furthermore, a conflict set memory is used to store conflict set elements obtained as a result of pattern matching.

[0005]

However, the conventional production system rule-based control device that performs forward inference has a problem in its operation when facts are added or deleted. In other words, when a change occurs in the contents of the working memory (when facts are added or deleted), all the facts in the working memory after the change are compared with all the condition parts in the rule base, and a new competitive set is created. Due to this generation, the following problems have arisen.

(1) Since the conflict set is calculated every time the contents of the working memory change, information regarding the previously calculated conflict set cannot be used effectively.

(2) Furthermore, it takes time to delete the contents of the working memory and the conflict set memory.

(3) Since calculations are also performed for rules unrelated to the attributes of the contents of the working memory, speeding up of operations is hindered.

[0009] As described above, conventionally, there has been a problem that hinders high-speed comparison between the rules on the rule base and the facts on the working memory. An object of the present invention is to realize high-speed matching by effectively utilizing information regarding the previously calculated conflict set.

0010

[Means for Solving the Problems] In order to achieve such an object, the present invention provides a rule base, a working memory, a conflict set memory, a rule matching data management unit,
Each component has the following configuration and function.

First, the rule base stores rules. In addition to a conditional part and an execution part, a rule in the present invention has a working memory attribute that represents the attribute of the fact referred to by the conditional part. Furthermore, each rule is also given a rule identifier.

Next, the facts are registered in the working memory. In addition to the fact data representing the content of the fact, the fact registered in the working memory includes a competition set element identifier representing a related competition set element and an attribute referenced by the condition part of the rule. Furthermore, a working memory fact identifier is assigned to each fact.

The conflict set memory stores conflict set elements having rule identifiers representing related rules and working memory fact identifiers representing related facts, as well as conflict set elements to which conflict element identifiers are individually assigned.

[0014]Furthermore, the rule matching data management section includes a first
, a second indicator, and a third indicator. The first indicator is an indicator for scanning rules on the rule base. The second indicator is an indicator for specifying a new fact on the working memory. The third indicator is an indicator for scanning old facts in working memory.

[0015] When a new fact is registered in the working memory, the rule matching data management unit first registers the first indicator,
A comparison between the rule base and the working memory is performed using the second indicator and the third indicator. This verification is performed for each attribute. In other words, a rule that has the same working memory attribute as the attribute of the registered fact (that is, the condition part refers to the attribute) and a fact that has the same attribute as the attribute of the fact to be registered. Matching is performed. If the verification is successful, a predetermined identifier is written into each member in accordance with this. That is, a rule identifier representing a rule related to successful matching is written in the rule identifier of the competitive set element, and a working memory fact identifier representing a fact related to successful matching is written in the working memory fact identifier of the competitive set element. A new conflicting element identifier is assigned to the conflicting set element. The rule matching data management unit writes the conflicting element identifier into the conflicting element identifier of the related fact on the working memory.

When deleting a fact from the working memory, the rule matching data management section first searches the working memory for a fact having a working memory fact identifier representing the fact to be deleted. This search is performed for each attribute. That is, a search is performed based on facts that have the same attributes as the attributes of the fact to be deleted. Furthermore, a conflict set element having the same conflict element identifier as the conflict element identifier of the fact discovered by this search is searched on the conflict set memory. The rule matching data management unit deletes the conflicting set element thus discovered from the conflicting set memory. After deleting all conflicting set elements discovered by the search, the rule matching data management unit deletes the facts to be deleted from the working memory.

[0017]

[Operation] In the present invention, rules and facts are checked for each attribute, and facts to be deleted are searched for. In other words, rules and facts related to different, unrelated attributes are excluded from the verification target. This results in
Verification operation becomes faster. In addition, in the present invention, as a result of writing a predetermined identifier at the time of new fact registration, it is possible to save previously calculated conflict set elements and use predetermined information for deletion operations, resulting in an operation with low redundancy. And speeding up of the deletion operation is realized.

That is, when a match is made during fact registration, the rule match data management section writes a rule identifier, and as a result, information representing the relationship with the rule is attached to the conflict set element, and the working memory fact identifier is written. As a result, information representing the relationship with facts is attached to the competitive set element. Furthermore, as a result of writing the competing element identifier, information representing the relationship with the competing set element is added to the related facts in the working memory.

In the fact deletion operation executed after the fact registration operation, a fact to be deleted is searched for among facts with the same attribute using a working memory fact identifier. A competing element identifier has been attached to the fact obtained as a result of the search in the previous fact registration operation. Therefore, by using this, related conflict set elements can be searched on the conflict set memory. If you sequentially delete the conflicting set elements obtained through the search from the conflicting set memory and then delete the fact from the working memory after deleting a series of conflicting set elements, you can delete the conflicting set obtained by the previous operation. There is no need to recalculate elements, and information can be used more effectively.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows the configuration of a rule-based control device according to an embodiment of the present invention. As shown in this figure, this embodiment is composed of a rule base 10, a working memory 12, a conflict set memory 14, and a rule matching data management section 16.

The rule base 10 is a means for storing a set of rules, and the rules are composed of a condition part and an execution part as described above. In this embodiment, a rule identifier is further assigned to each rule, and each rule has a working memory attribute. This working memory attribute indicates the fact attribute referred to in the condition part.

Working memory 12 is a memory that stores a set of factual data. In this embodiment, as shown in FIG. 2, data including not only the fact data itself but also working memory fact identifiers, competing set element identifiers, and attributes are stored as facts in the working memory 12. The attribute indicates the attribute of the fact data body. For example, the fact data body "Japan Radio Co., Ltd." corresponds to the attribute "company name." A working memory fact identifier is an identifier assigned to each fact.
The conflicting set element identifier is an identifier representing a related conflicting set element.

The conflict set memory 14 is a memory that stores a set of conflict set elements. A competitive set element is a set element that is constructed when the condition part of a rule is matched with a fact and the match is established. As shown in FIG. 3, the conflict set element includes a conflict set element identifier, a rule identifier, and a working memory fact identifier. The competition set element identifier is an identifier given to each competition set element, and the rule identifier and working memory fact identifier represent rules and facts related to successful matching. A conflict set element may have multiple working memory fact identifiers.

The rule matching data management section 16 has three types of indicators. The first is an indicator for scanning rules on the rule base 10, the second is an indicator for specifying an area on the working memory 12 to add a new fact, and the third is an indicator for scanning rules on the rule base 10.
has an indicator for scanning an area in the working memory 12 where registered (old) facts are stored. The rule matching data management unit 16 uses these indicators to perform matching, searching, and the like. Next, the operation of the rule matching data management section 16 in this embodiment will be explained. In this embodiment, the fact registration and fact deletion operations are performed according to the result of the classification of rules based on attributes. That is, the operation for each attribute, which is the first feature of the present invention, is realized as a result of classification using the working memory attribute of each rule. Rules that refer to different attributes each form a separate set of rules, and the following operations are performed for each set.

First, let us consider the case where a new fact is registered in the working memory 12. In this case, working memory 1
2. The new fact is registered in the area specified by the indicator related to the new fact. The rule matching data management unit 16 then starts matching the rule base 10 and the working memory 12.

In other words, the rule matching data management unit 16 sequentially specifies rules by scanning the indicators related to the rules, and sequentially specifies the rules by scanning the indicators related to old facts, in accordance with the method of production system forward reasoning. The facts on the working memory 12 (however, only the areas where old facts are registered) are sequentially specified. Verification calculations regarding new facts are performed using a known method for the specified rules and facts. However, as described above, this verification is performed only for one unit classified by attribute.

In the case of this embodiment, the number of collations C is determined by the number of rules being Rn, the total number of facts in the working memory 12 being Wn,
Assuming that the number of attributes is a and the number of facts referred to by one rule is Pn, it is expressed as the following equation (1).

[0029]

[Math 1]

[0030] However, Wn≧1.

[0031] When a match is established between a rule on the rule base 10, a new fact newly added to the working memory 12, and an old fact, the rule matching data management unit 16 stores the combination. It is written on the conflict set memory 14 as an identifier. At this time, the rule matching data management unit 16 assigns a conflict set element identifier to the new conflict set element related to the writing, and also adds this to the conflict set element of the fact related to the conflict set element in the working memory 12. Write. In this way, facts, competing set elements, and rules are related to each other during registration.

Next, consider the case of deleting facts. In this case, the rule matching data management unit 16 is sent from a device (not shown).
In response to being given a fact deletion command, the rule matching data management unit 16 searches the working memory 12 using the working memory fact identifier and discovers the fact to be deleted. Note that, as mentioned above, this operation is performed on a group of facts that have the same attributes as the facts to be deleted.

When a fact is discovered, the conflict set memory 14 is searched using the conflict set element identifier of the discovered fact. That is, the rule matching data management unit 16 searches for, discovers, and deletes a conflict set element having the same conflict set element identifier as the conflict set element identifier of the discovered fact. It is assumed here that a discovered fact may generally have multiple conflicting set element identifiers. The rule matching data management unit 16 deletes the discovered fact from the working memory 12 when the deletion of the conflicting set element is completed for all the conflicting set element identifiers that the discovered fact has. In this way, facts are deleted.

As described above, according to this embodiment, the registration deletion operation is automatically executed at high speed without the user being aware of it at all. The high speed is due to the fact that the number of facts to be processed is reduced by attribute classification, the previously calculated competing set elements are saved, and the identifiers are assigned and used effectively. .

Comparing this effect numerically, it is as follows. In the conventional method, that is, the method of calculating all competitive set elements when adding facts, the number of matchings C′ is
C′=RnWnPn
…
(2) becomes. The number of verifications C in this example (however,
(does not include decrease due to rule classification) ratio C/C'
By dividing equation (1) by equation (2),
C/C'={1-((Wn-a)/Wn)Pn}/
It is expressed as aPn+1 (3). Furthermore, by classifying rules by attribute, this can be further improved.
It can be reduced by /a times.

Furthermore, when deleting facts, there is no need to perform a comparison operation by using conflicting set element identifiers, so speeding up can be achieved. Furthermore, it is possible to use a hash function for searching, and it is easy to implement.

[0037]

[Effects of the Invention] As explained above, according to the present invention,
By assigning identifiers, classifying based on attributes, and searching operations,
Since we are trying to reduce the number of objects to be processed, adding facts,
The deletion operation becomes faster and a device with higher usability can be realized.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the configuration of a rule-based control device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a data structure of facts in this embodiment.

FIG. 3 is a diagram showing the data structure of a competitive set element in this embodiment.

[Explanation of symbols]

10 Rule base 12 Working memory 14 Contention set memory 16 Rule matching data management department

Claims (1)

[Claims] Claim 1: A production rule having a conditional part representing a condition, an execution part representing an action when the conditional part is true, an individually assigned rule identifier, and a working memory attribute representing an attribute of the fact referenced by the conditional part. A rule base that stores a working memory fact identifier assigned to each fact, fact data representing the content of the fact, a competing set element identifier representing the related competing set element, and a fact having attributes referenced by the condition part of the production rule. a working memory in which is registered; a conflict set memory that stores conflict set elements having conflict element identifiers assigned to each individual, a rule identifier representing a related production rule, and a working memory fact identifier representing a related fact; and a rule base. a rule matching data management unit having a first indicator for scanning the above production rule, a second indicator for specifying a new fact in the working memory, and a third indicator for scanning the old fact in the working memory; When a new fact is registered in the working memory, the rule matching data management unit creates a production rule that has the same working memory attribute as the attribute of the registered fact, and a production rule that has the same working memory attribute as the attribute of the registered fact. and the fact that has been verified using the first indicator, the second indicator, and the third indicator, and if the match is successful, a rule identifier representing a production rule related to the success of the match and a fact are displayed. Writes the working memory fact identifier into the rule identifier and working memory fact identifier of the conflicting set element, adds a new conflicting element identifier to the conflicting set element, and writes the conflicting element identifier to the conflicting element identifier of the related fact in the working memory. , when deleting a fact from the working memory, a fact having a working memory fact identifier representing the fact to be deleted is searched from among the facts having the same attributes as the fact to be deleted in the working memory, and the fact found by the search is Search the conflict set memory for a conflict set element that has the same conflict element identifier as the actual conflict element identifier, delete the conflict set elements found by the search from the conflict set memory, and delete all the conflict set elements found. A rule-based control device that deletes a fact to be deleted from a working memory after the fact is deleted.