JPH03179527A - Cooperative decentralized type inference device - Google Patents

Abstract

PURPOSE:To improve the reliability for the solution of problems by providing plural secondary problem solving parts and connecting these problem solving parts to each other in steps for cooperative solution of problems. CONSTITUTION:When a problem is inputted to the secondary problem solving part of the highest order, the problem solving part decides whether it can solve a due problem by itself or not. If so, the secondary problem solving part solves the problem by itself and sends this solution to a high-order secondary problem solving part. If not, the problem is decomposed into plural secondary problems via a problem decomposing part 46 and these secondary problems are sent to a low-order secondary problem solving part. This low-order problem solving part repeats the processing to the received secondary problems. When the problems are solved at the low-order problem solving parts, these solutions are sent back to a high-order problem solving part. Then plural solutions are synthesized at a result synthesizing part 47.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a collaborative distributed inference device applied to an expert system.

[Prior Art] Conventional expert systems can be broadly divided into those that perform inference based on rules and those that perform inference based on frames.

In the expert system 1 based on the former rule, as shown in FIG.
Knowledge was extracted from the knowledge base 4 by the inference engine 3. The flow of such reasoning is as follows.

(1) Set initial facts.

(2) Apply the rules existing in Knowledge Base 4 to derive new facts. Such processing continues until there are no more rules to apply.

(3> Display the facts obtained by inference.

In this manner, the conventional expert system 1 has the knowledge of experts in the knowledge base 4, and the inference engine 3 performs a process of extracting that knowledge.

On the other hand, the latter frame-based expert system is different in that the method of expressing knowledge is a frame rather than a rule. Same as system. However, in frame-based reasoning, the target fact is retrieved while searching upper and lower frames of the frame.

[Problem to be solved by the invention] In conventional expert systems, a single knowledge base4
Because it only had 3 and inference engine 3, it had the following drawbacks.

(1) When one part of the system breaks down, the whole system goes down.

(2) Because they only have a single point of view, they give up when faced with a problem they cannot solve.

(3) Since it has only a single viewpoint, the reliability of the inference results inferred by the inference engine 3 is low.

(4) Since everything is processed by one processor, if there is a bottleneck in the middle, the speed will drop significantly.

(5) For geographically dispersed systems, by adopting such a cooperative and distributed configuration, communication time can be eliminated and a substantial response speed can be obtained. Furthermore, if a failure occurs in the communication path, the entire system may go down.

The present invention has been made in view of the above points, and its purpose is to eliminate the drawbacks of the conventional expert systems described above,
An object of the present invention is to provide a cooperative distributed reasoning device that improves the reliability of problem solving.

[Means for solving problems] Consists of a problem solving part that solves a given problem, a problem decomposition part that decomposes the problem into sub-problems, and a result synthesis part that synthesizes the results of solving the sub-problems, making a knowledge base and inference. It has a plurality of sub-problem solving units each having a problem processing unit equipped with an engine, a communication unit for exchanging information with other sub-problem solving units, and a control unit that controls the operation of the sub-problem solving unit. This is a collaborative distributed inference device characterized by cooperatively solving problems by step-by-step combination of problem solving units.

[Operation] The system is distributed to a plurality of sub-problem solving sections and the sub-problem solving sections are made to cooperate to solve the problem.

[Embodiment] A collaborative distributed inference device according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a collaborative distributed inference device. In FIG. 1, numeral 10 is the topmost sub-problem solving unit.

Sub-problem solving units 11 and 12 are connected below this sub-problem solving unit IQ. Furthermore, sub-problem solving units 21 and 22 are connected below the sub-problem solving unit 11. Furthermore, a sub-problem solving unit 21.2 is located below the sub-problem solving unit 12.
2 is connected.

Further, below the sub-problem solving unit 21, there is a sub-problem solving unit 3.
1.32, sub-problem solving units 33.34 are below the sub-problem solving unit 22, sub-problem solving units 34.35 are below the sub-problem solving unit 23, and sub-problem solving units 34.35 are below the sub-problem solving unit 23. Sub-problem solving units 36 and 37 are connected to the lower level.

Next, the configuration of the sub-problem solving section will be explained by taking the sub-problem solving section 11 as an example with reference to FIG. In Figure 2, 41
42 is a communication unit for exchanging information with the upper sub-problem solving unit, and 42 is a communication unit for exchanging information with the lower sub-problem solving unit. The outputs of these communication units 41 and 42 are connected to the control unit 43. This control section 43 includes a problem processing section 44.
is connected and done. The problem processing unit 44 is composed of a problem solving unit 44 that solves a given problem, a problem decomposition unit 46 that decomposes the problem into subproblems, and a result synthesis unit 47 that synthesizes the results of solving the subproblems. An inference section 48 is connected to the problem processing section 44 . The inference unit 48 is composed of an inference engine 49 and a knowledge base 50. This knowledge base 50 stores expert knowledge in the form of rules. The inference engine 49 also uses the knowledge base 5
Pull out the knowledge stored in 0.

Note that communication with humans is performed via the communication unit 41 of the sub-problem solving unit IO in FIG.

Next, the operation of an embodiment of the present invention configured as described above will be explained. First, when a problem is input to the topmost sub-problem solving unit IO, processing as shown in the flowchart of FIG. 3 is started in the collaborative distributed inference device of FIG. 1. First, the sub-problem solving unit IO receives a problem via the communication unit 41 (step Sl). This separate problem solving unit IO determines whether the problem to be solved can be solved by oneself (step S2).

Determining whether or not you can solve the problem yourself is based on 50 knowledge bases.
The decision is made based on whether or not there is a rule base that can be applied to the problem to be solved. If it is determined in step S2 that the problem can be solved by itself, the separate problem solving unit 1O solves the problem by itself (step S3) and sends the solution to the higher level separate problem solving unit (step S4). . Here, if the problem is solved by the top-ranking separate problem-solving section to, the top-ranking separate problem-solving section does not exist, so it is displayed on, for example, a display section (not shown).

By the way, if the separate problem solving unit 10 determines that it cannot solve the problem by itself, that is, in step S2, rN
If it is determined that the problem is OJ, the problem decomposition unit 46 decomposes the unsolvable problem into a plurality of separate problems (step S5).
). Each separate problem is then sent to the lower separate problem solving section (steps 86 to S8). Then, each separate problem solving unit that has sent the separate problem enters a waiting state (νaft). Then, when the lower-level separate problem solving section receives the sent separate problem, it similarly repeats the processing from step S1 onward in the flowchart of FIG. If the problem is solved in the lower-level separate problem-solving section, the solution is sent back to the higher-order separate problem-solving section via the communication section 41. The solution is then received by the upper-level separate problem solving unit (steps S9 to
S11). The solutions solved by the plurality of lower-level separate problem solving units are then synthesized in the result synthesis unit 47 (step S L2). The combined solution is then sent to another higher-level problem solving unit (step S4). As described above, when a problem is given to a separate problem-solving section and it cannot be solved by that separate problem-solving section, after decomposing the problem, Send to. If the problem is solved by the lower-level separate problem solving unit, the higher-order separate problem solving unit synthesizes the solutions. Further, problems that cannot be solved by the lower-level separate problem solving unit are sent to the lower-level separate problem solving unit to find the solution.

Next, FIG. 4 shows an example in which the collaborative distributed reasoning device is applied to an automobile fault diagnosis device. In Figure 4, 61
is a diagnostic device that serves as a top-level separate problem-solving unit. A fuel supply system diagnostic device 62 that diagnoses the fuel supply system, a power storage system diagnostic device 64 that diagnoses the power storage system, and a starter motor diagnostic device 66 that diagnoses the starter motor as separate problem solving units below this diagnostic device 61. , an inspection system diagnostic device 67 that diagnoses the inspection system, and a power train diagnostic device 70 that diagnoses the power train. Furthermore, a carburetor diagnostic device 63 for diagnosing a carburetor brake is connected to the fuel supply system diagnostic device 62 as a lower separate problem solving section. Further, a battery diagnostic device 65 for diagnosing the battery is connected to the power storage system diagnostic device 64 as a lower-level separate problem solving unit. Furthermore, as separate problem solving units below the inspection system diagnostic device 67, a distributor diagnostic device 68 for diagnosing the distributor and a plug diagnostic device 69 for diagnosing the plug are connected.

By configuring the cooperative distributed inference device in this way, a fault diagnosis of an automobile can be performed as follows. When the symptoms of the automobile are given to another problem solving unit at the highest level, that is, the diagnostic device 61, diagnosis begins.

The diagnostic device 6L then breaks down the symptoms into symptoms suitable for solving by the lower-level diagnostic device, and sends them to the lower-level diagnostic device. For example, it sends symptoms such as ``The inside of the air filter is wet,'' which is related to diagnosis of the fuel supply system.

The lower-level diagnostic device that receives the symptom solves the problem itself if it can, and sends the results back to the higher-level diagnostic device. If you cannot solve the problem yourself, divide it into symptoms that are suitable for lower-level diagnostic equipment and send them. After dividing and distributing the problem in this manner and having each diagnostic device solve the problem, the results are sent back to the higher-level diagnostic device to narrow down candidates for the cause of the failure. For example, if the diagnostic device 61 returns a result saying "air filter is clogged" and the diagnostic device 64 returns a result "failure cause unknown," then the diagnostic device 61 determines that "air filter is clogged." It will be lowered by

As described above, reliability can be improved because a large number of separate problem solving units can be prepared and problems can be analyzed from multiple viewpoints. Furthermore, since problems can be solved in parallel in terms of processing speed, processing speed can be improved. Furthermore, for geographically dispersed systems, by adopting such a cooperative and distributed configuration, communication time can be reduced and a substantial response speed can be obtained.

Furthermore, even if a failure occurs in the communication path, the entire system will not go down, and it has excellent fault tolerance.

[Effects of the Invention] As detailed above, according to the present invention, it is possible to provide a collaborative distributed inference device that eliminates the drawbacks of conventional expert systems and improves the reliability of problem solving. .

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a collaborative distributed inference device according to an embodiment of the present invention, FIG. 2 is a block diagram showing the detailed configuration of the sub-problem solving section, and FIG. 3 is a block diagram showing the processing contents of the sub-problem solving section. FIG. 4 is a block diagram showing an example in which the cooperative distributed reasoning device is applied to an automobile failure diagnosis device, and FIG. 5 is a block diagram showing the configuration of a conventional expert system. 10-12...Sub-problem solving unit, 41.42...Communication unit, 43...Control unit, 44...Problem processing unit, 45.
...Problem solving unit, 46...Problem decomposition unit, 47...Result synthesis unit, 48...Inference unit, 49...Inference engine, 50...Knowledge base.

Claims (1)

[Claims] The problem processing unit includes a problem solving unit that solves a given problem, a problem decomposition unit that decomposes the problem into subproblems, a result synthesis unit that synthesizes the results of solving the subproblems, and a problem processing unit that has a knowledge base and an inference engine. A plurality of sub-problem solving units each having a communication unit for exchanging information with the sub-problem solving unit and a control unit controlling the operation of the sub-problem solving unit, and the sub-problem solving units are coupled in stages. A collaborative distributed inference device characterized by solving problems collaboratively.