Decision making method

• US 5,870,730 A
• Filed: 07/07/1995
• Issued: 02/09/1999
• Est. Priority Date: 07/11/1994
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Komura, Fuminobu, Furuya, Masatoshi
• Primary Examiner(s): Cosimano, Edward R.
• Application Number:
  US08/499,203
• Time in Patent Office: 1,313 Days
• Field of Search: 364/130, 382/159, 395/10, 395/11, 395/50, 395/51, 395/54, 395/61, 395/200.1, 395/203, 395/900, 395/905, 395/906, 395/907, 705/1, 705/3, 706/14, 706/47, 706/45
• US Class Current: 706/47
• CPC Class Codes: G06N 5/048 Fuzzy inferencing

Abstract

When an intention is decided by a computer, the conclusion part of a rule is expressed by using an intention level such as an auxiliary verb of intention and an intention is decided on the basis of the expression. A rule for a certain action is described in a format of "IF (status of A), THEN (action B on the basis of the status A)+(intention level)" beforehand. A fitness degree indicating the fitness state of the status A to the status to be processed is calculated. When the fitness degree at the strongest intention level is highest, the corresponding rule is assumed as an intention in the status to be processed. When the fitness degree at the strongest intention level is not highest, it is converted to a fitness degree equivalent to the intention level of the next strength and a rule which is an intention to an action is decided.

Claims

1. A decision making method of deciding intentions to actions using rules including condition parts indicating statuses and conclusion parts indicating actions according to said condition parts, comprising:

a first step of deciding a strength of an intention by an auxiliary verb which indicates a strength of intention by a numerical value, to an action for each said conclusion part;

a second step of calculating a fitness degree indicating a fitness state of the status of each said condition part to the target status; and

a third step of deciding a first rule as the intention to an action when the fitness degree of said first rule has a biggest fitness value and the auxiliary verb in said first rule has a biggest numerical value among said rules.

2. The decision making method according to claim 1 wherein when there are a plurality of actions, the intentions of which are to be decided simultaneously, further comprising:

(a) classifying said actions per each predetermined attribute;

(b) executing said first step to said third step for each classified attribute and deciding predetermined rules which are intentions to actions;

(c) deciding all said predetermined rules as intentions to actions when all said predetermined rules can be executed; and

(d) deciding the intentions to be the actions of the executable rules on the basis of said fitness degrees and said strengths of intentions when all said predetermined rules cannot be executed.

3. The decision making method according to claim 1 further comprising:

a fourth step of calculating a new fitness degree from the fitness degree of the first rule on the basis of the fitness degree and a strength of intention of a second rule having a same strength of intention as said first rule when the fitness degree of the first rule is not highest among said rules;

a fifth step of deciding said second rule as an intention to an action when the fitness degree calculated at said fourth step or the fitness degree of said second rule, whichever is higher, is highest among said rules; and

a sixth step of executing said fourth step and said fifth step for said second rule when one of the fitness degree calculated at said fourth step and the fitness degree of said second rule is not highest among said rules.

4. The decision making method according to claim 3, wherein when it is assumed that the strength of the intention of said first rule is L, and the fitness degree thereof is X, and the strength of the intention of the second rule is M, said fourth step calculates a new fitness degree from the formula 1-M/L (1-X).

5. The decision making method according to claim 3, wherein when it is assumed that the strength of the intention of said first rule is L, and the fitness degree thereof is X, and the strength of the intention of the second rule is M, and the fitness degree thereof is Y, said fourth step calculates a new fitness degree from the formula Y+-(M/L) X and when the calculated fitness degree is more than the predetermined value, said fourth step sets the fitness value to said predetermined value.

6. A vehicle controller comprising:

sensors for calculating data relating to the motion state of a vehicle at a predetermined time, a storage device for storing a rule comprising a condition part indicating a status and a conclusion part comprising motion of said vehicle corresponding to said status and an intention level indicating a strength of intention for performing said motion, a fitness degree calculator for calculating a fitness degree indicating a fitness state of said status of the condition part to the data calculated by said sensors at said predetermined time, a weight calculator for calculating the weight of the intention of the motion of said conclusion part on the basis of the fitness degree calculated by said fitness degree calculator, and a controller for controlling the motion of said vehicle according to the following conditions at said predetermined time;

(1) according to a first rule when said weight of intention of said first rule at the highest intention level is biggest among the rules stored in said storage device;

(2) means for converting the weight of intention of said first rule so as to fit a second rule at the next intention level of said first rule, by a formula 1-M/L (1-X), where L is the strength of intention of said first rule, X is the fitness degree and M is the strength of intention of the second rule, and ascertaining whether one of said calculated weight of intention and the intention level of said second rule is biggest when said weight of intention of said first rule at the highest intention level is not biggest among the rules stored in said storage device; and

(3) means for repeating said converting until the weight of intention of the rule at the highest intention level is maximized according to the second rule when one of them is biggest or for said second rule when both of them are not biggest.

7. An analyzer of a public opinion survey assigning a strength of intention to each answer beforehand, comprising:

a support rate calculator for collecting answers to each question of a public opinion survey and calculating a support rate of each answer; and

a representative answer decider for ascertaining whether the strength of intention of a first answer, the calculated support rate of which is highest, is biggest or not among all the answers, assuming said first answer as a representative answer of said public opinion survey when it is biggest, converting the strength of intention of said first answer, by a formula 1-M/L (1-X) (where L is the strength of intention of said first answer, M is the strength of intention of a second answer having a support rate next to said first answer, and x is the support rate of said first answer, when it is not biggest, and deciding a representative answer of said public opinion survey on the basis of the converted strength of intention.