JPH06161756A - Restriction relaxing inference method - Google Patents

Abstract

PURPOSE:To obtain a solution of high precision in consideration of all restriction conditions in inference which relaxes restriction conditions. CONSTITUTION:A generation rule is applied to a selection node, which is developed (102); and the restriction conditions are applied to the obtained node to determine the degree of sufficiency showing the degree of sufficiency of the restriction conditions (104). Evaluation points are given (106) to the node and when there is a node satisfying the reference degree of sufficiency, the process is repeated. When not, on the other hand, back tracking is carried out to search for another node satisfying the reference degree of sufficiency (116) and when the target node can not be reached even after the frequency of back tracking exceeds a limited value, the reference degree of sufficiency is relaxed (114) to repeat the inference by using the relaxed reference degree of sufficiency.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a constraint relaxation inference method, and more particularly, it relaxes a constraint condition when a search tree is generated by a search and inference is performed, and a solution satisfying all constraint conditions cannot be obtained. The present invention relates to a constraint relaxation inference method for promoting inference.

[0002]

2. Description of the Related Art When inferring using a plurality of constraint conditions, there are cases where a solution satisfying all of these constraint conditions cannot be obtained. In such cases, it is necessary to somehow relax the constraints and find an appropriate solution. Conventionally, as a method of relaxing constraint conditions, a method of defining constraint strength representing the strength of the constraint condition for all constraint conditions and removing the constraint condition from weak constraint strength to proceed with inference Is known (Japanese Patent Laid-Open No. 3-67334).
issue). This method removes the application of the constraint with the weakest constraint strength, that is, does not consider it as a constraint when the solution cannot be obtained by considering all the constraints, and advances the inference to obtain a solution again. When it is no longer possible, the constraint condition with the weakest constraint strength is removed next, and so on.

[0003]

However, when the constraint strength is defined in the constraint condition as in the above-mentioned conventional method, the constraint conditions are ordered from the viewpoint of strong and weak. There is a problem that it is not possible to give an appropriate constraint strength when there are a plurality of constraint conditions.

Further, in the method of using the constraint strength and sequentially removing the constraint condition from the weak constraint strength to proceed with the inference, since the order of removing the constraint conditions occurs, only the constraint condition with the strong constraint strength is satisfied. Even if it is not done, all constraint conditions with weaker constraint strength must be removed, and the accuracy of the solution becomes worse. further,
Since the solution can be obtained without considering the removed constraint condition at all, the accuracy of the solution obtained by the removal of the constraint condition decreases.

The present invention has been made in order to solve the above problems, and when relaxing the constraint conditions, there are a plurality of constraint conditions having the same constraint strength or a case where the constraint strength has an order. It is an object of the present invention to provide a constraint relaxation inference method that can obtain a highly accurate solution by considering all constraint conditions.

[0006]

In order to achieve the above object, the present invention provides a step of expanding a node according to a generation rule to generate a new node, and each step generated based on a predetermined constraint condition. The degree of satisfying the constraint condition at a node is expressed in a plurality of stages from a degree of completely satisfying the constraint condition to a degree of not satisfying the constraint condition at all, and is stored as a satisfaction degree, and the satisfaction degree is While selecting as a node for developing a node satisfying the standard satisfaction degree, when the satisfaction degree does not have a node satisfying the standard satisfaction degree, a value obtained by relaxing the reference satisfaction degree is the satisfaction degree as the reference satisfaction degree. The target node is inferred by repeating the step of selecting a node satisfying the standard satisfaction degree as a node for developing the node and repeating the process.

[0007]

According to the present invention, the nodes are expanded according to the generation rule to generate new nodes, and each newly generated node is defined with a degree of sufficiency representing the degree to which the predetermined constraint condition is satisfied. . The degree of sufficiency is an index showing how much the constraint condition is satisfied, and is expressed in a plurality of stages from a degree of completely satisfying the constraint condition to a degree of not satisfying the constraint condition at all. Remembered.

The sufficiency of each node is compared with the reference sufficiency, and a node whose satisfaction satisfies the reference sufficiency is selected as a node for expansion. On the other hand, when there is no node whose satisfaction degree satisfies the standard satisfaction degree, the value of the standard satisfaction degree is relaxed, the relaxed standard satisfaction degree is compared with the satisfaction degree, and the value is relaxed. The nodes satisfying are selected as the nodes for expansion. Then, by repeating the above, the solution that is the target node is inferred.

[0009]

EXAMPLES Examples of the present invention will be described below. In the present embodiment, the present invention is applied to a product allocating device for allocating a large number of products to a plurality of coating lines. As shown in FIG. 1, this embodiment includes a condition storage unit 10 that stores constraint conditions and a rule storage unit 12 that stores generation rules and evaluation rules. The condition storage unit 10 and the rule storage unit 12 are connected to an inference unit 14 that performs inference while expanding nodes by search to generate a search tree. The inference unit 14 is connected to a schedule display unit 16 that displays the inferred result as a schedule.

An example of the operation of this product allocating device is to appropriately allocate eight kinds of product ac different in color, production amount and size to four automated coating lines A to D. I will explain it. Painting line A-D, product A-
The characteristics of the black circles are shown in Table 1 and Table 2, respectively.

[0011]

[Table 1]

[0012]

[Table 2]

The inference by the inference unit 14 is performed while expanding the nodes to generate a search tree, and the allocation is completed as the node goes to the lower part of the search tree, and the optimum allocation determined by the lowest node, That is, it becomes a solution.

The rule storage unit 12 stores rules for performing inference. The generation rules and evaluation rules used in the above example are shown below.

Generation Rule: (1) If there is still room in the coating line and there are 30 or more products remaining, the product is assigned to the coating line with room.

(2) If there are products with red or purple color remaining,
Since those products have limited coating lines, the products are assigned to compatible coating lines.

Evaluation rule: (1) If a product whose color is the same as the reference color of the coating line is allocated, the allocation amount x 1 is added to the score.

(2) If the difference between the maximum size and the minimum size of the products allocated in the same coating line is within 30, the score of 30 is added to the score.

Further, the condition storage unit 10 stores the following constraint conditions. Constraints: (1) If the difference between the maximum size and the minimum size of the products allocated in the same coating line is 60 or more, the satisfaction level is 0.2.

(2) Satisfaction = if a product whose color is neither the standard color nor the corresponding color is assigned to the coating line
Set to 0.

(3) Since the coating lines A and D are coating lines for small sizes, the size of the coating lines A and D is 40 to 6
If 0 products are assigned, the satisfaction level is 0.8.

(4) Since the coating lines A and D are coating lines for small sizes, if a product having a size of 70 or more is allocated to the coating lines A and D, the degree of satisfaction is set to 0.6.

The inference unit 14 performs inference by relaxing constraint conditions and backtracking backward in the search tree. This inference will be described with reference to the flowcharts of FIGS. 9 and 10.

In the initial state, no product is assigned to all painting lines. Therefore, step 100
In, the node in the initial state is the selected node, and the standard satisfaction is 1. It should be noted that the satisfaction level = 1 means that the constraint condition is 100%.
Satisfaction, that is, complete satisfaction, means that the satisfaction degree = 0 does not satisfy the constraint condition at all, and 1> satisfaction degree> 0 does not satisfy the constraint condition at all. Means a condition between degrees.

In step 102, the generation rule read from the rule storage unit 12 is applied to the selected node in the initial state to develop the node, and a plurality of nodes are created by allocating the product to any one painting line. In FIG. 2, the generation rule (1) is applied, and 30 or more products C, D, E, and F are assigned to the coating line A with a standard production of 50, and 50 products C are assigned to the painting line A. A node in which 30 products are assigned to the coating line A A node in which 40 products are assigned to the coating line A A product is shown in an example in which 4 nodes are created, 30 nodes assigned to the coating line A .

At the next step 104, the constraint condition is applied to the created nodes to determine the "sufficiency" representing the satisfaction of each node. It should be noted that the initial value of the degree of sufficiency is 1, and if none of the constraint conditions is satisfied, the degree of sufficiency of the node is left at the initial value, that is, 1. In the example of FIG. 2, the constraint condition (3) conforms to the node and the sufficiency of the node becomes 0.8, and the constraint condition (2) and the constraint condition (3) conform to the node and the sufficiency of the node is minimum. The value is 0.
The degree of sufficiency is 1 because no constraint condition is satisfied for node and
It remains.

In step 106, the above-mentioned evaluation rule is applied to the nodes for which the degree of sufficiency is determined and the degree of sufficiency is not 0, and points are given to each node. In the example of FIG.
Since the blue product C, D, and E, which are the same color as the standard color of the line, are assigned to the coating line A whose standard color is blue, the evaluation rule (1) conforms to the nodes ,, and the allocation amount x 1 By calculation, 50 points for each of the nodes ,, 3
0 points and 40 points (evaluation points) are given. Since the degree of sufficiency is 0 for a node, no score is given.

In step 108, it is judged whether or not there is a node whose satisfaction degree satisfies the reference satisfaction degree. In the example of FIG. 2, since there is a node satisfying the criterion satisfaction = 1, the process proceeds through step 120 to step 110, and the node having the highest score among the evaluated nodes is selected as the selected node. In the above example, the node with the highest score is selected as the selected node. In the above example, since the reference satisfaction level is 1, the selected node has the highest score among the nodes having the satisfaction level of 1. If there are multiple nodes with the same score, all the nodes with the same score are selected.

Next, returning to step 102, the above generation rule is applied to the selected node again to expand the node, thereby extending the next branch. In this example, the generation rule (2) is satisfied, and as shown in FIG. 3, a node where 30 product powers are assigned to the coating line C is a node where 30 product powers are assigned to the coating line D. Three nodes are obtained, the two assigned nodes. The three nodes obtained,
On the other hand, in step 104, the constraint condition is applied as described above. In the above example, the satisfaction condition is 0.8 because the constraint condition (3) matches the node. Since there is no constraint that matches the node, the sufficiency remains 1. Then, in step 106, the evaluation rule is applied in the same manner as described above to give a score. In this case, the evaluation rule (1) is satisfied, and the scores of the nodes ,, are 50 points, 50 points, 52 points, respectively.
It becomes a point.

Then, step 108 and step 110
As described above, the node having the highest score is selected as the selected node from among the nodes whose sufficiency satisfies the reference sufficiency, that is, the node having the highest score is selected, and the process returns to step 102 to apply the generation rule to branch Proceed with the postponement search.

It is assumed that the search tree as shown in FIG. 4 is obtained by repeating the above steps 102 to 110.

In FIG. 4, the two nodes b and c generated by applying the generation rule to the thick-lined node a below the search tree are both satisfied with a constraint condition of 0. That is, it is impossible to extend the branch any further and proceed with the search. Therefore, the process proceeds from step 108 to step 112 to perform backtracking. When the process first proceeds to step 112, the backtrack is the first time and the number of backtracks is equal to or less than the limit value (for example, 2).
Proceed to step 16 to backtrack and search for another branch.
That is, backtracking is performed as shown by the broken line arrow in FIG. 4, and the node X is selected as the selected node.

FIG. 10 shows the details of the backtrack in step 116 of FIG. 9. In step 200, the currently selected node is set to the node X, and in step 210, the node X has a horizontally arranged node. It is determined whether or not there is a node which satisfies the standard satisfaction degree and whose evaluation point is lower than the evaluation point of the node X. If the determination in step 210 is negative, in step 220, the parent node of the node X (node whose expansion is one before) is set as the node X in step 21.
The judgment of 0 is made again. On the other hand, when the determination in step 210 is affirmative, in step 230, a node that is a horizontal array of the nodes X and that satisfies the reference satisfaction level and has the next highest evaluation point of the node X is set as a new selected node. By this backtracking, the node X that satisfies the reference satisfaction degree below the search tree and has the next highest evaluation point after the node a is selected as the selected node.

However, even if the backtracking is performed for the first time as described above, it may not be possible to proceed with the search. In the example of FIG. 5, the node d generated ahead of the node X due to the expansion of the node X has a degree of sufficiency of 0 and cannot proceed with the search. In this case as well, backtracking is performed again as indicated by the dashed arrow to select the node Y as the selected node.

However, as shown in FIG. 6, the degree of sufficiency may be 0 at the tip of the node Y, and eventually the search may not be able to proceed. In such a case, backtracking may be performed and another search may be performed again, but there are cases where the solution cannot be reached forever. Therefore, the number of backtracks is limited as described above. For example, the number of backtracks is limited to two. Then, in the example of FIG. 6, the search cannot be completely advanced. Therefore, the process proceeds from step 112 to step 114 to relax the limiting condition. That is, the value of the reference satisfaction degree is relaxed, the constraint of the reference satisfaction degree is set to 0.9 ≦ reference satisfaction degree <1, and the node satisfying this condition is searched in step 118. If there are multiple nodes, 1) a node with a high degree of sufficiency 2) a node at the bottom of the search tree 3) a node with a high evaluation score will be prioritized and selected. If there are a plurality of nodes that have exactly the same condition, all of the plurality of nodes may be selected. In the example of FIG. 6, the node Z having a satisfaction level of 0.9 is selected as the selected node. After that, the process proceeds to step 108 and the above steps 102 to 110 are repeated to proceed with the search in the range of 0.9 ≦ satisfaction degree <1.

The search may be continued as it is, and the situation as shown in FIG. In this case, the satisfaction degree of the generated node W does not satisfy the condition of 0.9 ≦ satisfaction degree <1. In addition, even if an attempt is made to backtrack, there are no nodes that satisfy the condition of 0.9 ≦ satisfaction degree <1. In this case, the constraint condition is relaxed again and the reference satisfaction degree is set to 0.8 ≦ reference satisfaction degree <0.9, and the search is advanced in this range. Then, the above steps are repeated until it is determined in step 120 that the allocation is completed, and finally, in FIG.
A node V as a solution that is a target node shown in is obtained, and in step 122, the allocation schedule is set in the schedule display unit 1.
Display in 6.

Although the present invention has been described above with reference to an example in which the present invention is applied to a product allocating device, the present invention is not limited to this and is applied to artificial intelligence and various systems utilizing artificial intelligence. It is possible.

[0038]

As described above, according to the present invention, by defining the degree of sufficiency in a constraint condition, the constraint condition that must be absolutely complied with is relaxed if desired. Since it can be expressed flexibly, even when there are multiple constraints with similar strength, it is possible to give an appropriate degree of satisfaction, and always consider all constraints without removing them. Since a node whose degree satisfies the standard satisfaction degree is selected, an effect that a highly accurate solution can be obtained is obtained.

[Brief description of drawings]

FIG. 1 is a block diagram of a product allocating device according to this embodiment.

FIG. 2 is a diagram showing a state in which selected nodes in an initial state are expanded to generate nodes ˜.

FIG. 3 is a diagram showing a state in which selected nodes are expanded to generate nodes ˜.

FIG. 4 is a diagram showing a state of backtracking from a node a to a node X.

FIG. 5 is a diagram showing a state in which backtracking is performed from a node X to a node Y.

FIG. 6 is a diagram showing a state of backtracking from a node Y to a node Z.

FIG. 7 is a diagram showing a state in which a node W is obtained by expanding the node Z.

FIG. 8 is a diagram showing a state in which a target node V is obtained.

FIG. 9 is a flowchart showing an inference routine in the inference unit.

10 is a flow chart showing details of step 116 of FIG.

[Explanation of symbols]

 10 Condition Storage Section 12 Rule Storage Section 14 Inference Section 16 Schedule Display Section

Claims (1)

[Claims] 1. A step of expanding a node according to a generation rule to generate a new node, and a degree of satisfying the constraint condition at each node generated based on a predetermined constraint condition, A step of expressing the degree of complete satisfaction to the degree of not satisfying the constraint at all in a plurality of stages and storing as a satisfaction degree, and a step for developing a node in which the satisfaction degree satisfies a reference satisfaction degree Along with the selection, when there is no node whose satisfaction satisfies the standard satisfaction, the value satisfying the standard satisfaction is defined as a value satisfying the standard satisfaction as a value satisfying the standard satisfaction as a value satisfying the standard satisfaction. A constraint relaxation inference method in which a target node is inferred by repeating the selecting step and.