JP2002041291A - Setting method of computation rule, computation rule porting unit using the method and artificial animal - Google Patents

Abstract

PROBLEM TO BE SOLVED: To set versatile individuality to each individual maintaining the tendency of population. SOLUTION: A characteristic value distribution that indicates the range of a degree of the character of a population is to be set for every type of plural characteristics that represents the tendency of the population, and a characteristic value included in each characteristic value distribution for an arbitrary individual attached to the population is to be extracted randomly. Further, a correlation data, a relation of combination of extracted characteristic values to both input pattern and behavior pattern is to be prepared. The behavior pattern that varies with individuals to each type of the input pattern appears by porting a characteristic value table that shows a combination of different characteristic values for every individual and a correlation data for each characteristic value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a moving object (hereinafter referred to as a "moving object") that executes a predetermined behavior pattern under the control of a computer, such as a breeding character presented on a display such as a pet robot or an electronic portable device. Artificial animal "
That. ), And particularly relates to a method of setting a calculation rule for determining the behavior pattern of these artificial animals for this type of calculation device.

[0002]

2. Description of the Related Art In recent years, various pet robots imitating animals such as dogs and cats have been developed. These pet robots are provided with a large number of sensors for obtaining sensations such as sight, hearing, touch, and the like, and are set so as to take a predetermined action according to a stimulus given through these sensors.

Further, in such a robot, a plurality of input signal patterns (hereinafter, referred to as "input patterns") obtained by each sensor are provided so that different responses are made to the same stimulus from time to time. A variety of behavior patterns are prepared, and each behavior pattern is set so as to appear with a predetermined probability.

FIG. 33 shows an example of a table (hereinafter, referred to as an “individuality table”) in which the appearance probabilities of action patterns for various input patterns for an arbitrary individual A are set. The “input pattern” specifically refers to the type of sensor that obtained the input signal, the installation position, the intensity of the sensor output, etc.
Although shown as data indicating the content of the sensor output, it can be converted as a pattern of a stimulus given to an individual from an inference operation using these data. Therefore, in the following description, each input pattern will be described by replacing it with a stimulus pattern obtained by inference (stroking the head, cuddling, ticking the belly, calling the name, etc.).

[0005] The "action pattern" indicates the content of the action according to the type of the individual, such as "escape", "jump", and "bark". As shown in FIG. Probability is set. In the drawing, there is an action pattern whose appearance probability is set to zero depending on the input pattern, such as action pattern 1 for input pattern 2.

When the arithmetic unit in the robot determines the content of the input pattern based on the input signal from the sensor, it determines an action pattern corresponding to the input pattern using the personality table, and activates a predetermined actuator to activate the predetermined actuator. Execute the action pattern. Therefore, even for the same stimulus, the response of the robot changes from time to time, so that a setting corresponding to the "mood" of a living animal can be performed.

Further, since a plurality of individuality tables are created, if the individuality tables incorporated between individuals differ, as shown in FIG. 34, even with the same input pattern, various action patterns have different probabilities. Appears and individuality can be set for each individual.

The setting of the personality table is determined with reference to the properties of the population of pet robots. For example, if there are dog-type robots and cat-type robots, populations according to each type are assumed, and for each population, "behavior that looks like a dog" and "behavior that looks like a cat"
By setting a personality table in which a predetermined appearance rate is set for each individual with respect to the behavior pattern considered to be, a personality table adapted to each population is set (see FIG. 35).

However, when actually moving the robot,
The appearing behavior pattern may not be adapted to the population (for example, a behavior that does not look like a dog even though it is a dog robot). For such individuals, the setting of the individuality table is modified until the expected behavior pattern appears in the population (see FIG. 36).

The above-mentioned individuality table is prepared in advance by a programming device such as a personal computer, and is read into a memory before being mounted on the computing device in the programming device, or transmitted to the computing device via a cable. By such a method, it is transplanted to the robot side.

FIG. 37 conceptually shows functions of the programming device. "Individuality setting function" in the figure
This is a function of creating the individuality table while supporting a process of setting an appearance probability of various action patterns for the input pattern by an operator of the programming device. The “reset function of personality” is a function of rewriting the personality table based on the reset while supporting the reset processing of the personality table shown in FIG. The “personalization transplanting function” refers to a function of incorporating the above-described personality table into the arithmetic device on the robot side.

[0012]

In the above method, there is no specific definition as to what action the population can take in response to various input patterns.
The setting of the behavior pattern of each individual is likely to deviate from the appearance range of the behavior pattern (“individuality description rule” in FIG. 35) assumed for the population. In particular, if the variation in the appearance probabilities of the behavior patterns among individual individuals is increased in order to provide individuality with diversity, many individuals that deviate from the population are generated.
As described above, when the setting of the individuality table does not conform to the population, the correction work must be performed many times until the personality table is adapted. Therefore, a great deal of labor is required.

On the other hand, if the method of reducing the variation of the individuality table is adopted so that each individual is included in the population, individuality cannot be diversified, and any individual takes the same operation pattern and has a unique individuality. The problem that an individual cannot be provided arises.

That is, in the conventional personality setting method, FIG.
As shown in FIG. 8, the maintenance of the population tendency (refers to the appearance tendency of the behavior pattern taken by each individual) and the diversification of individuality are in an opposite relationship, and in order to give priority to the maintenance of the population tendency, There has been a problem that the individuality of the individual becomes poor.
The present invention has been made in view of the above-mentioned problems, and has as its object to set various personalities for each individual while maintaining the tendency of the population.

[0015]

SUMMARY OF THE INVENTION The present invention provides an arithmetic operation for responding to a plurality of individuals responding to a predetermined input signal in a predetermined manner determined by arithmetic processing in a different manner for the same input signal depending on the individual. Applied when setting rules.

The input signal is a signal obtained from a means for outputting a signal in response to an external stimulus, such as a tactile sensor, a visual sensor, a sound sensor, or an operation switch in an artificial animal, and is given as a single signal. In some cases, the signal may be given as a combination of a plurality of types of signals. The reaction in a predetermined manner with respect to the input signal is realized by a control signal (hereinafter, referred to as a “reaction signal”) obtained by performing arithmetic processing on the input signal. The operation rule defines what kind of response signal is output for a predetermined input signal, and if necessary, for the input signal,
It is also possible to set a rule that outputs a plurality of types of reaction signals with a predetermined probability.

In the present invention, the "characteristic" of each individual is regarded as "a degree of a certain property", and data obtained by digitizing the degree is defined as a "characteristic value". Further, in the present invention, it is considered that the individuality of an arbitrary individual includes a plurality of types of characteristics, and the individuality of the individual is represented by a combination of characteristic values for various characteristics. Then, after setting a distribution state of characteristic values in the population for each of a plurality of types of characteristics as a range of characteristics of the population to which the plurality of individuals belong, for each individual, a characteristic included in the population for each characteristic Randomly extract the values,
Calculation rules for a plurality of types of input signals are set based on combinations of the extracted characteristic values.

The distribution of characteristic values in the population is as follows:
It can be indicated by the range of characteristic values that can be possessed by individuals belonging to the population and the ratio of individuals having each characteristic value. The random extraction of characteristic values means extracting an arbitrary characteristic value from the distribution of characteristic values in the population according to the ratio of individuals having the characteristic. In this process, for example, a predetermined number of random numbers are assigned to each characteristic value included in the distribution according to the ratio, and a random number is randomly generated in an arithmetic circuit, and a characteristic value corresponding to the random number is selected. (A so-called lottery process).

The calculation rule can be set, for example, such that a specific signal of a plurality of types of reaction signals is output in response to a predetermined input signal. It is also possible to set a rule that generates a plurality of types of reaction signals with a predetermined probability for a predetermined input signal. When determining a response signal to a plurality of types of input signals, the calculation rule includes an inference rule for determining the specific content of the stimulus from a combination of the input signals.

According to the above-described method, it is possible to set an operation rule that shows a different response to the same input signal depending on a combination of characteristic values. In addition, since the combination of the characteristic values is extracted from the range of the characteristic in the population set in advance, a reaction that deviates from the possible response of the population does not occur, and the tendency of the population does not occur. , And various personalities can be set.

The above method can be applied to an apparatus for transplanting an operation rule for generating a reaction signal to a plurality of operation apparatuses that generate and output a response signal to a predetermined input signal by an arithmetic processing. . In this case, the apparatus includes a setting data input unit that receives, as setting data indicating a range of characteristics of a population of each arithmetic device, data for setting a distribution state of characteristic values in the population for each of a plurality of types of characteristics. And storage means for storing the setting data received by the setting data input means, for each arithmetic device,
An operation rule generation unit for randomly extracting characteristic values included in a population for each characteristic from the storage unit and generating operation rules for a plurality of types of input signals based on a combination of the extracted characteristic values; Calculation rule output means for writing the calculation rules generated by the means into the memory of each processing device.

The setting data input means includes, as setting data of a distribution of characteristic values for each characteristic, data relating to at least a range of characteristic values that can be taken by individuals in a population and a ratio of individuals having each characteristic value. Accept the input of. Note that this type of data input can be performed by operating a keyboard or a mouse, but is not limited to this, and it is possible to input transmission data from another device via a network line,
By reading data recorded on a predetermined recording medium, data input can also be performed.

The calculation rule output means is, for example, electrically connected to the calculation device and writes the calculation rule into a memory via the CPU of the calculation device, or is stored in the memory mounted on the calculation device before mounting. It is constituted by means for writing rules.

Further, the above-mentioned operation rule transplantation device has the following configuration based on the operation rules and the setting data in the storage means.
A response pattern output unit that creates data indicating an output pattern of a reaction signal that can be output by the population with respect to a predetermined input signal and outputs the data to the outside; and an input of data for correcting setting data in the storage unit. It is possible to include a correction data input means for accepting, and a setting data rewriting means for accepting the input of the correction data after the data output by the reaction pattern output means and rewriting the setting data in the storage means. According to these configurations, after properly setting the range of the characteristics in the population, the calculation rule for each processing device can be reset by the characteristic value extracted from the reset range. It is possible to set an appropriate calculation rule corresponding to the population.

Further, the present invention is applied to an artificial animal having an arithmetic unit for generating and outputting a response signal to a predetermined input signal by arithmetic processing, and executing means for executing an action pattern according to the response signal. be able to. The calculation rule incorporated in the calculation device includes, as a range of characteristics of the population of artificial animals, a distribution of characteristic values in the population for each of a plurality of types of characteristics, and then a characteristic included in the population for each characteristic. Since the values are set based on the combination of the characteristic values obtained by randomly extracting the values, it is possible to provide an artificial animal that does not deviate from the population.

[0026]

1 to 9 show the concept of a method for setting personality in one embodiment of the present invention. In this embodiment, a population is assumed for each species such as a dog and a cat with respect to an artificial animal such as a pet robot, and data indicating what characteristics the population tends to have (hereinafter referred to as data). After setting “population tendency”), the individuality of each artificial animal belonging to the population is extracted from the population tendency.

As shown in FIG. 1, the population tendency can be broken down into a plurality of types of characteristics. In this embodiment, each characteristic is defined as “degree of property”, and “characteristic items” indicating the contents of each characteristic are set.
For each of these characteristic items, population tendency is indicated by data obtained by quantifying the degree of the property (hereinafter referred to as “characteristic value”).

FIG. 2 shows a setting example of each characteristic. The left column in the figure is the above-described characteristic item, which functions as a “headline” indicating the content of the target characteristic. In each characteristic item, a word indicating a contradictory concept in the item is set. In this embodiment, these contradictory words are used as both poles to set an axis indicating the spread range of the property (hereinafter, referred to as a "characteristic axis"). For example, one pole is "100" and the other pole is "0". By associating each pole with a numerical value, the degree of characteristics is quantified.

FIG. 3 shows the principle of a method for expressing a population tendency using the characteristic axis. The horizontal axis in the figure is the characteristic axis, and the vertical axis shows the selection probability of the characteristic value. In this example,
When setting the population tendency, first, as shown in FIG. 3A, a standard value of the characteristic value in the population is set. Subsequently, as shown in FIG. 3 (2), the spread range of the population on the characteristic axis is set based on the standard value, and the selection probability in the population for each characteristic value included in this range is set. Set. (Hereinafter, the two-dimensional data represented by the characteristic value and the selection probability will be referred to as “characteristic value distribution”.)

Therefore, the total of the selection probabilities (total of the integrals of the characteristic value distribution pattern) shown in this characteristic value distribution is 10
0%. However, as shown in FIG. 4, the characteristic value distribution for one population may be divided into two or more distribution curves.

When a plurality of populations such as "dog" and "cat" are assumed, a characteristic value distribution for each population for the same characteristic item is obtained, as shown in FIG. Different distributions can be obtained. The total of the selection probabilities in the characteristic value distribution of each means is 100%.

In this embodiment, as shown in FIG. 6, characteristic value distributions are set for a plurality of types of characteristics for each population, and the population tendency is represented as a set of characteristic value distributions. Note that the dashed line in FIG. 6 indicates the position of the standard value in each characteristic value distribution, but there are also cases where the standard value is not at the center of the distribution, as in the characteristic value distribution of the second population in FIG. is there. Further, as shown in FIGS. 6 (4) and (5), no characteristic value distribution is set for characteristics that do not apply to the population.

In this embodiment, an arbitrary characteristic value is extracted from the characteristic value distribution for each characteristic item representing the population tendency, and a combination of the extracted characteristic values is used to determine the combination of any individual in the population. Set as personality. FIG. 7 shows a specific example of the setting of the individuality, in which a characteristic value table indicating the characteristic value of each characteristic item for an arbitrary individual in the population is set. This table stores, for each characteristic item, characteristic values extracted from the characteristic value distribution by a lottery process corresponding to the selection probability. Of course, the set values for the characteristic value distribution not set in the population are blank.

Further, in this embodiment, as shown in FIG. 8, a combination pattern of a predetermined number of characteristic items is associated with an input pattern and an action pattern in the characteristic value table (hereinafter referred to as "correlation data"). )). This correlation data indicates a combination of characteristics related to various input patterns, and for the input pattern,
It shows the types of action patterns that can appear depending on the characteristic values of the combined characteristics and the degree of appearance.

FIG. 9 shows a specific example of the correlation data.
In the illustrated example, the action of “sitting” in response to an instruction input of “sitting” as an input pattern is a degree of understanding, a degree of obedience,
It indicates that it is related to characteristic items such as the degree of whims and the degree of anti-hunger. Although not shown here, between the characteristic items and the behavior patterns, the degree of probability that “sitting” appears with each characteristic value is also set.

That is, the degree of influence of each characteristic on the appearance of the behavior pattern differs depending on the characteristic value possessed by each individual. For example, in the example of FIG. 9, in the case of an individual having a high degree of understanding and obedience, an instruction input by the owner is
Although it is set so that “sit” appears with a high probability, the setting is made so that the frequency of execution of “sit” is low in individuals with low characteristics or individuals with a high whim. In addition, in an individual with a high degree of hunger resistance, if an input signal indicating the hunger state is separately received, the appearance rate of the action pattern in response to the instruction input of “sitting” can be set to be low.

As will be described later, by learning the behavior pattern taken according to the past input pattern in the individual, the past learning result is added to the appearance rate of the behavior pattern derived from the characteristic value table and the correlation table. The appearance rate of the behavior pattern can be determined in consideration of the addition.

FIG. 10 shows the function of a device constituting an artificial animal to which the present invention is applied. This device includes an input unit 2, an input result processing unit 3, an output generation unit 4, an output unit 5, and the like, in addition to an action pattern determination unit 1 for realizing a movement of an individual. Behavior pattern determination unit 1
Is an arithmetic circuit 12 mainly composed of a CPU and a memory unit 1
3 (shown in FIGS. 11 and 12). The data processing functions of the input result processing unit 3 and the output generation unit 4 are also set in the arithmetic circuit 12.

The input unit 2 is a means for generating a predetermined input signal such as a switch, a sensor, a microphone, and the like. The input result processing unit 3 includes hardware such as an A / D conversion circuit, a filter for removing noise, and an interface unit. The input result processing unit 3 performs a pattern matching process on the input signal processed by these circuits, for example. To get the input pattern.

The output unit 5 causes the device to perform a predetermined operation, such as an actuator for moving a joint, a speaker for outputting a cry, a display for displaying an image of a character, a lamp for decoration, and the like. It consists of. The output generation unit 4 implements the behavior pattern determined by the behavior pattern determination unit 1,
This is for directly controlling the operation of the output unit 5.

The behavior pattern determination section 1 includes an operation section 6 for executing specific processing, an individuality storage section 7 for storing "individuality" peculiar to each individual in association with each other, and an action pattern for a past input pattern. And a learning unit 8 for learning the history of. The calculation unit 6 determines the execution of a predetermined operation pattern by processing the input pattern given from the input result processing unit 3 based on the personality set in the personality storage unit 7 and the history data of the learning unit 8. .

The individuality storage unit 7 incorporates a characteristic value table, correlation data, and the like imported from an external programming device. On the other hand, the history data of the learning unit 8 is set to an empty state in an initial state, and data is sequentially accumulated by operating this individual.

FIG. 11 shows a hardware configuration of the artificial animal and the programming device. Artificial animal computing device 10
(Hereinafter, simply referred to as “arithmetic device 10”) includes an input interface unit 14, an output driver unit 15, a communication interface circuit 16, and the like, in addition to an arithmetic circuit 12 and a memory unit 13 mainly including a CPU. In the memory unit 13, a storage area is set for each of the characteristic value table, the correlation data, and the history data of the behavior pattern, and a storage area (not shown) for a program related to control of the arithmetic circuit 12 is set. The input interface unit 14 and the output driver unit 15 have a hardware configuration included in the input result processing unit 3 and the output generation unit 4 of FIG. 10, respectively, and are connected to the input unit 2 and the output unit 5, respectively.

The programming device 11 is composed of a personal computer or the like, and includes an input unit 20, a display unit 21, and interface circuits 22 and 23 for these in addition to an arithmetic circuit 18 and a memory unit 19 mainly composed of a CPU. , A communication interface circuit 24, and the like. In the memory unit 19, a storage area of a program necessary for a process of transplanting an individual to an artificial animal, a work area of the program, and the like are set. The input unit 20 is used for inputting characteristic value setting data and the like described later.
It consists of a keyboard and a mouse. Display 21
Is composed of a CRT or a liquid crystal display device,
It is used for displaying a screen for inputting setting data and a display screen for a simulation result described later.

Each of the communication interface circuits 16 and 24 of the arithmetic unit 10 and the programming unit 11 having the above-described configuration.
Are communicably connected via connectors 17 and 25. As shown in FIG. 13A, the programming device 11 creates a characteristic value table of the artificial animal using the arithmetic circuit 18 and transmits the characteristic table to the arithmetic device 10. The arithmetic circuit 12 of the arithmetic device 10 stores the characteristic value table received from the programming device 11 in the memory unit 1 in its own device.
Write 9

FIG. 12 shows another example of the configuration of the arithmetic unit 10 and the programming unit 11. In this example, instead of the connectors 17 and 25 connecting the devices and the communication interface circuits 16 and 24, the programming device 11
On the side, an interface circuit 28 for memory read / write is incorporated. The interface circuit 28 is connected to an external memory read / write device (not shown) via the pin connector 26. In this configuration, as shown in FIG.
The information is written in the memory chip 13A before being mounted on the memory chip 0, and then the chip 13A is mounted. When this method is used, the memory chip 13A
In addition, not only the characteristic value table and the correlation data, but also a control program are written in the table. In the above configuration, the written data cannot be changed after the memory is mounted, but is suitable for producing a large number of artificial animals.

The characteristic value table thus transplanted to the artificial animal acts on the input pattern obtained from the input unit 2 to cause the artificial animal to take an action according to the personality, as shown in FIG. Works like that. FIG.
The functions set in the calculation unit 6 are shown. In the figure, an input pattern acquisition unit 29 is for receiving an input pattern obtained by the input result processing unit 3 and a specific value extraction unit 31.
Extracts an arbitrary characteristic value from the characteristic value table 32 for the characteristic item corresponding to the input pattern. The conversion processing unit 30 converts the input pattern into a predetermined behavior pattern based on the correlation data for the characteristic value. The conversion result output unit 33 takes in the conversion result and supplies it to the output generation unit 4.

Next, with reference to FIGS. 16 to 18, a procedure of processing performed until the artificial animal receives a predetermined signal input and reacts will be described. FIG. 16 shows the processing procedure of the input signal, which corresponds to the processing in the input result processing unit 3 in FIG. In the first ST1, an input signal from the input unit 2 (which has already been subjected to digital conversion processing) is received, and in the next ST2, signal noise is removed by filtering processing. Here, when a plurality of input signals are given, the processes of ST1 and ST2 are individually performed for each signal.

In the next ST3, the input pattern is determined by comparing the type, strength, and combination of the input signal with a predetermined model pattern. Note that the input pattern determination method is not limited to the matching process, and fuzzy inference or the like can be used.

FIG. 17 shows a procedure for determining an action pattern by the operation unit 6 shown in FIG. First, after acquiring the input pattern obtained by the procedure of FIG. 16 in the first ST4, in the next ST5, the characteristic value corresponding to the input pattern is obtained by using the function of the characteristic value extracting unit 28 shown in FIG. Extract.

Next, in ST6, an action pattern corresponding to the input pattern and its appearance probability are extracted based on the correlation data for the extracted characteristic values. Further S
At T7, the history of the action pattern for the same input pattern is read from the history data learned in the past, and ST
At 8, an action pattern to be finally executed is determined. In ST8, for example, the extraction result in ST6 and ST7
Then, if an action pattern different from the extraction result is learned, a process of selecting the learned action pattern is performed.

FIG. 18 shows a procedure of processing performed by the output generation unit 4 after the action pattern is determined. First, ST9
Then, when the action pattern determined by the arithmetic unit 6 is received, in step ST10, the contents of a specific instruction for executing the action pattern are determined. Note that what is determined here is the part to which the drive signal is to be given and the type of the signal, and in ST11, the output unit outputs a control signal based on the determination to execute the action pattern.

FIG. 19 shows a processing procedure by the learning unit 8. This process is executed in conjunction with the processes in FIGS. As shown in FIG. 10, the input result processing unit 3 converts the input pattern obtained from the input signal into the operation unit 6.
Also output to the learning unit 8 at the same timing as. Learning unit 8
When the input pattern is received in ST12 of FIG. 19, the history data stored in the past (refers to the history of the action pattern output with respect to the input pattern) in response to a request from the arithmetic unit 6. Is output (ST13,
14).

In the next step ST15, the output generating section 4 fetches the action pattern in conjunction with the operation of receiving the action pattern from the arithmetic section 6. In the next ST16, the acquired behavior pattern is stored as new history data in association with the input pattern.

Next, the function and processing of the programming device 11 will be described in detail. FIG. 20 shows functions set in the arithmetic circuit 18 in the programming device 11. In this embodiment, as the information for setting the characteristic value table, the input unit 20 sets various characteristic items included in the characteristics of the population and characteristic value distribution setting data for each characteristic item (see FIG. 3). Standard value, distribution range, selection probability). When performing a simulation using the set characteristic value table, calculation conditions for the simulation are input.

In the figure, a characteristic item acquiring section 35 is a section for receiving an input about the characteristic item, and a characteristic value distribution setting section 36 receives the setting data of the characteristic value distribution, and This is for setting dimensional data. The population tendency storage unit 38 is for creating data indicating a population tendency by associating each characteristic item with a characteristic value distribution and storing the data in the memory unit 19.

The characteristic value extracting unit 39 randomly extracts characteristic values included in the population for each characteristic item from the population tendency by a lottery process using random numbers. Here, a plurality of combinations of characteristic values are extracted for the population.
The characteristic value storage unit 40 has a function of storing each extracted combination in the memory unit 19. The transplant processing unit 41 creates a characteristic value table by calling a predetermined combination of characteristic values in the fictitious space with respect to an arbitrary individual, and transplants this to the memory unit of the individual.

The calculation conditions for the simulation are as follows:
For example, it refers to a desirable appearance probability when an arbitrary behavior pattern appears for an arbitrary input pattern in a population. The calculation condition acquisition unit 37 receives the calculation condition input from the input unit 20 and outputs the calculation condition to the output tendency calculation unit 42 and the display unit 21 corresponding to the population. The population-based appearance tendency calculation unit 42 calculates, for the set population, the appearance tendency of the behavior pattern with respect to the pseudo-given input pattern based on the correlation data, and calculates the calculation result by the calculation result. Compare with condition. On the other hand, the output tendency calculating unit 43 corresponding to the individual stores the characteristic value storage unit 40.
Is calculated based on the characteristic value table set in (1). The calculation results of the output tendency calculation units 42 and 43 are output to the display unit 21 via the result output unit 44.

The display section 21 outputs, in addition to the calculation results, the calculation conditions, the characteristic value distribution indicating the population tendency, the setting contents of the characteristic value table, and the like. That is, on the display unit 21, the appearance tendency of the behavior pattern in the population with respect to the predetermined input pattern and the appearance tendency of the behavior pattern in one individual are displayed in a comparable state, and are set for each characteristic value distribution and individual. The contents of the specified characteristic value table are displayed. The operator refers to these displays and determines whether or not the behavior pattern taken by the entire population and each individual is appropriate as the behavior taken by the original animal (eg, dog). If there is an inappropriate behavior pattern, the setting of the population tendency is corrected so that the personality that causes the behavior pattern to appear is not included in the population tendency.

FIG. 21 shows the principle of correcting the population tendency based on the simulation. When the appearance tendency of the behavior pattern in the population does not match the behavior of the animal indicated by the population, it is necessary to remove the individuality that causes the undesirable behavior pattern to appear from the population tendency. Since a correlation is set between the various input patterns and the behavior patterns and the characteristic values, the characteristic items to be deleted are determined based on the relationship between the undesired behavior patterns and the input patterns that have caused the behavior. In addition, the range of the characteristic value to be deleted (the area indicated by oblique lines in the figure) can be easily specified. In addition, from the characteristic value table of each individual and the appearance tendency of the behavior pattern of the individual, it is possible to extract an individual having an individuality that takes an undesirable behavioral pattern, so that the individuality of the individual can be easily corrected. . In this way, by providing an appropriate population tendency to the set population tendency by feeding back the appearance tendency of the behavior pattern appearing according to the population tendency, the characteristics included in the population tendency are set. Individuality of each individual can be determined by a combination of values, and various individualities can be set without departing from the population.

FIG. 22 shows the functions that need to be set in the programming device 11. The functions of the programming device 11 can be roughly classified into four functions: a function of setting individuality of each individual, a function of verifying an output result by each individuality (simulator function), a function of resetting individuality, and a function of transplanting individuality. . Although not shown, it is also possible to set a function of sucking up individuality from each individual.

The function for setting the individuality includes a function for setting the characteristic value distribution for each characteristic item based on the setting data input after the operator has decomposed the characteristics of the population to be processed for each item (characteristic setting). Function), a function of extracting characteristic values for setting individuality from the respective characteristic value distributions, and a function of creating the correlation data by associating the extracted characteristic values with input / output patterns. It is necessary to transplant at least the characteristic value table and the correlation data among the above data as data for specifying the tendency of the individual to appear in each artificial animal.

The output result simulator function includes a function of receiving and setting the above-mentioned calculation condition input and a function of calculating the tendency of the behavior pattern to appear in the population and each individual. The individuality set by the individuality setting function is corrected and rewritten according to the processing result of the simulator function by the function of the individuality resetting function. Thereafter, the rewritten personality is transferred to the memory unit 13 of each individual by the function of the personality transfer function.

Next, a procedure from setting of personality to transplantation by the programming device 11 will be described with reference to FIGS. The steps in each figure are the same as those in FIGS.
"St" to distinguish it from the processing procedure on the artificial animal side. Further, “other processing” in each drawing represents waiting for data input or the like, but is not particularly mentioned in the following description.

First, as the first processing, the item name of the characteristic item is input from the input unit 20. When this input is accepted, st1 in FIG. 23 becomes "YES" and the process proceeds to st3, and the input item name is stored in the population tendency setting area in the memory unit 19. In further subsequent st4, a new characteristic value axis corresponding to the input characteristic item is set in the memory unit 19.

The set characteristic item may be deleted in some cases. When this deletion instruction is received, st5 in FIG.
Becomes "YES" and proceeds to st7, where a list of characteristic item names registered in the display unit 21 is displayed. When a characteristic item to be deleted is selected for this list display, s from st8
Proceeding to t10, the selected characteristic item name and the characteristic axis corresponding to the selected item are deleted from the memory unit 19.

Next, when it is instructed to set the characteristic value distribution for the set characteristic item, st1 in FIG.
1 becomes "YES" and proceeds to st13, where a list of characteristic item names registered in the display unit 21 is displayed. When a characteristic item to be set is selected for this list display, st1
From 4 to st16, the characteristic axis corresponding to the characteristic item is read from the memory unit 19 and set in an imaginary work space. Further, when the standard value, the range of the characteristic value distribution, the selection probability, and the like are input for this characteristic item, the process proceeds from st17 to st19 to create a characteristic value distribution based on the data input on the characteristic axis. , Stored in the memory unit 19. Thereafter, the processing of st14 to st19 is repeated until the end of setting is instructed in st20.

Next, when an instruction to generate individuality is given, FIG.
6 st21 is "YES" and the process proceeds to st23, and the display unit 21 displays a list of registered population names in a list. When the population to be processed is selected from the list display, the process proceeds from st24 to st26, where the population tendency of the selected population is read out, and then the characteristic value distribution is obtained from each characteristic value distribution included in this population tendency. Are randomly extracted one by one. As described above, this extraction process is performed in a plurality of ways. In the next st27, a characteristic value table in which each characteristic value is set as one individual characteristic is created for each extracted unit and stored in the memory unit 19. To save.

Next, the operation shifts to a simulation operation.
When an instruction for an operation for the population is issued, FIG.
Is "YES" and the process proceeds to st30, where the display unit 21 displays a list of registered population names. When a population to be processed is selected for this list display, the process proceeds from st31 to st33, and the memory unit 19
The characteristic value distributions forming the population tendency of the population selected from the population are read. In the next st34, correlation data indicating the correlation between each characteristic value included in these characteristic value distributions and the input pattern and the action pattern is read.

In the next st35 and st35, in accordance with the input of the calculation condition from the input unit 20, the condition is obtained, and the following calculation process is executed. First, in st37, a standard value of each characteristic value distribution is extracted, and in the next st38, one characteristic item to be calculated is extracted. In st39, first, the occurrence probabilities of various behavior patterns with respect to various input patterns are calculated from the correlation between the characteristic values corresponding to the standard values and the input patterns and the behavior patterns for the characteristic items. That is, the appearance probabilities of the action patterns that appear when an individual having the standard characteristic value is given a predetermined input pattern are individually calculated.

When the above operation is completed, the process proceeds to st40, where each operation result is compared with the operation condition. In the next st41, each calculation result is stored in the memory unit 19 in association with the selection probability of the characteristic value under consideration and the comparison result.

Thereafter, the characteristic values of the characteristic item of interest are shifted by one unit within the range of the distribution (st42, 45), and the processing of st39 to 41 is executed for the characteristic values in order. Further, when the calculation process for the characteristic item under attention is completed, the target of interest is changed to the unprocessed characteristic item (st4).
2 to 44), the same processing is repeated. When the calculation for all the characteristic items is completed, the process proceeds from st43 to st46, and the calculation results are integrated and stored in the memory unit 19.

Next, when a simulation for an individual is instructed, st47 in FIG.
Proceeding to t49, each personality registered by the processing of FIGS. When a predetermined personality is selected for this list display, st50 to st5
Proceeding to 2, the respective characteristic values constituting the individuality selected from the memory unit 19 are read. Further, in the next st53, the correlation between these characteristic values and the input pattern and the behavior pattern is read, and in the following st54, the occurrence probabilities of various behavior patterns with respect to the various input patterns are calculated based on the correlation. When the operation is completed for all combinations of input / output patterns, the process proceeds to st55, and each operation result is stored in the memory unit 19 together with the characteristic value selection probability (population selection probability) used for the operation. I do.

When the output of each calculation result is instructed after the simulation processing, "st56" in FIG. 29 becomes "YES", and a menu of the display method is displayed first in st58. When a predetermined display method is selected from this menu,
Proceeding from st59 to st61, each data is processed so that the calculation results obtained in the processes of FIGS. 27 and 28 are displayed by the selected display method. In st62,
The processed data is displayed.

The operator corrects the population tendency while referring to this display. For example, if any of the behavior patterns whose appearance probabilities for the various input patterns are calculated according to the procedure of FIG. The population tendency is corrected by deleting the characteristic value to be performed from the population tendency or changing the selection probability of the characteristic value. Further, the operator specifies an individual who takes the behavior pattern targeted for the correction from the appearance tendency of the behavior pattern in each individual obtained in the procedure of FIG. 28, and performs a process of correcting the individuality of the individual.

When it is instructed to transplant the personality set in this way to a predetermined artificial animal, st63 in FIG. 30 becomes “YES”, and the display unit 21 displays each target to be transplanted. A list of personalities is displayed (st65). When a predetermined personality is selected from this list display, s from st66
Proceed to t68. In st68, for the selected personality, after reading the characteristic value table and the correlation data,
By outputting these data to the memory unit 13 of the target individual, the individual is transplanted to the individual.

By the way, according to the above-mentioned individuality setting method,
Since the debugging after the individuality is transplanted to the artificial animal becomes unnecessary, the characteristic value table showing the individuality of each individual can be incorporated into the chip constituting the arithmetic unit of the individual.

FIG. 31 shows a configuration example of the arithmetic circuit 12A in which the characteristic value table is incorporated. In the figure, CPU5
0, address register 51, input / output buses 52, 53
This is a normal configuration applied to the arithmetic circuit 12 having the configuration shown in FIGS. In addition to these configurations, the arithmetic circuit 12A of this embodiment incorporates an eigenvalue section 54 and an A / D conversion circuit 55.

The eigenvalue section 54 is a section in which analog data indicating the contents of the characteristic value table is set. The setting data is digitally converted by the A / D conversion circuit 55 and provided to the CPU 50.

FIG. 32 shows a specific example of a method of forming the arithmetic circuit 12A. In this embodiment, the silicon wafer 57
A characteristic value table having a different configuration is set for each of a plurality of chips 56 which are separated from each other, and mask data obtained by converting a digital code of each characteristic value in the characteristic value table into an analog signal is written to each chip 56 to thereby obtain the unique value. The unit 54 is constituted. Note that the logic section 60 in the figure includes a configuration other than the eigenvalue section 54 in FIG. 31, and all the chips are configured in the same manner. The above chip 56
Is mounted on the control board 59 in a state sealed in the resin case body 56.

According to the above-described configuration, an artificial animal having a different personality can be created between individual animals only by incorporating a chip into each individual animal, and the man-hour and cost for creating an artificial animal can be greatly reduced. can do. It will also be possible to sell this chip alone as "artificial intelligence with personality."

[0082]

As described above, according to the present invention, a plurality of individuals belonging to a population are given a response to an input signal based on a combination of characteristic values randomly extracted from the range of characteristics of the population. Since the calculation rules for determining the signal are set, each individual must be set to have a personality that does not deviate from the range of characteristics set for the population, and present various responses to the same input signal. Can be. In addition, since the individuality of each individual is set after setting the range of characteristics in the population, there is no need to repeat debugging for each individual, greatly reducing the labor and cost of manufacturing arithmetic devices for artificial animals. Can be.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the concept of population tendency.

FIG. 2 is a diagram showing a comparison between a characteristic item and a setting standard of a characteristic axis for each item.

FIG. 3 is a diagram showing the principle of a method for setting a characteristic value distribution.

FIG. 4 is a diagram showing an example in which a characteristic value distribution is differentiated.

FIG. 5 is a diagram illustrating that a characteristic value distribution differs depending on a population.

FIG. 6 is a diagram illustrating a data configuration of a population tendency.

FIG. 7 is a diagram illustrating a specific example of setting of individuality.

FIG. 8 is a diagram illustrating a correlation between a characteristic value, an input pattern, and an action pattern.

FIG. 9 is a diagram specifically illustrating the correlation of FIG. 8;

FIG. 10 is a functional block diagram showing a configuration example of an artificial animal to which the present invention is applied.

FIG. 11 is a block diagram illustrating a hardware configuration of an artificial animal and a programming device.

FIG. 12 is a block diagram illustrating a hardware configuration of an artificial animal and a programming device.

FIG. 13 is a diagram illustrating a method of implanting a characteristic value table into an artificial animal.

FIG. 14 is a diagram showing a principle for causing an artificial animal to act on an input pattern.

FIG. 15 is a functional block diagram illustrating a detailed configuration of a calculation unit in FIG. 10;

FIG. 16 is a flowchart illustrating a processing procedure of an input signal in an artificial animal.

FIG. 17 is a flowchart showing a procedure for determining a behavior pattern in an artificial animal.

FIG. 18 is a flowchart showing a procedure for generating an output instruction for executing a behavior pattern in an artificial animal.

FIG. 19 is a flowchart showing a control procedure using a learning function in an artificial animal.

FIG. 20 is a functional block diagram showing a configuration set in an arithmetic circuit on the programming device side.

FIG. 21 is a diagram illustrating a principle of correcting a population tendency by a simulation process.

FIG. 22 is a diagram illustrating functions set in the programming device.

FIG. 23 is a flowchart showing a procedure for receiving an input of a characteristic item in the programming device.

FIG. 24 is a flowchart showing a procedure when a characteristic item is deleted in the programming device.

FIG. 25 is a flowchart showing a procedure of a characteristic value distribution setting process in the programming device.

FIG. 26 is a flowchart showing a characteristic value extraction processing procedure in the programming device.

FIG. 27 is a flowchart showing a procedure for calculating an appearance tendency of a behavior pattern in a population in the programming device.

FIG. 28 is a flowchart showing a procedure for calculating an appearance tendency of an action pattern in an individual in the programming device.

FIG. 29 is a flowchart showing a procedure for outputting a calculation result in the programming device.

FIG. 30 is a flowchart illustrating a procedure of a personality transfer process in the programming device.

FIG. 31 is a block diagram illustrating a configuration of an arithmetic circuit in which a characteristic value table is incorporated.

FIG. 32 is a diagram illustrating a method of creating the arithmetic circuit in FIG. 31.

FIG. 33 is a diagram illustrating a data configuration related to the setting of personality in the related art.

FIG. 34 is a diagram illustrating a data configuration related to the setting of personality in the related art.

FIG. 35 is a diagram illustrating a conventional data configuration related to setting of personality.

FIG. 36 is a diagram illustrating a conventional method of correcting individuality.

FIG. 37 is a diagram illustrating functions set in a conventional programming device.

FIG. 38 is a diagram illustrating a problem in a conventional method of setting personality.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Behavior pattern determination part 2 Input part 3 Input result processing part 4 Output generation part 5 Output part 10 Arithmetic unit 11 Programming unit 12, 12A Arithmetic circuit 18 Arithmetic unit 13 Memory part 13A memory 20 Input part 24 Communication interface circuit 28 Memory read / write Interface circuit 32 characteristic value table 38 population tendency storage unit 39 characteristic value extraction unit 40 characteristic value storage unit 41 transplantation processing unit 42 output tendency calculation unit 44 result output unit

Claims (4)

[Claims] 1. A method for setting, to a plurality of individuals responding to a predetermined input signal in a predetermined manner determined by arithmetic processing, an operation rule for reacting the same input signal in a different manner depending on the individual. After setting the distribution of characteristic values in the population for each of a plurality of types of characteristics as a range of characteristics of the population of each individual, for each individual, the characteristic values included in the population for each of the characteristics are set. A method for setting an operation rule, characterized in that operation rules for a plurality of types of input signals are set based on a combination of extracted characteristic values at random. 2. An apparatus for transplanting operation rules for generating different reaction signals for the same input signal into a plurality of operation apparatuses that generate and output a reaction signal corresponding to a predetermined input signal by operation processing. Setting data input means for receiving an input of data for setting a distribution state of characteristic values in the population for each of a plurality of types of characteristics as setting data indicating a range of characteristics of the population of each arithmetic device; A storage unit for storing the setting data received by the data input unit, and for each arithmetic unit, randomly extracting characteristic values included in a population for each characteristic from the storage unit based on a combination of the extracted characteristic values. Calculation rule generation means for generating calculation rules for a plurality of types of input signals; and a calculation rule generated by the calculation rule generation means in a memory of each calculation device. Calculation rule implant device formed by and a calculation rule output means for writing. 3. The operation rule transplanting device according to claim 2, wherein an output pattern of a reaction signal that a population can output with respect to a predetermined input signal is shown based on the operation rule and setting data in a storage unit. Reaction pattern output means for creating and outputting data to the outside, correction data input means for accepting input of data for correcting setting data in the storage means, and correction after data output by the reaction pattern output means An operation rule transplantation device comprising: a setting data rewriting means for receiving input of data and rewriting setting data in the storage means. 4. An artificial animal, comprising: an arithmetic unit that generates and outputs a response signal to a predetermined input signal through arithmetic processing; and execution means that executes an action pattern according to the response signal. The calculation rule incorporated in the device is that, after setting the distribution of characteristic values in the population for each of a plurality of types of characteristics as a range of characteristics of the population of artificial animals, the characteristic values included in the population for each characteristic are set. An artificial animal that is set based on a combination of characteristic values obtained by random sampling.