JP2001334482A - Robot device and method of determining action of robot device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a robot device having improved a feeling of a living thing for performing animal-like action. SOLUTION: An action determining system 70 comprises a perception information acquisition part 90 and a motivation information acquisition part 81 as external or internal information detected by a CCD camera 20, a distance sensor 22, a microphone 23, etc., for acquiring factors as information which gives influence to action and acquiring tendencies to development of actions influenced by the factors in accordance with the acquired factors, an action selection arithmetic part 82 for comparing the tendencies to development corresponding to two or more actions acquired by the perception information acquisition part 90 and the motivation information acquisition part 81, as the tendencies to development in the same group, and selecting one of the actions in accordance with the comparison result of the tendencies, and an output semantics converter module 68 for controlling an operation part in accordance with the action selected by the action selection arithmetic part 82 to make the selected action develop.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot device that behaves autonomously and a method for determining the behavior of a robot device that determines the behavior of such a robot device.

[0002]

2. Description of the Related Art In recent years, a robot device having a shape imitating an animal, that is, a so-called pet robot has been proposed. Such a robot device has a shape similar to a dog or a cat bred in an ordinary household, and is autonomous in response to a user's (owner's) action such as "slapping" or "stroking", and the surrounding environment. To act on. For example, as an autonomous action, an action such as “barking” or “sleeping” is performed in the same manner as an actual animal.

By the way, if the robot device can approach the behavior of an actual animal more, the biological feeling of the robot device is further increased, and the user (owner) feels a greater sense of closeness and satisfaction with the robot device. . Thereby, the amusement of the robot device is improved.

[0004] For example, it is conceivable to determine the behavior of a robot device from an ethological approach in order to make a behavior like an actual animal behave.

[0005] The results of behavioral studies made from an animal behavioral approach include, for example, the behaviorist Sibly, Mcfarl.
and et al. published a paper on the state of motivational spatial representation in 1975 ("Animal Communication", Nishimura Shoten). Ludlow also announced a competing model of behavior in 1976. And these achievements are Bruce Mitchell Blumb
erg (Bruce of Arts, Amherst College, 1977, Master of S
iencs, Sloan School of Management, MIT, 1981)
ricks, New Dogs: Ethology and Interactive Creature
s "(April 1997). Bruce Mitch
ell Blumberg applies the above-mentioned theory to 3D CG (computer graphic) dogs and verifies it as a behavior selection mechanism.

[0006]

[Problems to be solved by the invention] By the way, Bruce Mitc
Hell Blumberg has verified the behavior selection mechanism of animals with a CG model, and is not adapted to robotic devices existing in the real world.

In a virtual creature displayed on a display of a computer system such as a computer graphics (CG), it is possible to directly connect action selection and action realization (action selection = action realization). However, in an actual robot, action selection and action realization cannot always be directly linked (that is, action selection = action realization is not always possible). This is because of the following reasons.

[0008] Actions that are performed independently of planned actions, such as reflex actions, can be counteracted.

[0009] Unless feedback from input from a sensor is used, there is a case where it is impossible to know whether the action has been realized in a true sense.

[0010] For the above reason, as a specific example of the latter, the action of "kicking the ball with the foot" is selected when the distance from the ball reaches a distance at which the ball can be kicked. Even if is output, there is a case where the ball is not kicked if the ball is placed on a slope or the like. In order to recognize "the ball was kicked" as a result of the action of "kicking the ball with the foot", the ball and the robot device came into contact, recognizing that the ball advanced forward,
It is possible for the first time to be "kicked". That is, it is necessary to evaluate the behavior using the sensor information of the sensor of the robot device and change the internal state of the robot device.

Thus, Bruce Mitche as described above
The technology proposed by ll Blumberg alone cannot determine the behavior of robotic devices that exist in the real world.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide a robot apparatus having a more improved biological feeling and a method of determining the behavior of the robot apparatus.

[0013]

SUMMARY OF THE INVENTION A robot apparatus according to the present invention is a robot apparatus for controlling an operating unit to cause an action to appear, and for detecting external or internal information. Means, external or internal information detected by the detecting means, a causal factor obtaining means for obtaining a causal factor which is information affecting behavior, and a causal factor based on the causal factor obtained by the causal factor obtaining means. Appearance tendency acquisition means for acquiring the appearance tendency of an action affected by a factor, and appearance tendency corresponding to two or more actions acquired by the appearance tendency acquisition means, wherein the appearance tendency is regarded as being in the same group. Based on the comparison result of the tendency of appearance by the tendency comparison means and the appearance tendency comparison means, an action selection means for selecting one action, and controlling the operation unit based on the action selected by the action selection means, And a operation section control means for the appearance of the selected action, the appearance tends one behavioral action selection means has selected changes according to the causative factors that change the actual appearance of the action.

A robot device having such a configuration is as follows.
The external or internal information is detected by the detecting means, and the external or internal information detected by the detecting means is acquired by the causal factor acquiring means as a causal factor which is information affecting behavior,
Based on the causal factors acquired by the causal factor acquiring means, the appearance tendency of the behavior affected by the causal factors is acquired by the appearance tendency acquiring means.

The robot apparatus compares the appearance trends corresponding to the two or more actions acquired by the appearance tendency acquisition unit, which are considered to be in the same group, by the appearance tendency comparison unit. Based on the comparison result of the appearance tendency according to the above, one action is selected by the action selecting means, and based on the action selected by the action selecting means, the operation section is controlled by the operation section control means, and the selected action is Let it appear. Then, the appearance tendency of one action selected by the action selection means changes according to the causal factor that changes due to the actual appearance of the action.

[0016] Such a robot apparatus compares the appearance tendencies determined by the causal factors,
One behavior has been selected, and the behavior emerges as an ethological approach.

In order to solve the above-mentioned problems, a method for determining an action of a robot apparatus according to the present invention includes an information detecting step of detecting external or internal information of the robot apparatus by a detecting means, and detecting the information by an information detecting step. A causal factor acquiring step for acquiring causal factors which are information affecting the behavior of the robot device of the external or internal information, and an action affected by the causal factor based on the causal factors acquired in the causal factor acquiring step. An appearance tendency acquisition step of acquiring the appearance tendency of the two, and an appearance tendency comparison step of comparing the appearance tendency corresponding to two or more actions acquired in the appearance tendency acquisition step, which are considered to be in the same group; An action selection step of selecting one action based on the comparison result of the appearance trends in the appearance tendency comparison step, and a robot apparatus based on the action selected in the action selection step. Controlling the working unit to cause the robot device to cause the selected action to appear, and the action tendency of the one action selected in the action selecting step is determined by the actual appearance of the action. It changes according to changing causal factors.

[0018] In such a method for determining the behavior of a robot device, the detection means detects external or internal information, which is information external or internal to the robot device, in an information detecting step,
In the causal factor acquisition step, a causal factor that is information that affects the behavior of the robot device of the external or internal information detected in the information detection step is acquired, and based on the causal factor acquired in the causal factor acquisition step, The appearance tendency of the behavior affected by the causal factor is acquired in the appearance tendency acquisition step.

The method for determining the action of the robot apparatus is as follows.
Appearance tendencies corresponding to two or more actions acquired in the appearance tendency acquisition step, which are in the same group, are compared in the appearance tendency comparison step, and the results of the appearance tendency comparison in the appearance tendency comparison step are compared. Based on the action selected in the action selecting step, based on the action selected in the action selecting step, the operating unit of the robot device is controlled in the operating unit control step, and the selected action is selected by the robot apparatus. Make the action appear. Then, the appearance tendency of one action selected in the action selection step changes according to the causal factor that changes due to the actual appearance of the action.

According to such a method for determining the behavior of a robot device, the robot device selects one behavior by comparing the appearance tendency determined by being influenced by the causal factors. As a behavior comes to appear.

[0021]

Embodiments of the present invention will be described below in detail with reference to the drawings. In this embodiment, the present invention is applied to a robot apparatus that changes the state of instinct and emotion based on external factors and internal factors, and acts according to the state of instinct and emotion. .

In the embodiment, first, the configuration of the robot device will be described, and thereafter, the application portion of the present invention in the robot device will be described in detail.

(1) Configuration of Robot Apparatus According to the Present Embodiment As shown in FIG. 1, a so-called pet robot having a shape imitating a "dog" is provided, and leg units 3A, 3A, 3B, 3C, and 3D are connected, and a head unit 4 and a tail unit 5 are connected to the front end and the rear end of the body unit 2, respectively.

As shown in FIG. 2, a CPU (Central Processing Unit) 10 and a DRA
M (Dynamic Random Access Memory) 11, Flash ROM (Read 0nly Memory) 12, PC (Personal Co.)
A controller 16 formed by interconnecting a card interface circuit 13 and a signal processing circuit 14 via an internal bus 15 and a battery 17 as a power source of the robot apparatus 1 are housed therein. . The body unit 2 also houses an angular velocity sensor 18 and an acceleration sensor 19 for detecting the acceleration of the direction and movement of the robot device 1.

Also, the head unit 4 receives a charge (CCD) camera 20 for capturing an image of an external situation and a pressure applied by a physical action such as "stroke" or "hit" from the user. A touch sensor 21 for detection, a distance sensor 22 for measuring a distance to an object located ahead, a microphone 23 for collecting external sounds, and a speaker 24 for outputting a sound such as a squeal And an LED (Light Emitting Diode) (not shown) corresponding to an “eye” of the robot apparatus 1 are arranged at predetermined positions.

Further, the joints of the leg units 3A to 3D, the leg units 3A to 3D, and the trunk unit 2
, The connecting portion of the head unit 4 and the body unit 2 and the connecting portion of the tail 5A of the tail unit 5, etc.
5 _1, 25 _2, ..., and the potentiometer 26 _1, 26
₂ ,... Are provided. For example, each of the actuators 25 ₁ , 25 ₂ ,... Has a servomotor. By driving the servo motor, the leg unit 3A
3D is controlled to transit to a desired posture or motion.

The angular velocity sensor 18, acceleration sensor 19, touch sensor 21, distance sensor 22, microphone 23, speaker 24 and each potentiometer 2
The various sensors such as 6 ₁ , 26 ₂ ,... And the LEDs and the actuators 25 ₁ , 25 ₂ ,... Are connected to the control unit 16 via the corresponding hubs 27 _{1 to} 27 _n.
, And the CCD camera 20 and the battery 17 are directly connected to the signal processing circuit 14, respectively.

The signal processing circuit 14 sequentially takes in the sensor data, image data and audio data supplied from each of the above-described sensors, and sequentially stores them at predetermined positions in the DRAM 11 via the internal bus 15. In addition, the signal processing circuit 14 sequentially takes in remaining battery power data indicating the remaining battery power supplied from the battery 17 and stores the data at a predetermined position in the DRAM 11.

The sensor data, image data, audio data, and remaining battery data stored in the DRAM 11 in this manner are thereafter transmitted to the robot apparatus 1 by the CPU 10.
It is used when controlling the operation of.

Actually, at the initial stage when the power of the robot apparatus 1 is turned on, the CPU 10 executes the control program stored in the memory card 28 or the flash ROM 12 inserted in the PC card slot (not shown) of the body unit 2 into a P program.
The data is read out via the C card interface circuit 13 or directly and stored in the DRAM 11.

The CPU 10 then determines the status of itself and its surroundings and the user based on the sensor data, image data, audio data, and remaining battery data sequentially stored in the DRAM 11 from the signal processing circuit 14 as described above. Judgment of the instruction from the person and the presence or absence of the action.

Further, the CPU 10 determines a subsequent action based on the determination result and the control program stored in the DRAM 11, and drives the necessary actuators 25 ₁ , 25 ₂ ,... Based on the determination result. Thereby, actions such as swinging the head unit 4 up, down, left, and right, moving the tail 5A of the tail unit 5, and driving and walking each leg unit 3A to 3D are performed.

At this time, the CPU 10 generates audio data as necessary, and supplies the generated audio data to the speaker 24 as an audio signal via the signal processing circuit 14, thereby outputting an audio based on the audio signal to the outside. LED above
Is turned on, off or blinking.

In this way, the robot device 1 is capable of acting autonomously in accordance with the situation of itself and the surroundings, and instructions and actions from the user.

(2) Software Configuration of Control Program The software configuration of the above-described control program in the robot device 1 is as shown in FIG. This figure 3
, The device driver layer 30 is located at the lowest layer of the control program and includes a device driver set 31 including a plurality of device drivers. In this case, each device driver is CC
An object that is allowed to directly access hardware used in a normal computer, such as the D camera 20 (FIG. 2) and a timer, and performs processing in response to an interrupt from the corresponding hardware.

The robotic server object 32 is located at the lowest layer of the device driver layer 30 and includes, for example, the various sensors and actuators 2 described above.
5 _1, 25 _2, the virtual robot 33 comprising a software group that provides an interface for accessing the hardware, such as ..., and Ba word manager 34 made of a software suite for managing the switching of power supply,
It is composed of a device driver manager 35 which is a software group for managing various other device drivers, and a designed robot 36 which is a software group for managing the mechanism of the robot apparatus 1.

The manager object 37 includes an object manager 38 and a service manager 39.
It is composed of In this case, the object manager 38 has the robotic server object 32,
Middleware layer 40 and application
The service manager 39 is a software group that manages activation and termination of each software group included in the layer 41. The service manager 39 stores connection information between objects described in a connection file stored in the memory card 28 (FIG. 2). Is a group of software that manages the connection of each object based on.

The middleware layer 40 is located on the upper layer of the robotic server object 32 and is composed of a group of software for providing basic functions of the robot apparatus 1 such as image processing and sound processing. I have. Further, the application layer 41 is located above the middleware layer 40, and
It is composed of a software group for determining an action of the robot apparatus 1 based on a processing result processed by each software group constituting the layer 40.

FIG. 4 shows a specific software configuration of the middleware layer 40 and the application layer 41, respectively.

In the middleware layer 40,
As shown in FIG. 4, each signal processing module 50 for noise detection, temperature detection, brightness detection, scale recognition, distance detection, attitude detection, touch sensor, motion detection, and color recognition A recognition system 60 having an input semantics converter module 58 and an input semantics converter module 59; and an output semantics converter module 68 and an attitude management, tracking, motion reproduction, walking, fall recovery,
Each signal processing module 61 for LED lighting and sound reproduction
And a recognition system 60 having -67 and the like.

Each of the signal processing modules 50 to 50 of the recognition system 60
Numeral 58 fetches corresponding data among the sensor data, image data, and audio data read from the DRAM 11 (FIG. 2) by the virtual robot 33 of the robotic server object 32, and performs predetermined processing based on the data. Is given to the input semantics converter module 59. Here, for example, the virtual robot 33 is configured as a part that exchanges or converts signals according to a predetermined communication protocol.

The input semantics converter module 59 determines "noisy", "hot", "bright", "ball detected", "fall" based on the processing results given from each of the signal processing modules 50-58. Self and surrounding conditions such as "detected", "stroked", "hitted", "hearing the scale of Domiso", "detected moving object" or "detected obstacle", and the user , And outputs the recognition result to the application layer 41 (FIG. 2). In the application layer 41, an action determination system for action determination to be described later is constructed.

On the other hand, in the output system 69, the output semantics converter module 68 outputs
Each of the signal processing modules 61 to 67 is controlled. That is, for example, according to the recognition result of the recognition system 60, “noisy”, “hot”, “bright”, “ball detected”,
"Detected fall,""stroke,""beaten,"
"I heard the domeso scale,""Detected a moving object."
Or, it outputs control information (commands) to each of the signal processing modules 61 to 67 in response to the self and surrounding conditions such as “detected an obstacle”, commands and actions from the user, and the like.

The behavior information input to the output semantics converter module 68 includes “forward”, “please”,
"Squeal" or "tracking (chasing the ball)"
And the output semantics converter module 68 gives the action commands to the corresponding signal processing modules 61 to 67. The behavior information input to the output semantics converter module 68 is from the behavior determination system that is the upper information processing unit, but is a main part according to the present invention, and will be described later in detail.

The signal processing modules 61 to 67
Outputs a control signal for controlling each device to the virtual robot 33 based on the action command from the output semantics converter module 68. Specifically, when an action command is given, the signal processing modules 61 to 67 perform the corresponding actuators 25 ₁ , 25 ₂ ,... To perform the action based on the action command.
..Servo command value to be given to (FIG. 2) or speaker 2
4 (Fig. 2) and / or "eye"
, And the corresponding actuators 25 ₁ , 25 ₂ ,... Sequentially through the virtual robot 33 of the robotic server object 32 and the signal processing circuit 14 (FIG. 2). Alternatively, the data is sequentially transmitted to the speaker 24 or the LED.

Each device is controlled based on a signal (command) from the virtual robot 33, so that the robot apparatus 1 makes a predetermined action appear.

Next, the next action (transition action or target action) is determined based on the recognition result from the input semantics converter module 59, and information on the determined action is output to the output semantics converter module 68.
Will be described.

(3) Configuration of Behavior Determination System of Robot Device The robot device 1 determines a behavior by a behavior determination system 70 as shown in FIG. Action decision system 70
Determines an action based on the recognition result from the input semantics converter module 59, and outputs the action to the output semantics converter module 68. Specifically, the robot device 1 includes an action selecting unit 80,
An internal state model unit 71 and a modulator 72 are provided.

The action selecting section 80 selects a desired action from a plurality of actions. Specifically, the action selecting unit 80 selects a desired action based on the recognition result of the input semantics converter module 59. For example, the action selection unit 8
Numeral 0 includes a perception information acquisition unit 90, a motive information acquisition unit 81, and an action selection calculation unit 82, and selects an action.

In such an action determination system 70, the perceptual information acquiring unit 90 and the motive information acquiring unit 81
Causal factor obtaining means for obtaining external or internal information detected by the detecting means for detecting external or internal information, such as the CD camera 20, the distance sensor 22, the microphone 23, etc., and which is assumed to be information affecting behavior; and Based on the causal factor acquired as the causal factor acquiring unit, an appearance tendency acquiring unit configured to acquire an appearance tendency of an action affected by the causal factor is configured.
And an appearance tendency corresponding to two or more actions acquired by the motivation information acquisition unit 81, based on an appearance tendency comparison unit that compares the appearance trends included in the same group and an appearance tendency comparison result by the appearance tendency comparison unit. The output semantics converter module 68 controls the operation unit based on the action selected by the action selection calculation unit 82 to cause the selected action to appear. An operation unit control means is configured.

Here, the action selecting section 80 is configured such that the perceptual information acquired by the perceptual information acquiring section 90 from the recognition result and the motivation information acquired by the motivation information acquiring section 81 from the internal state information from the internal state model section 71. Based on the action selection operation unit 82
Select an action in. The action selecting section 80 will be described later in detail.

On the other hand, the internal state model unit 71 has an internal state model that changes the instinct and emotional state (internal state) of the robot apparatus 1 according to external factors and internal factors. Here, the external factor is, for example, information that is hit, information that is stroked, or a command from a user. Further, the internal factor is information such as a low battery level or a rise in body temperature.

More specifically, the internal state model section 71 changes the internal state based on the recognition result of the input semantics converter module 59, and uses the information of the internal state as the action selecting section 80 and the modulator 72. Output to

The motive information acquiring section 81 acquires motive information based on the internal state information. This will be described later in detail.

On the other hand, the modulator 72 is a part for generating action information (action command) to be finally displayed. Specifically, the modulator 72 includes the action selection unit 80
The behavior information finally appearing is generated from the behavior selected in and the information on the internal state from the internal state model unit 71, and is output to the output semantics converter module 68.

The modulator 72 causes the behavior determined by the behavior selection unit 80 to be added to the instinct and emotional state obtained by the internal state model unit 71, so that the behavior appears in the robot apparatus 1. Will be able to do that. That is, for example, the action selecting unit 80 selects an action such as “eat an apple” as the next action from the recognition result or the like, while the internal state model unit 71 selects the current internal state of the robot device 1 based on the recognition result. For example, "angry" is acquired. Then, based on the information, the modulator 72 adds the internal state of “angry” to the action of “eat apple” to generate information of the action of “eat apple badly”. , And outputs the information to the output semantics converter module 68. The output semantics converter module 68 sends a signal to each of the signal processing modules 61 to 67 and controls each device to control the operation unit, so that the robot apparatus 1 is disgruntled as the next action (target action). The behavior of eating apples will appear.

The internal state information indicating the state of instinct and emotion generated by the internal state model unit 71 is also used as information when the robot apparatus 1 determines (selects) an action. In addition, it is used in a form that is expressed (added) to the determined action.

As described above, the action determining system 70 determines an action based on the recognition result. Hereinafter, each unit constituting the action determination system 70 will be described in more detail.

(3-1) Configuration of Internal State Model Unit The internal state model unit 71 changes internal states such as instinct and emotion according to external factors and internal factors. The state of instinct and emotion obtained by the internal state model unit 71 is used as information when the robot apparatus 1 determines an action, as described above, and is further expressed for the determined action. Also used in the form of (added).

The internal state model unit 71 has a plurality of elements related to instinct (desire) and personality that change according to external factors and internal factors.

More specifically, the internal state model unit 71 includes: fatigue, heat or body temperature, pain, appetite or hunger, and dryness (thirs).
t), nine instinct factors such as affection, curiosity, elimination and sexuality, and happiness and sadness.
s), anger, surprise, disgus
t), fear, frustration, boring
redom), sleep (somnolence), sociability (gregariousnes)
s), patience, tension, relaxed, alertness, guilt, spite, spy, loyalty, submission
And 18 emotional elements such as jealousy, etc., and 27 elements indicating the internal state as a whole.

The emotion element holds a parameter representing the intensity of the emotion for each emotion. Then, the internal state model unit 71 compares the parameter value of each of these elements with a specific recognition result such as “hit” or “stroke” given from the input semantics converter module 59 and elapsed time. Is changed periodically based on

Specifically, the emotional element is determined based on the recognition result given from the input semantics converter module 59, the behavior of the robot device 1 at that time, the elapsed time since the last update, and the like. The amount of change of the emotion at that time calculated by the arithmetic expression is expressed as △ E [t],
E [t] of the current parameter value of the emotion, the coefficient representing the sensitivity of the emotion as k _e, (1) to calculate the emotion parameter value E [t + 1] in the next cycle by formula, which currently Of the emotion is replaced with the parameter value E [t] of the emotion.

[0064]

(Equation 1)

The internal state model unit 71 updates the parameter values of all emotions such as the above-mentioned “happy” by the equation (1) shown above.

The degree to which each recognition result and the notification from the output semantics converter module 68 affect the variation ΔE [t] of the parameter value of each emotion is determined in advance. Is the amount of change in the parameter value of the emotion of “anger” △ E
[T] is greatly affected, and the recognition result such as “stroke” is the variation amount of the parameter value of the emotion of “joy” 喜 び E
[T] is greatly affected.

Here, the notification from the output semantics converter module 68 is so-called action feedback information (action completion information), information on the appearance result of the action, and the internal state model unit 71 Change emotions with information. Note that the internal state model unit 71 also changes the instinct described later.

This is, for example, such that the behavior of "barking" lowers the emotional level of anger. The feedback of the action result may be made by the output of the modulator 72 (the action to which the emotion is added).

On the other hand, the instinct element holds a parameter indicating the strength of the desire for each desire (instinct). Then, the instinct element periodically updates the parameter values of these desires based on the recognition result given from the input semantics converter module 59, the elapsed time, the notification from the output semantics converter module 68, and the like.

Specifically, the instinct factors include “tired”, “affection”, and the recognition results of “curiosity”, “libido”, and “excretion”, elapsed time, and notification from the output semantics converter module 68. The amount of change in the desire at that time calculated by a predetermined arithmetic expression based on
Assuming that the parameter value of the current desire is I [k] and the coefficient k _i representing the sensitivity of the desire is a parameter value I [k +
1] is calculated, and the calculation result is replaced with the current parameter value I [k] of the desire to update the parameter value of the desire.

[0071]

(Equation 2)

The internal state model unit 71 updates the parameter values of all instinct (desirability) such as fatigue described above by using the equation (2).

The degree to which the recognition result and the notification from the output semantics converter module 68 affect the variation ΔI [k] of the parameter value of each desire of each instinct element is determined in advance. For example, the notification from the output semantics converter module 68 is
This greatly affects the amount of change ΔI [k] in the parameter value of “fatigue”.

As shown below, a parameter value of a predetermined desire can be determined.

Of the instinct factors, “pain”
A signal processing module 55 for attitude detection of middleware layer 40, based on the number of times of abnormal posture notified via the input semantics converter module 59, the number is N, and the intensity of pain and K _1, the rate constant pain decreases and K _2, the equation (3), "pain"
Is calculated, and the parameter value of the “pain” is changed by replacing the calculation result with the parameter value I [k] of the current pain. Where I
When [k] <0, I [k] = 0, t = 0, and N = 0.

[0076]

(Equation 3)

As for “heat”, the temperature is set to T, the outside air temperature is set to T _0, and the temperature is set to T ₀ based on the temperature data from the signal processing module 51 for temperature detection supplied via the input semantics converter module 59. as an increase coefficient K _3, the equation (4), and calculates a parameter value I [k] of the "heat", of the "hot" and to replace the result of the calculation parameter value I of the current heat [k] and Update parameter values. When T−T ₀ <0, I
[K] = 0.

[0078]

(Equation 4)

[0079] In addition, the "appetite", based on the remaining battery amount data supplied via the input semantics converter module 59 (information module was obtained for the battery remaining amount detecting not shown), a battery remaining amount B _L
In a predetermined period, the parameter value I [k] of “appetite” is calculated by the equation (5), and the calculation result is replaced with the current appetite parameter value I [k]. Update the parameter value of.

[0080]

(Equation 5)

As for “dry”, the remaining battery level at time t is set to B _L (t) based on the rate of change of the remaining battery level provided via the input semantics converter module 59, and the current times t ₁ and t ₁ Assuming that the remaining battery data is obtained in step ₂ , the parameter value I [k] of “dry” is calculated by the equation (6), and this calculation result is replaced with the current dry parameter value I [k]. The parameter of “dry” is updated.

[0082]

(Equation 6)

In the present embodiment, the parameter value of each emotion and each desire (instinct) is regulated to fluctuate in the range of 0 to 100, and the coefficient k
_e, the value of k _i is also set individually for each emotion and each desire.

As described above, the internal state model unit 71 is configured, and the robot apparatus 1
According to 1, it is possible to perform an autonomous action by changing an instinct (desired) or an emotional state (parameter) in accordance with the situation of the user and the surroundings, and instructions and actions from the user.

(3-2) Change of instinct and emotion according to environment In addition to the above configuration, in the case of the robot apparatus 1, for example, when the surroundings are "bright", the surroundings become cheerful, whereas the surroundings are "dark". Depending on the degree of the three conditions (hereinafter, referred to as environmental conditions) of “noise”, “temperature”, and “illuminance” in the surrounding environment, such as sometimes being quiet,
It is designed to change your instinct.

That is, the robot apparatus 1 has the above-described C as an external sensor for detecting the surrounding situation.
In addition to the CD camera 20, the distance sensor 22, the touch sensor 21, the microphone 23, and the like, a temperature sensor (not shown) for detecting an ambient temperature is provided at a predetermined position. The corresponding configuration is middleware layer 4
The 0 recognition system 60 is provided with signal processing modules 50 to 52 for noise detection, temperature detection, and brightness detection.

The noise detection signal processing module 50 determines the ambient noise level based on the sound data from the microphone 23 (FIG. 2) given via the virtual robot 33 of the robotic server object 32. It is configured to detect and output the detection result to the input semantics converter module 59.

The signal processing module 5 for temperature detection
1 detects the ambient temperature based on the sensor data from the temperature sensor provided via the virtual robot 33 and outputs the detection result to the input semantics converter module 59.

Further, the signal processing module 52 for detecting brightness detects surrounding illuminance based on the image data from the CCD camera 20 (FIG. 2) given through the virtual robot 33, and inputs the detection result. Output to the semantics converter module 59.

The input semantics converter module 59 recognizes the degree of “noise”, “temperature” and “illuminance” of the surroundings based on the output of each of the signal processing modules 50 to 52, and determines the recognition result as described above. Internal state model part 7 of application module 41 (FIG. 5)
Output to 1.

More specifically, the input semantics converter module 59 includes a signal processing module 5 for noise detection.
Recognizing the degree of surrounding “noise” based on the output of 0,
The recognition result such as “noisy” or “quiet” is output to the internal state model unit 71.

The input semantics converter module 59 recognizes the degree of surrounding "temperature" based on the output of the temperature detection signal processing module 51, and outputs the recognition result such as "hot" or "cold" to the internal state. Output to the model unit 71 and the perceptual information acquisition unit 90.

Further, the input semantics converter module 59 includes a signal processing module 5 for detecting brightness.
2. Recognize the degree of "illuminance" around based on the output of 2,
The recognition result such as “bright” or “dark” is output to the internal state model unit 71 or the like.

As described above, the internal state model unit 71
Based on the various recognition results provided from the input semantics converter module 59, each parameter value is periodically changed according to equation (1).

Then, based on the recognition results of “noise”, “temperature”, and “illuminance” given from the input semantics converter module 59, the internal state model unit 71 calculates a predetermined corresponding emotion ( 1) increase or decrease the value of the coefficient k _e of expression.

[0096] Specifically, internal state model unit 71, for example, emotion values of the coefficient k _e is increased by a predetermined number with respect to the "noisy" when such recognition result is given "anger", whereas reducing the predetermined number of values of the coefficient k _e for the emotion "angry" when the recognition result such as "quiet" is given Te. As a result, the parameter value of “anger” changes due to the influence of the surrounding “noise”.

Further, the internal state model unit 71 indicates “hot”
For recognition when the result is given reduces a predetermined number of values of the coefficient k _e for emotion "joy", the emotion of "sadness" when the recognition result such as "cold" is given to this such the value of the coefficient k _e is increased by a predetermined number.
As a result, the parameter value of “sadness” changes due to the influence of the surrounding “temperature”.

Further, the internal state model unit 71 sets
If a recognition result such as
Coefficient k for emotion_eIs increased by a predetermined number.
Then, if a recognition result such as [dark] is given,
Coefficient k for emotion of "fear" _eIncrease the value of
You. As a result, "fear" is affected by the surrounding "illuminance"
Will change.

Similarly, the internal state model 71
, As described above, periodically changes the parameter values of each desire in accordance with the equations (2) to (6) based on the various recognition results and the like provided from the input semantics converter module 59.

Further, based on the recognition result of the degree of “noise”, “temperature” and “illuminance” given from the input semantics converter module 59, the internal state model unit 71 determines a predetermined corresponding desire. (2) increasing or decreasing the value of the coefficient _{k i} of expression.

[0102] Specifically internally state model unit 71, for example, when the recognition result such as "noisy" or "bright" is given, the value of the coefficient k _i for "tired" is decreased a predetermined number, to "quiet" or when the recognition result such as "dark" is given to increase the predetermined number of values of the coefficient k _i for "tired" against. Also, the internal state model unit 71
Increases the value of the coefficient k _i for “fatigue” by a predetermined number when a recognition result such as “hot” or “cold” is given.

As a result, in this robot device 1,
For example, when the surroundings are "noisy", the parameter value of "anger" tends to increase, and the parameter value of "tired" tends to decrease. When the surroundings are “quiet”, the parameter value of “anger” tends to decrease, and the parameter value of “tired” tends to increase, so that the behavior is “settled” as a whole.

When the surroundings are “hot”, “joy”
Parameter values tend to decrease, and the parameter value for “fatigue” tends to increase.
Sometimes the parameter value for "sadness" tends to increase,
Since the parameter value of “tired” is likely to increase, the behavior is “cool” behavior as a whole.

When the surroundings are "bright", the parameter value of "joy" tends to increase and the parameter value of "tired" tends to decrease, so that the behavior becomes "cheerful" as a whole. On the other hand, when the surroundings are "dark", the parameter value of "joy" tends to increase, and the parameter value of "tired" tends to increase, so that the behavior becomes a "quiet" behavior as a whole.

As described above, the robot apparatus 1 can change the state of instinct and emotion according to the environment (external factors and internal factors) by the internal state model unit 71. Can be expressed in actions. Further, the state of instinct and emotion obtained by the internal state model unit 71 is also used as information for selecting an action in the action selection unit 80.

(3-3) Configuration of Action Selector The action selector 80 selects one action from a plurality of actions prepared in advance. The behavior selection unit 80 is configured to select (determine) behavior by an ethological approach.

In general, it is considered that an animal determines behavior based on a plurality of external causal factors and internal causal factors (hereinafter collectively referred to as causal factors) that affect behavior. Determines behavior in a state where these factors are intertwined in a complex manner. The robot apparatus 1 adopts such a general mechanism for determining the behavior of an animal to determine its behavior.

The robot device 1 in which the mechanism of behavior determination is constructed by an ethological approach, for example, as shown in FIG. 6A, appears when there is a puddle in front of the eyes by the following procedure. Decide what action to take.

The robot apparatus 1 perceives and recognizes (evaluates) an external causal factor (external perceptual element, for example, based on perception) of “discover water” and “water is 10 cm”. On the other hand, for example, the motive as the internal causal factor (internal motivation element, for example, based on instinct or emotion) is “robbing throat dry” or “moderately angry”. In state. The motive is obtained using the parameter values of the internal state model unit 71 described above. This will be described in detail later.

Here, according to the ethological approach, at least the following judgment is made in determining the behavior.

Even in the case of “throat drying is high” and “find water”, for example, when the distance to the water is long, the action of taking water (water intake action) Not exclusively. For example, when the distance to the water is long, the physical strength is consumed correspondingly, and the instinct that the throat becomes more severe is dealt with.

On the other hand, when the state is "throat is low" and "the water is in front of the eyes", the water may be taken. In this way, it is not determined from the internal causal factors such as "throat dryness" whether or not water intake behavior is performed in a straightforward manner. The decision of the action decision is made from the external causal factors. That is, the behavior is determined based on a plurality of external causal factors and internal causal factors, and is determined in a state where those factors are intertwined in a complicated manner.

At this time, the action is finally determined as compared with other actions. For example, in the state of "want to drink water" and "want to eat food", the degree of "want to drink water" or its feasibility and the degree of "want to eat food" or its feasibility By comparing gender, for example, a watering behavior is selected as one behavior.

[0114] By such an ethological approach, the robot apparatus 1 finally determines the behavior. That is, the robot apparatus 1 performs the “water discovery” under the situation “the throat dries high”,
Comprehensively judge from information such as "m" and exclude other behaviors such as "feeding behavior" to make "watering behavior" appear.

Further, the robot apparatus 1 causes the user to take a water action with an angry action as the state of “medium angry”. This is realized by the modulator 72 described above. And, for example, the robot apparatus 1 detects water as an internal state,
Decrease the level of that "anger". This is realized by the feedback of the action completion information from the output semantics converter module 68 to the internal state model unit 71 described above.

FIG. 6B shows a procedure up to selection of "walk forward" as a specific action of "water intake behavior" by the ethological approach as described above. .

First, when the robot apparatus 1 is in the state shown in FIG. 6A, the robot apparatus 1 selects "ingestion behavior" from a plurality of behaviors. The other actions not selected include, for example, a “fighting action” and a “searching action”. For example, the robot apparatus 1 is set as an action group in which such “ingestion action”, “fighting action”, “searching action”, and the like can be selected, and is held as a subsystem (subsystem layer).

Here, the action group is composed of a plurality of lower-level action groups that embody a higher-level action. For example, the lower-level actions have a relation of suppressing each other. Hereinafter, the same applies.

Next, the robot apparatus 1 selects “water intake behavior” as one of the selected intake behaviors. The other ingestion behavior not selected includes, for example, “eating behavior”. For example, the robot apparatus 1 is set as an action group in which such “water-feeding action”, “feeding action”, and the like can be selected, and holds it as a mode (mode layer). That is, “water intake behavior” and “feeding behavior” are set as behavior groups, and are held as lower behaviors of “intake behavior”.

Next, “approaching water” is selected as one of the selected watering behaviors. As another water intake behavior, for example, "put water in the mouth" can be mentioned. For example,
The robot apparatus 1 is a group (selectable) of “approaching water” and “pour water”, and is held as a module (module layer).

Next, "go forward" is selected as a specific action of "approaching water" to appear as an actual action. Examples of the other actions of “approaching water” that are not selected include “go backward”, “turn right”, and “turn left”. For example, robot device 1
Are groups that can be selected from among such “forward”, “reverse”, “turn right”, “turn left”, and the like, and are held as motor commands (motor command layers).

According to such a procedure, the robot apparatus 1
Makes the behavior of the lowest layer, such as "walk forward" from the ethological approach, appear as the final behavior of "ingestion behavior" that defines the higher behavior.

FIG. 7 shows an action selection system constructed for the action determination of the robot apparatus 1. This action selection system is constructed, for example, in the action selection unit 80.

The action selection system is constructed such that a plurality of actions have a hierarchical structure (tree structure). In the action selection system having the hierarchical structure, an action group including a plurality of lower actions is configured from an action group that embodies the upper action. For example, if the higher-level action is “fighting action”, the lower-level action is “fighting / predatory action”,
It seems to consist of actions such as "defense and flight behavior".

For example, the action selection system may hold information on each action in a data form (for example, a database form) or may be designed as an object-oriented one. For example, when the action selection unit is designed as object-oriented, in the action selection system, each action is independently configured as a unit of an object, and various processes for action selection are performed for each object.

As shown in FIG. 7, in the action selection system in which a plurality of actions are constructed in a hierarchical structure, the action of the higher layer is an abstract action like a desire, and the action of the lower layer is the abstract action. It is a concrete action to fulfill such a desire.

In such an action selection system, an action is selected while proceeding to a lower layer, that is, an action that realizes an upper action is selected, and a final action is selected. That is, regarding the actions in the middle class, the information on the selected route from the highest action to the lowest action is selected.

In this way, when the action is selected in each layer while proceeding from the upper layer to the lower layer, each layer performs one action by the external causal factor and the internal causal factor as described above. The choices are being made.

[0129] Specifically, as shown in FIG. 5, the behavior selecting section 80 includes a perceptual information obtaining section 90 and a motivation information obtaining section 81.
And an action selection calculation unit 82. Each unit constituting the action selecting unit 80 functions roughly as shown in FIGS. 8 and 9.

[0130] The perceptual information acquisition section 90 acquires perceptual information for each action. For example, as acquisition of perceptual information, an RM value indicating evaluation of perception in a release mechanism (Release Mechanism) described later is calculated. For example, the perceptual information acquisition unit 90 finds water and recognizes that the distance to the water is 10 cm. This allows
The value of the ingestion behavior (watering behavior) increases, that is, the selectivity increases.

The motivation information acquisition section 81 acquires motivation information for each action based on the internal state. For example, as acquisition of motivation information for each action, a motivation for each action is calculated based on the instinct values and emotion values described above. Specifically, M indicates the state of the motive in the motivation creator described later.
Calculate the ot value. For example, the motivation information acquisition unit 81 acquires a dry throat state. As a result, the motivation value of the eating behavior is increased, and among the eating behaviors, the drinking behavior is further increased.

The action selection calculation unit 82 selects a desired action from the motive information (motivation value) from the motive information acquisition unit 81 and the perception information (value) from the perception information acquisition unit 90 for each behavior. . Then, when selecting a desired action, the action selection calculation unit 82 exclusively controls other actions in the same group (action group). For example, the behavior selection operation unit 82 selects an intake behavior in the subsystem layer,
In addition, the water intake behavior in the intake behavior is selected.

Further, the action selection calculating section 82 plans an actual action group based on the selected action. For example,
"Move-forward" is selected, and so on.

As described above, the internal state model unit 71 acquires information on internal states such as instinct and emotional state in the body. For example, an instinct value and an emotion value are calculated as information on an internal state. Specifically, the parameter value of the instinct (desired) described above, the parameter value of the emotion, or the IE value described later is calculated. For example, the internal state model unit 71 acquires information on a state that the throat is dry because of movement.

The output semantics converter module 68 is a module for converting an action into an operation sequence according to the type of the robot device 1, as shown in FIG. For example, the output semantics converter module 68 now recognizes that the type of the robot device 1 is a quadruped robot, and converts it into an action sequence according to the input behavior and emotional state. That is, the output semantics converter module 68 sends a command from the action command from the higher-level action determination system 70 to the corresponding signal processing modules 61 to 67.

The modulator 7 shown in FIG.
2. The posture management module will be described later in detail.
In FIG. 9, the “input” column indicates the form of the input command, and the “output” column indicates the form of the output command.

As described above, the action selecting section 80 is configured. Next, the ethological approach adopted by the behavior selection unit 80 as a behavior selection system will be described in more detail.

(3-4) Behavior Selection by Ethological Approach Generally, the determination (selection) of the behavior of an animal is made by intricately intermingling a plurality of factors as described above. . As a simple example, as shown in FIG. 10, an action is determined from a perception and a motive.

Here, the perception is information from the outside that affects the behavior, and can be considered as a condition induced or restricted by the input external environment. The motive is internal information such as “I am hungry” and expresses an internal state, and can be considered as an internal intention to want to take the action. As described above, it can be considered that a behavior is determined as an action due to a perception or a motive.

The action decision based on the perception and the motivation is made specifically as follows. The principle of action decision (selection) described below is based on Silby and Mcfarland (197
This is based on the state space approach proposed in 5).

[0141] Silby and Mcfarland (1975) start with a theory based on the assumption that animals perform activities (behaviors) that are most likely to occur. One direction that clearly shows the arrangement of the appearance tendency is a vector space. Therefore, the magnitude of the vector indicates the strength of the so-called appearance tendency based on an index having a certain commonality. The appearance tendency is
For example, there is a tendency (degree) in which eating behavior appears and a tendency (degree) in which watering behavior appears. All appearance states are represented as points on the appearance space.

This space is divided into regions in which the behavior having the highest appearance tendency is, and the boundary line is defined as a “switching boundary line (switching line).
ne) ".

On the other hand, the appearance tendency depends on various causal factors. For example, eating habits depend on food restrictions, eating opportunities, the possibility of feeding, and the like. A clear way to express all these causal factors is to use another vector space. Determining the appearance tendency from the causal factor is synonymous with the following expression. In other words, there is a mapping from the state space of the causal factor to the appearance space, and there is a state space of the appearance tendency corresponding to any state of the causal factor. Then, it is possible to determine which action is to be performed. The relationship between the causal factor and the appearance tendency will be described with reference to FIG.

(A) and (C) of FIG. 11 show a causal factor state space indicating the state of the causal factor. This causal factor state space is constituted by causal factors that influence a certain action to be guided. The causal factors include the above-mentioned “perception” and “motivation”. Note that FIG.
In No. 1, only the two-dimensional space is considered for simplicity, but in reality, many of the appearance tendencies of actions are determined by the causal factor state space of three or more dimensions.

FIG. 11A specifically shows the tendency of eating behavior to appear, that is, the tendency of “feeding behavior” to appear (hereinafter referred to as eating tendency). "Hunger" is taken on the horizontal axis as "motivation" as a factor, and "deliciousness" is taken on the vertical axis as "perception" as a causal factor. In FIG. 11, (C) shows the tendency of the appearance of the water intake behavior, that is, the appearance tendency of the “water intake behavior” (hereinafter referred to as the water intake tendency). The abscissa represents "kawaki" and the ordinate represents "distance to water" as "perception."

FIG. 11B is a diagram corresponding to FIG.
6C shows the appearance tendency space of “feeding tendency” and “water intake tendency” based on the state of the causal factor in (C). That is, the appearance tendency of the action affected by the causal factor is mapped, and the space in which the appearance tendency can be compared is shown.

First, regarding the causal factor state space, FIG.
This will be specifically described with reference to FIG. The causal factor state space shown in FIG. 12 corresponds to the “feeding behavior” shown in FIG.
Is the causal factor state space.

[0148] As shown in FIG. 12, when it is in there is what really delicious (m ₂ state), but not crowded so much stomach condition (cause state) (n ₁ state), on the other hand, we are hungry is extremely stomach Is (n ₂ (> n ₁ ) state),
When there is some food that is not so delicious (cause state) (m ₁ (<m ₂ ) state), eating behavior appears. In other words, eating behavior does not appear to depend solely on motivation "hunger" but also depends only on perception "deliciousness". It is not a thing, but "hungry" and "deliciousness" interact, and eating behavior appears.

In other words, even if the degree of “hunger” is different, the eating behavior appears, and depending on the state of “hunger” and “deliciousness”, the eating behavior is within the causal factor state space. There are causal states in which the degree of appearance of the action is the same, that is, causal states in which the eating tendency is the same. For example, there is a similar tendency to eat when feeding very good food when not hungry, or when feeding very bad food when very hungry.

For example, in the eating behavior, when “hungry” and “deliciousness” are considered as causal factors, the appearance tendency is almost the same. Is low (less), and when "hunger" is low (less), "deliciousness" is high (many). Therefore, when considering that the eating tendency is the same, “hunger” and “deliciousness” have an inversely proportional relationship, and when the points where the eating tendency is the same are connected, for example, in FIG. Is shown as a curve. That is, as shown in FIG. 12, there are a plurality of causal factor states that make the so-called strength (vector strength) y of the eating tendency the same, and these are shown as curves in the causal factor state space. It is.

Further, in the causal factor state space, there are a plurality of different feeding tendency intensities y ₁ , y ₂ ,.
As shown in Fig. 2, it is the distribution of the tendency of eating tendency.
This is shown as a so-called contour line.

In FIG. 12, it is shown that the intensity of eating increases as going to the upper right in the causal factor state space. This means that if you are very hungry, and if you have very good food in front of you, everyone will eat.

As described above, the strength of the tendency to eat can be defined by the causal factors, and similarly, the strength of the tendency to drink water can be defined.

[0154] That is, the water intake behavior occurs when the throat is intense and appears even when the distance to the water (the place where the water is present) is long, and when the throat is small, the distance to the water is short. Appears, and “throat throat” and “distance to water” interact, and a water intake behavior appears.

In other words, even when the degree of “throat throat” is different, the water intake behavior appears,
Depending on the states of "throat throat" and "distance to water", there are causal states at a plurality of points in the causal factor state space that have the same tendency to water intake. For example, there is a tendency for water to be comparable when water is in front of you when water is not scarce, or when water is far away when water is very scarce. That is.

When considering that the tendency of water intake is almost the same, “throat throat” and “distance to water” are inversely proportional, and the point that the tendency of water intake is the same is considered. When they are connected, for example, they are shown as curves in the causal factor state space in FIG. That is, as shown in FIG. 11 (C), there are a plurality of causal factor states in which the so-called strength x of the water intake tendency is substantially the same, and these are shown as curves in the causal factor state space. . Then, as shown in FIG. 11 (C), the strengths x ₁ , x ₂ ,... Of different water intake tendencies are shown as contour lines in the causal factor state space.

As described above, the strength of “feeding tendency” and the strength of “water intake tendency” are obtained based on the state of each causal factor, and the comparison is made based on these strengths. , An action is determined (selected). Then, such an appearance tendency is compared in the appearance tendency space shown in FIG. The appearance tendency space is made up of behavior tendency that can appear.

[0158] For example, in some causes state, when the intensity of x ₂ Strength y ₁ phrase water consumption tendency of a feeding trend was obtained, as shown in FIG. 11 (B), the causative agent state the strength x of strength y ₁ and the water consumption trend of eating trend that has been mapped from space
₂ are combined and compared in the appearance tendency space. Specifically, the action is selected as follows.

As shown in FIG. 11B, the appearance tendency space is divided into two regions by setting the switching boundary line. A region surrounded by y = 0 (x-axis indicating a tendency to water intake) and a switching boundary (hereinafter, referred to as a water-supplying action selection region), and x = 0 (y-axis indicating a tendency to eat) and a switching boundary. (Hereinafter, referred to as an eating behavior selection area).

As described above, in each area formed by setting the switching boundary line in the appearance tendency space, one action is determined by the position of the value (x, y) mapped from the causal factor state space. That is, when the value (x, y) is within the watering action selection area, the watering action is selected, and when the value (x, y) is within the watering action selection area, the watering action is selected. Eating behavior is selected. Therefore, (C) in FIG.
In the example shown in ( ₁ ), since the value (x ₂ , y ₁ ) is in the watering action selection area, the watering action is selected.

For the sake of simplicity, the state space of the causal factor is shown by dividing it into state variables (causal factors) related to eating and drinking, respectively. Influences the appearance of behavior. The curves in the causal factor space lead to conditions that lead to the same level of propensity for certain behaviors.

Also, depending on the finally selected action,
It may affect the causal factor involved and several other causal factors. For this reason, information exclusion processing is performed.

As an ethological approach as described above, a method of action determination (selection) using causal factors has been proposed by, for example, Silby and Mcfarland (1975 paper) and Ludlow (as a competition model). .

(3-5) Formula for Realizing Behavior Decision by Ethological Approach The above-described ethological approach to behavior decision is only a theory, and is applied to the actual robot apparatus 1. For this purpose, it is necessary to convert the above ethological approach into information as a database or the like, or to make a mathematical expression. Therefore, in order to realize the present invention, the above-described behavioral determination based on the ethological approach is expressed by the following mathematical formula.

As shown in FIG. 13A, the state (degree) of “hunger” which is a causative factor of “feeding behavior” is represented by M
ot [0], and the evaluation of “deliciousness” is RM [0]. The feeding tendency (strength of the tendency) at a certain value of certain Moto [0] and RM [0] is set to Be [0].

Similarly, as shown in FIG. 13 (B), the state (degree) of “throat throat”, which is a causative factor of water intake behavior, is set to Moto [1], and “distance to water” is set. Is set to RM [1]. And Mot [1] and RM
[1] The water intake tendency (strength of the tendency) at a certain value is expressed by Be.
[0]. These relationships are as shown in the following table.

[0167]

[Table 1]

In this example, since the two appearance trends to be compared are “feeding behavior” and “water intake behavior”, the perception is represented by two values of RM [0] and RM [1], and the motivation is represented by RM. Although the two values, that is, [Mot [0]] and [Mot [1]], are used, it is possible to compare many appearance trends. From this, the perception (external intellectual element) is expressed as RM [i],
The motivation (internal motivation element) is Moto [i] and the appearance tendency is B
Let e [i], where i is an integer, generalized. In the following, when not specifically referring to the types of actions that appear,
These generalized versions are shown below.

In the above-described example, in the case of "feeding behavior", a similar appearance tendency indicates that "hunger" and "deliciousness" are established in an inversely proportional relationship. However, in order to have the same appearance tendency, the causal factors affecting the appearance tendency are not necessarily in an inversely proportional relationship. That is, Be [i], RM [i], and Mot
Although [i] can be expressed as a relationship as shown in Expression (7), the relationship between RM [i] and Mot [i] is not always inversely proportional. The point is that the appearance tendency is not only affected by the motive (internal motive element) but also by the perception (external intellectual element).

[0170]

(Equation 7)

The perceptual evaluation RM [i] such as “deliciousness” and “distance to water” is acquired by the perceptual information acquiring unit 90, and includes “hungry” and “throat throat”. The motive Mot [i] is the motive information acquisition unit 8
1 is obtained. Perceptual information acquisition unit 90
Specific processing for acquiring these pieces of information in the motivation information acquisition section 81 will be described later in detail.

As described above, the eating tendency and the water tendency obtained from the perception (external intellectual element) and the motive (internal motive element) are shown in the appearance tendency space as shown in FIG. Is shown.

Here, in the appearance tendency space shown in FIG. 14, two switching boundary lines such as a first switching boundary line (y = αx) and a second switching boundary line (y = βx) are set. I have. That is, the appearance tendency space is divided into three regions. On the other hand, in the appearance tendency space shown in FIG. 11B, there was only one switching line. This is for the following reasons.

Theoretically, as described above, it is possible to select between different actions even with one switching boundary. However, when such a theory is applied to the actual robot apparatus 1 as it is, if the appearance tendency of each action exists near the set switching boundary line, switching between the currently selected action and the other action becomes difficult. , Robot device 1
Calms in his actions. Such a phenomenon occurs on the premise that when an action is selected and performed, the appearance tendency of the action becomes relatively smaller than other actions. That is, if the motivation (aspiration) is achieved, the degree of the motivation is reduced, and as a result, the tendency of the behavior influenced by the motivation is reduced.

As described above, by dividing the area by the two switching boundaries, the area where the “feeding action” is selected (feeding action selection area) and the area where the “watering action” is selected ( Watering behavior selection area), "feeding behavior" or "watering behavior"
Are selected (hereinafter, referred to as eating or watering action selection areas). Thereby, it is possible to prevent the behavior of the robot apparatus 1 from becoming restless. The reason why the setting of the two switching boundaries prevents the behavior of the robot apparatus 1 from becoming restless will be described later in detail.

In the appearance tendency space shown in FIG.
The action with the strongest appearance tendency is selected from the following relationship.

The appearance tendency space shown in FIG.
e [0] is set on the x-axis, and the water intake tendency Be [1] is set on the y-axis, and is composed of the eating tendency Be [0] and the water intake tendency Be [1]. Then, in such an appearance tendency space, the first and second switching boundaries are respectively defined as y = αx
And y = βx. For example, the slope coefficients α and β
Is determined as an arbitrary value, and can be determined according to, for example, the growth of the robot apparatus 1 or the like.

Here, the feeding tendency Be [0] is represented by “Hungry” Moto [0] and “Taste” R shown in FIG.
It is a value determined by M [0], and the water intake tendency Be
[1] is the “Nodokadaki” Mo shown in FIG.
t [1], a value determined by “distance to water” RM [1].

In such an appearance tendency space, FIG.
As shown in (1), when the value (a, a ′) mapped from the causal factor state space is in the eating action selection area (point C), the eating action is selected as one action, while the value When (a, a ′) is in the water-feeding action selection area (point D), the water-feeding action is selected as one action.

The value (a, a ') is, for example, as shown in FIG.
As shown in (A) of FIG. 3, “Hunger” is Mot [0] = n
₀ , “deliciousness” is in the state of RM [0] = m ₀ , and at this time, the “feeding tendency” Be [0] is set to a, and further, as shown in FIG. Which one is Mot
[1] = n ₁ , “distance to water” is RM [1] = m ₁ , and the “water intake tendency” Be [1] is a ′.

The above action selection can be realized by the following mathematical expression.

First, a '/ a (Be [1] / Be
[0]). The condition that the value (a, a ′) is located in the water-supplying action selection area surrounded by x = 0 and the second switching boundary (y = βx) is as follows: ∞> a ′ / a>
β. Also, y = 0 and the first switching boundary line (y =
αx) and the value (a,
The condition where a ′) is located is α> a ′ / a> 0.

From the relational expression, the following expression can be derived. If α> a ′ / a> 0, that is, the value (a,
When a ′) is in the eating behavior selection area, the result is as shown in FIG. 15A, and the expressions (8) and (9) are satisfied.

[0184]

(Equation 8)

[0185]

(Equation 9)

Here, the inclination α of the first switching boundary line can be expressed as α ′ having a relationship as shown in equation (10). Note that α ′ is the water intake tendency Be, as described later.
The exclusive control gain (> 1) for the feeding tendency Be [0] is obtained from [1].

[0187]

(Equation 10)

From such a relationship, it is derived that the condition for selecting the “feeding behavior” satisfies the expression (11).

[0189]

[Equation 11]

Next, (B) in FIG. 15 shows the case of the water intake behavior, and the slope β of the second switching boundary is (12)
Given as an expression. Β is an exclusive control gain (> 1) from the feeding tendency Be [0] to the water feeding tendency Be [1], as described later.

[0191]

(Equation 12)

From such a relationship, it is derived that the condition for selecting the “water intake behavior” satisfies the expression (13).

[0193]

(Equation 13)

The above condition is satisfied by (14)
Equation (15) and Equation (15). Therefore, when taking a feeding action, the expression (14) is satisfied,
When taking a water-supplying action, the expression (15) is satisfied.

[0195]

[Equation 14]

[0196]

(Equation 15)

Here, when (a-a'α ') and (a'-aβ) as described above are described as a matrix, the expression (16) is obtained.

[0198]

(Equation 16)

If this is calculated discretely, the appearance tendency Be _t [i] at time t and the time t−1
(17) can be described by using the appearance tendency Be _(t-1) [i].

[0200]

[Equation 17]

[0201] Here, α 'is, exclusive control gain for the feeding trend _Be t [0] from the water consumption trend Be [1] (> 1)
Are shown, also, beta is as shown exclusive control gain (> 1) for feeding tends _Be t [0] from the water consumption tends _Be t [1]. For example, the image, as shown in FIG. 16, alpha 'serves as the exclusive control gain for feeding tends Be t _[0], also, beta acts as the exclusive control gain for feeding tends Be t _[1] .

As described above, the appearance tendency of a plurality of actions can be represented as a determinant. The positive Be in the matrix on the left side of the determinant
When there is a _{t [i],} actions to be corresponding to the occurrence tendency Be t _[i] is selected as an action.

In the case of the above-described determinant, since the value of one of the appearance trends is negative, 0 is substituted for the negative appearance tendency, and recursively. Make a calculation.

By performing the exclusive control recursively using the above equation (17), the action selection as shown in FIG. 17 can be made.

Here, it is assumed that when one selected action is executed, the influence of a causal factor on the one action is reduced, and the tendency of the executed one action is reduced. That is, for example, when the “feeding behavior” is selected as one action, the eating behavior is realized and the motive or the like for eating is satisfied.
Is less influenced by the causal factor (motivation), and consequently, the tendency to eat is reduced (weakened).
By performing the exclusive control recursively using the equation (17), the action is selected as follows.

As shown in FIG. 17, for example, (eating tendency Be [0], watering tendency [1]) = (a, a ′) is within the eating behavior selection region (y = 0 and y = αx In this case, the feeding behavior is selected as one action as long as the value (a, a ′) is in the eating behavior selection area. Here, the value (a, a ') be in a feeding behavior selection area, eating tendency Be t _[0] of the left-hand side of the above equation (17) is a positive.

[0207] When the feeding behavior continues to be selected, feeding tends Be t _[0] from the influence of the causative agent is reduced by implementation of the eating behavior as described above becomes smaller (weaker) . When eating tendency Be _t becomes smaller, the value (a, a ') leads to fed or water consumption selection.
That is, as shown by the arrow P ₁ in a graph shown in FIG. 17 value (a, a ') is changed.

In the eating or watering action selection area, an eating action is selected. Further, feeding the tendency of the left side at the above (17) _Be t [0] is also shown a positive. When the feeding behavior continues to be selected, eating trend since the influence of the causative agent by implementation of the eating behavior decreases Be t _[0] becomes smaller. Then, the value (a, a ′) ranges from the eating or watering action selection area to the watering area (an area surrounded by x = 0 and y = βx). That is, as shown by the arrow P ₂ in the graph shown in FIG. 17 value (a, a ') is changed.

[0209] In the water supply selection area, a water supply action is selected. Here, the value (a, a ') be in a water consumption behavior selection area, water consumption tends _Be t [1] of the left-hand side of the above equation (17)
But this time shows positive.

[0210] Then, if the water intake behavior is continuously selected, in the water intake behavior selection area, the effect of the causal factor is reduced due to the realization of the water intake behavior.
_t [1] becomes smaller. And the value (a, a ')
From the water intake behavior selection area to the eating or water intake behavior selection area. The feeding or water consumption behavior selection area, water consumption behavior is selected, feeding tendency of the left side at the above (17) Be t _[1]
Also indicates positive. Furthermore, the water consumption behavior continues to be selected, water consumption tends Be t _[1] decreases, the value (a, a ') leads to feeding behavior selected area from the feeding or water consumption behavior selection. In the eating behavior selection area, the eating behavior is selected again. That is, the change of the values from water consumption behavior selection area to eating behavior selection region (a, a ') is as shown by the arrow P ₃ in the graph shown in FIG. 17. Thereafter, the action is selected in this manner, and the action is switched.

Further, by setting two switching boundaries in the appearance tendency space, it is possible to prevent frequent switching of the behavior, and to prevent the behavior from becoming restless.

[0212] As described above, by the feeding tendency _Be t [0] and water consumption tends _Be t [1] is changed, the value from the relationship of these values (a, a ') = ( Be t [ 0], Be _t
[1]) is specified on the appearance tendency space, and one action is selected. Then, at this time, in equation (17),
Feeding tendency Be t _[0], it indicates the value of any positive water consumption tendency Be t _[1], the one action appearing tendency showed a positive is selected. Such an action decision is made by the decision selecting unit 71 shown in FIG.

[0213] In the embodiment described above, if two actions feeding behavior and water consumption behavior are switched has been described by ingestion tends Be t _[0] and water consumption tends Be t _[1]. However, actually, more actions (n actions) are compared in the appearance tendency space to select one action. That is, one action is selected by the appearance tendency space indicated by n dimensions. When one action is selected from n actions, a determinant as shown in Expression (18) is obtained.

[0214]

(Equation 18)

[0215] In this case, G [i] [j] is the emergence of behavior which is to a certain appearance tendency _Be t [j] of the action tendency Be _t
The exclusive control gain is [i].

With the above mathematical formula, the appearance tendency of each action is obtained based on causal factors such as perception and motivation, and one action is determined (or selected) from the strength (magnitude) of the appearance tendency. Behavioral decisions can be made using such an ethological approach.

For example, when an action is selected as shown in FIG. 17, the appearance tendency is finally 0, that is, the eating tendency Be [0] and the drinking tendency Be.
It is also conceivable that [1] converges to 0 (origin). This is because when the action is realized as described above,
As it is, causal factors (eg,
This is because the effect of (motivation) approaches endlessly.

However, it can be said that there is no problem because the influence of the causal factor on other unselected actions increases while the selected one action is being executed. That is, for example, while the watering action is being performed, the state of, for example, “hungry”, which is a causal factor for the unselected eating behavior, changes, and thus the evaluation of “hunger” is reduced. This is because the change increases the tendency to eat. Appetite seems to be restored by "sleeping" or "walking". That is, the appearance tendency of the action not selected while the selected action is being executed is so-called recovered. This is illustrated in, for example, FIG.

The inclinations α and β of the first and second switching boundaries can be arbitrarily determined. This makes it possible to appropriately indicate such behavior, for example, by setting according to the growth stage or according to the personality.

For example, the robot apparatus 1 is provided with a growth behavior model that changes the behavior to appear according to the growth stage. When the growth stage in the growth model is “infant”, the inclination α of the first switching boundary line is And the inclination β of the second switching boundary line are set to close values to narrow the eating or watering action selection area. In addition, when the growth stage is “adult”, the first step is to increase the feeding or watering behavior selection area.
Of the switching boundary line and the inclination β of the second switching boundary line are set.

As a result, the robot device 1 can be used as an “infant”
In the case of, switching between eating behavior and watering behavior is performed frequently, and behaviors that are less calm appear, and in the case of "adults," switching between eating behavior and watering behavior is appropriate. Choose at appropriate intervals to bring out calm behavior.

Further, the so-called recovery speed of the appearance tendency can be made variable. For example, the recovery speed is set according to the growth level such that the recovery speed is increased when the growth level is low, and the recovery speed is reduced when the growth level is high. With this setting, the switching between the eating behavior and the watering behavior is frequently performed in the case of "infant", and the switching between the eating behavior and the watering behavior is performed in the case of "adult". Is appropriately performed, and the same effect as in the above-described example can be obtained.

As described above, the convergence is prevented by using the recovery of the appearance tendency. However, the convergence may be similarly prevented by manipulating this in calculation.

The calculation formulas for realizing the behavior determination by the ethological approach in the actual robot device 1 have been described. Then, the action selecting unit 80 selects an action using such a calculation formula.

(3-6) Specific Processing in Action Selection Unit Hereinafter, specific processing in the action selection unit 80 will be described. As shown in FIG. 19, the action selection unit 80 includes a perception information acquisition unit (Release Mech) for acquiring perception information (RM).
anism, release mechanism) 90 and motivation information (Mo)
A motivation information acquisition section (motion creator, Motivation Creater) 81 for acquiring t) and an action selection calculation section 82 for selecting one action based on perception information (RM) and motivation information (Mot) are provided.

[0226] (3-6-1) on the basis of acquisition procedure perception evaluation of the appearance trend in the (perceptual information) RM [i] motivated state (motive information) Mot [i], the emergence tendency _Be t [i]
Will be described. Appearance trend _Be t [i]
Is roughly divided into calculation of a value before exclusive control and calculation of a value subjected to exclusive control. That is, (1)
8) the calculation of the occurrence tendency _Be t [i] of the first term of the right side of the equation is classified roughly into a calculation of (18) in the left side of the appearing tendency _Be t [i].

The calculation of the former and the calculation of the latter will be described by taking as an example a case where the appearance tendencies Be _t [i] of three different actions are obtained. The three different actions are the same action group. For example, as shown in FIG. 20, for each of three actions, three first to third perceptual evaluations RM [0],
From the RM [1] and RM [2] and the three first to third motivation states Mot [0], Mot [1] and Mot [2], the corresponding first to third appearance trends Be. _t [0],
Be _t [1], a case such that acquires _Be t [2].

For example, the three actions to be compared include “feeding action”, “water feeding action”, and “excretion action”. Regarding the “feeding behavior” as the first behavior, “deliciousness” is given as the first perceptual evaluation RM [0], and “hunger” is given as the first motivation state Mot [0]. As for the “water intake behavior” as the second behavior, “distance to water” is given as the second perception evaluation RM [1], and “throat throat” is given as the second motive state Mot [1]. Kawaki ". Further, regarding the “excretion behavior” as the third behavior, the third perception evaluation RM
[2] includes “distance to toilet”, and first motivation state Mot [2] includes “feces or urine clogged”. The appearance tendency space is based on the eating tendency B.
e _t [0], water consumption trend _Be t [1], excretion tendency Be
_t [2].

Each perceptual evaluation RM [i] and motivation state Mot
[I] based on the calculation of the "feeding behavior", each occurrence tendency Be t is corresponding to the "water consumption behavior" and "excretion behavior" _[i] is as follows.

Perception evaluation RM [i] and motivation state Mot
Since the [i], the emergence tendency Be _t as shown in (19)
[I] is calculated.

[0231]

[Equation 19]

Here, the perception evaluation RM [i] and the motive state M
If there is an inversely proportional relationship with ot [i], the relationship can be expressed, for example, as in equation (20).

[0233]

(Equation 20)

By substituting this perceptual evaluation RM [i] into equation (19), A [i] is calculated as _Bet [i]. That is, the perception evaluation RM [i] and the motive state Mot
When a relationship of inverse proportion between the [i], the coefficients A [i] is than is calculated as the appearance tends Be t _[i].

[0235] can be calculated by such a calculation, exclusive control before the emergence tendency _Be t [i]. And, for the exclusive control into account the emergence tendency Be t _[i] can be calculated by equation (21).

[0236]

(Equation 21)

As an image, as shown in FIG. 21, the exclusive control gain G [i] [j] (i = 0, 1, 2, j =
0, 1, 2), the appearance tendency B of the first to third actions
_{e t [0], Be t} [1], Be t [2] is calculated is performed as the exclusive control.

As described above, the appearance tendency before exclusive control and
Using the appearance tendency before the exclusion, an appearance tendency in which exclusive control is considered is calculated.

For example, according to the procedure shown in FIG.
These series of calculations are performed.

First, as shown in step S1, t =
0, Be _(t-1) [i] = 0, and initializes each value. Then, in steps S2 to S6, B
e _t [0] to Bet [2], the first on the right side of equation (21)
The value of the term is calculated. That is, exclusive control before the emergence tendency _Be t [i] is calculated. Step S2 to step S
The processing of No. 6 is specifically as follows.

In step S2, i = 0. This starts the calculation for _Bet [0].

Subsequently, in step S3, perception evaluation RM
[0] and the motive state Mot [0] are calculated. That is, for example, the evaluation RM [0] of “deliciousness” is obtained,
The state “Hungry” state Moto [1] is acquired.

Subsequently, in step S4, the appearance tendency Be of “feeding behavior” is set as the value of the first term on the right side of the equation (21).
Calculate _t [0].

Then, in step S5, it is determined whether or not i = 3. That, is made discrimination value is whether the calculation of all occurrences tend Be t to be compared _{[0] ~Be t [2]} .

Here, if i = 3, step S
In step 6, i = i + 1, and the processing from step S3 is started again.

By the processing in steps S1 to S6, the eating tendency Be is set as the value before exclusive control.
_{Following t} [0], the water intake behavior tendency Bet [1] and the excretion tendency Bet [2] are calculated.

When i = 3 in step S5, the processing of step S7 is executed. In step S7, the appearance tendency Be _t [i] on the left side of equation (21).
(I = 0 to 2) is calculated. That is, the (21) equation, the emergence exclusive control were considered trends _Be t [i] is calculated.

Subsequently, in step S8, Be
It is determined whether any of _t [i] is a positive value. Here, if none of Be _t [i] is a positive value, in step S9, t = t + 1, and the processing from step S1 is executed again. As a result, a recursive calculation as shown in Expression (21) is performed. That is, the Be obtained by the pre-processing
_The calculation is performed with _t [i] being Be _(t-1) [i].

On the other hand, if any one of the appearance trends Be _t [i] has a positive value, the appearance tendency Be _t [i]
_The action corresponding to _t [i] is selected as one action to actually appear, and the action selection processing ends.

As described above, the perception evaluation (perception information) RM
[I] and on the basis of the motive state (motive information) Mot [i], it is possible to determine the appearance tends _Be t [i].

(3-6-2) Processing in the Perceptual Information Acquisition Unit Next, the perceptual information acquisition unit 9 for acquiring the perceptual evaluation RM [i]
A specific configuration of 0 and the motivation information acquisition unit 81 that acquires the motivation state Mot [i] will be described. First, the perceptual information acquisition unit 90 will be described.

The perceptual information acquisition section 90 acquires perceptual information (evaluation) as a causative factor of the action according to external or internal information (recognition result). This perceptual information acquisition unit 90
As shown in FIG. 23, an action memory 91, an object name memory 92, an object determiner 93, an object information memory 94, and a perceptual information calculator 95 are provided.

The action storage device 91 stores a plurality of actions that can be selected. For example, a plurality of actions are stored as a database.

The action storage unit 91 outputs a plurality of actions to be compared as one action group to the object determiner 93 by inputting an action group number (signal).

For example, the example of “eating apples (eating behavior of apples)” is as follows.

Examples of the action that embodies the action of “eating an apple” include “approaching an apple”, “smelling an apple”, “putting an apple in a mouth”, and “touching an apple”. Can be For example, “approaching” is an action of shortening the distance to the object, “smell” is an action of bringing the nose closer to the object, for example, and “putting in the mouth” is The action of bringing an object into the mouth, "touching", is an action of bringing a hand (leg) into contact with an object. The actions corresponding to “approach”, “smell”, “put in mouth”, “touch”, and the like are actions applicable to the general target object. That is, when the target object is "Mikan", the action of shortening the distance is:
The action of approaching oranges, the action of approaching the nose is the action of smelling oranges, the action of bringing to the mouth is the action of putting oranges into the mouth, The action of touching the hand is the action of “touching oranges”.

The action memory 91 stores such an “approaching”
A plurality of pieces of behavior information applicable to the general target object are output to the target object determiner 93 as one behavior group. In other words, the action storage unit 91 extracts the action name information defined by extracting the information of the object of the lower-level action that embodies the higher-level action such as “eating an apple” or the like.
Output to The action name information output from the action storage device 91 corresponds to the action whose appearance tendency is compared in the action selection calculation unit 82. Therefore, the behaviors have a relation of suppressing each other.

As described above, holding action name information as information applicable to an object in general eliminates the need to define a plurality of signals (commands) for the same operation in which the object is different, and for each object. This is because, for example, a system such as a database is prevented from being enlarged when an action is specified, and conversely, when a similar action is performed, there is no large difference in the action itself for each object. As for a special action, the action may be defined again including the information of the target object.

On the other hand, the object name storage 92 stores the object names. The object name stored in the object name storage device 92 is the object name selected in the higher-level action. For example, when “eating an apple (eating behavior of an apple)” is selected as the top action, the robot apparatus 1 recognizes the presence of an apple. In this case, “apple” is the object name. Is stored in the object name storage device 92. Then, the object name storage 92 outputs the object name information (object signal) to the object determiner 93.

The above-mentioned action memory 91 outputs action information applicable to the object in general to the lower-order action to the object determiner 93. The object name storage 92 stores the object name in the object This is output to the object determiner 93. Therefore, the object determiner 93 uses the information (action name signal) output from the action memory 91 and the information (object signal) output from the object name memory 92 to generate a plurality of comparison targets. Actions are formed as complete information.

The object determiner 93 outputs a plurality of action information (action group signals) to the perceptual information calculator 95 in such a comparable form. That is, the object determiner 93 outputs the action name acquired by the action memory 91 and the object name acquired by the object name memory 92 to the perceptual information calculator 95 as a corresponding form (pair).

It should be noted that not all of the plurality of actions to be compared are actions requiring an object. In such a case, information such as “there is no object” is output from the object name storage 92 to the object determiner 93 in response to such an action. As a result, the target object determiner 93 uses the behavior information output from the behavior storage unit 91 as the behavior information having no target object as the perceptual information computing unit 9.
5 is output.

The behavior memory 9 configured as described above
1. The object name storage 92 and the object determiner 93 perform, for example, the following processing. For example, when the behavior group number “1” is input, the behavior memory 91 stores the behavior group “
The “Behavior 0”, “Behavior 0”, “Behavior 1”, “Behavior 2”, and “Behavior 3” that constitute “1” are output to the object determiner 93. On the other hand, the object name storage 92 stores “Behavior 0”. "Food" is output in response to "", and "Water" is output in response to "Action 1"
Is output, "no object" is output in response to the action group "2", and "no object" is output in response to "action 3". For example, such an example is a case where the higher-level action is “ingestion action”, and as described above, if the higher-level action is “eat apple”, “ Only "apples" are output. Then, in the object determiner 93, each "" output from the action memory 91 is output.
The "action" and the "object name" output from the object name storage unit 92 are output as a pair of meaningful action information to the perceptual information calculator 95.

The input semantics converter module 59 outputs information relating to perception input to the robot apparatus 1 to the object information storage 94.
The object information storage 94 stores perception information sent from the input semantics converter module 59. That is, for example, the object information storage 9
4 stores parameters for perceptual evaluation used for calculating an appearance tendency, such as “apple”, “distance to apple”, and “direction with apple”, which are target objects.

[0265] The perceptual information arithmetic unit 95 is based on the object information (object information signal) from the object information storage 94 and the action group information (action group information signal) from the object determiner 93. A perception evaluation RM [i] corresponding to each action whose appearance tendency is compared in the action selection calculation unit 82 is acquired. That is, for example, using "distance to an apple", a perceptual evaluation of "eating an apple (eating behavior of an apple)" or a perceptual evaluation of "approaching an apple" is performed.

Then, the perceptual evaluation RM [i] obtained by the perceptual information calculator 95 in this way is used as the action selection calculator 8.
2 is output. For example, the perceptual evaluation RM [i] is shown in FIG.
As shown in FIG. 9, the perceptual information acquisition unit 90
Is output to the action selection calculation unit 82.

Note that a synchronization signal can be output from the object determinator 93 to the object information storage 94. The output of the object determiner 93 and the object information memory 94 are output by the synchronization signal.
Can be synchronized, thereby allowing the perceptual information calculator 95 to input a parameter corresponding to the action from the object determiner 93 at a predetermined timing. .

Basically, the robot apparatus 1 is provided with only one perceptual information acquisition unit 90. However, the perceptual information acquisition unit 90 can be provided for each action. In such a case,
It is only necessary to consider that the perceptual information acquisition unit 90 applies one action to the target object in general, so that it is not necessary to provide the action storage unit 91. For example, such an example is a case where an action selecting unit described later is configured as a plurality of objects.

The procedure of the process performed by the perceptual information acquisition unit 90 will be described with reference to FIG.

First, in step S11, an action group name is obtained. Acquisition of an action group name means
An action group name indicating “approaching an apple”, “smell an apple” or the like, which is a lower action of “eat an apple”, is acquired.

Subsequently, an object selection routine is executed.
In the target object selection routine, in step S12, an action name group to be calculated is obtained. By acquiring the behavior name group, a plurality of behaviors (behavior information in a form applicable to the general target object) are stored in the behavior memory 91. For example, it is information that defines an action name such as “approach” or “smell”.

At step S13, an object name is obtained. With the acquisition of the object name, the object name acquired in the higher-level action is stored in the object name storage 92. For example, it is information on an object such as “apple”.

Thus, in the object selection routine,
Acquisition of an action name group and acquisition of an object name are performed. In the subsequent step S14, it is determined whether or not the perceptual evaluation RM [i] has been calculated in the perceptual information calculator 95 for all the selected actions. If the calculation of the perceptual evaluation RM [i] has been completed for all the selected actions, the process is terminated, and the calculation of the perceptual evaluation RM [i] for all the selected actions has been completed. Is not completed, a perceptual evaluation calculation routine is executed.

The perceptual evaluation calculation routine is executed by the perceptual information computing unit 95 and comprises the following processing.

At step S15, it is determined whether or not the target exists. If the target exists, the process proceeds to step S16. If the target does not exist, the process proceeds to step S18.

In step S16, the perceptual information computing unit 95 acquires the distance and direction (parameter for acquiring perceptual evaluation) of the object from the object information storage 94, and in step S17, perceptual evaluation (Value). ) RM
[I] is calculated. That is, for example, the evaluation RM [i] of “approaching an apple” is calculated from “distance to an apple”. Note that the distance is detected by the distance sensor 22 and the direction is detected using an image captured by the CCD camera 20 or the like.

On the other hand, in step S18, the perceptual information acquisition unit 95 performs the perceptual evaluation (Valu
e) RM [i] is calculated. For example, this process corresponds to the case where the action for perceptual evaluation does not require an object.

The perceptual evaluation calculation routine calculates the perceptual evaluation RM [i] for all the actions to be compared (a plurality of actions constituting the action group) in the above-described determination processing in step S14. It is executed until it is determined that it has been performed. That is, the perception evaluation RM [i] for all the actions in the action group is calculated by the processing in step S14 and the perception evaluation calculation routine.

[0279] If it is determined in step S14 that the perceptual evaluation RM [i] of all actions in the action group has been calculated, the process ends.

The perceptual information acquiring unit 90 is configured as described above, and the perceptual information acquiring unit 90 performs the perceptual evaluation R on a plurality of actions to be compared in the action group.
M [i] can be obtained.

(3-6-3) Processing in Motivation Information Acquisition Unit The motivation information acquisition unit 81 performs processing based on instinct and emotional states that change according to external or internal information (recognition results).
Acquire the motive which is one of the causal factors of the behavior. As shown in FIG. 25, the motivation information acquisition unit 81 has a plurality of instinct / emotional parameters IE [p] (instinct / emotional parameter group), and thereby has a plurality of motivation Mot for actions.
[I] has been obtained. Specifically, the motive of the action is acquired as follows.

The instinct / emotional parameter group IE [p] is composed of information influenced by instinct and emotion, and is specifically composed of a plurality of parameters determined by the internal state model as described above. Have been. In other words, as instinct and emotion parameters, for example,
"Body temperature", "Pain", "Hunger", "Dry", "Affection", "Submission", "Curiosity", "Excretion", "Happiness",
"Sadness,""anger,""surprise,""disgust,""fear,""irritation,""boring,""sleepiness,""socialism,
"Patience", "Tension / Relax", "Vigilance", "Sin",
Examples include “malice”, “sincereness”, “libido”, and “jealousy”.

[0283] The action motivation group Mot [i] is a motivation group corresponding to a plurality of actions in the same action group. For example, "hunger" for "feeding behavior", "throat squirrel" for "watering behavior", and the like.

The motivation information acquisition section 81 maps the instinct / emotional parameter IE [p] to calculate the motive Mot [i] for each action. In particular,
It is calculated by equation (22).

[0285]

(Equation 22)

According to the equation (22), the instinct / emotional parameter IE [p] is multiplied by a coefficient K [i] [p], and
Motivation Mot of each action by mapping as linear sum
[I] is calculated. The motivation Mot [i] calculated as such a determinant is, as shown in FIG. 19, a vector amount from the motivation information acquisition unit 81 to the action selection calculation unit 82.
Is output to

[0287] For example, "search", "need", "rest"
The motivation will be described in detail as an example. The motive Mot [0] of “search”, the motive Mot [1] of “need”, and the motive Mot [2] of “rest” are given by Expression (23).

[0288]

(Equation 23)

Also, K [i] [p] is given as shown in equation (24).

[0290]

(Equation 24)

The instinct / emotional parameter IE [p] is given as shown in equation (25).

[0292]

(Equation 25)

From such a relationship, each of the motives of “search”, “sloppy”, and “rest” is expressed by equation (26).

[0294]

(Equation 26)

In the equation (26), “search” indicates that “fatigue” is a function of instinct / emotional parameters in which “curiosity” acts as a negative factor and “curiosity” acts as a positive factor. Further, “sneaking” indicates that “love and lust” is a function of instinct and emotion parameters acting as a positive factor. Also, “rest” indicates that “fatigue” is a function of instinct and emotion parameters in which “curiosity” acts as a negative factor.

Here, as a first example, consider the case where the instinct / emotional parameter IE [p] is [10, 50, 20]. For example, such a state refers to a state of high curiosity. In such a situation, the "search" M
ot [0] becomes 400 (= −100 + 500 + 0), and “wet” Mot [1] becomes 300 (= 0 + 0 + 30).
0), and the "rest" mot [2] is -150 (= 10
0-250 + 0).

As a second example, instinct and emotion parameter I
Consider the case where E [p] is [70, 10, 30]. For example, such a state refers to a tired state. In such a state, the "search" Moto [0]
Becomes -600 (= -700 + 100 + 0), the "wet" Moto [1] becomes 450 (= 0 + 0 + 450), and the "rest" Moto [2] becomes 650 (= 700-50 +
0).

As a third example, instinct and emotion parameter I
Consider the case where E [p] is [30, 20, 60]. For example, such a state refers to a state in which tiredness is slightly recovered and affection is high. In such a state, the “search” Mot [0] is −100 (= −300 + 2).
00 + 0), and “Needle” Mot [1] is 300
(= 0 + 0 + 300), and the “rest” Moto [2] becomes 200 (= 300−100 + 0).

In this way, from the instinct / emotional parameter group IE [p] and the coefficient K [i] [m], the motive M
ot [i] can be obtained. Then, by appropriately setting the mapping of the instinct / emotion parameter group K [i] [p], a desired motive Mot [i] for acquiring the appearance tendency RM [i] can be obtained. That is, for example, motivations such as “throat throat” and “hunger” in the above example can be obtained.

[0300] The motive information obtaining unit 81 is configured as described above, and the motive Mot [i] for each action can be obtained by the motive information obtaining unit 81. The motivation obtained by the motivation information acquisition unit 81 changes according to a change in parameter values of instinct and emotion.
As a result, it is reflected in the selected action. For example, in the above example, the motive is reflected in the behavior as follows.

The desire is basically a time increase, and therefore increases if it is not satisfied. When the curiosity increases, the robot device 1 starts searching (the first example described above). As we walk around by searching, we get tired as much as we walk. Also, curiosity itself can be reduced by searching. If nothing is input even after walking for a while, the curiosity is reduced and the fatigue is increased, thereby switching to a resting behavior (the second example described above). By taking a rest after a while, fatigue is reduced, and the behavior of the robot apparatus 1 is randomly switched by the affection that has increased with time (the third example described above). Thus, it can be seen that the motivation is reflected in the selected action.

The value of the coefficient K [i] [p] can be set arbitrarily. For example, by setting it arbitrarily, it becomes possible to variously change the mapping based on the instinct and emotion parameter IE [p] for acquiring the motive Mot [i]. Thus, depending on the setting of the coefficient K [i] [p], mapping can be performed in accordance with the type of animal, the growth level, and the like applied to the robot device 1.

The specific configuration of the perceptual information acquiring unit 90 for acquiring the perceptual evaluation RM [i] and the motivation information acquiring unit 81 for acquiring the motivation state Mot [i] has been described. Perceptual information acquisition unit 90 and motivation information acquisition unit 81 as described above
Evaluation RM [i] and motivation state M obtained by
According to ot [i], one action is selected in the action selection calculation unit 82.

[0304] Such an action selection process is executed until an action is selected in the lowest action layer. That is, as shown in FIG. 7 described above, the system for action selection is constructed as a hierarchical structure, and the perceptual perception is performed in each layer until the lowest action (actually output action) is determined. An action selection process based on the evaluation RM [i] and the motive information Mot [i] is executed. That is, as shown in FIG. 6 (B), the “ingestion behavior” is the perception evaluation RM [i] and the motive information M
ot [i], which is a result of the action selection based on the perception evaluation RM [i] and the action based on the motivation information Mot [i] in the mode layer composed of a further embodied action group. The result of the selection, “approaching water” is the result of the action selection based on the perception evaluation RM [i] and the motive information Mot [i] in the module layer composed of the action group that is further embodied. “Go forward” is the result of the action selection based on the perceptual evaluation RM [i] and the motive information Mot [i] in the motor command layer including the action group that has been further embodied. By such a selection process, the “feeding behavior” that is an abstract behavior (behavior as a desire) is realized by an actual behavior such as “moving forward”.

[0305] In selecting an action in each layer,
As described above, the appearance tendency is calculated by causal factors such as perception and motivation, and one action is selected based on the tendency, but the motivation information used when calculating the appearance tendency is calculated in all layers. A uniform one may be used. That is, for example, when “ingestion behavior” is a higher-level action, considering that all lower-level actions are for realizing “intake behavior”, the lower-level action is “ It is an action to satisfy the condition of “hunger (dry)”. Therefore, the lower-level behavior for realizing the “ingestion behavior” is that the state of “hunger (dryness)” is the motive information (causal factor).
Becomes

Note that perception does not always have such a relationship. The perceived information (external intellectual factor) of “approaching water” includes “distance to water”, and the perceived information of “advancing” below “approaching water” includes “direction of water” Is sometimes optimal.

(3-7) Processing in Modulator As described above, the processing for actually appearing one action selected by the action selection calculating unit 82 is performed by the modulator 72 and the output semantics converter module 68 described later. Done.

[0308] The modulator 72 determines an action to be finally caused to appear from the one action selected by the action selecting section 80 and the representative emotion information (representative emotion signal) output from the internal state model section 71.

Here, the representative emotion information output from the internal emotion model unit 71 indicates the current emotional state of the robot device 1. For example, the internal emotion model unit 71
The instinct or emotion having the largest instinct (desire) or emotion parameter value is output as representative emotion information.

The modulator 72 modulates one action selected by the action selecting section 80 based on such representative emotion. That is, by the processing of the modulator 72, the emotion is expressed in the action.

Thus, for example, it is not necessary to cause the current emotion to appear directly as an action of the robot apparatus 1, but this is effective when an action is caused to appear including an emotional expression. That is, for example, when the user is not really angry but is a little angry, the action selected by the action selecting section 80 is accompanied (added) by “displeased”.

[0312] The modulator 72 outputs information on the selected one action modulated by the emotion as described above.
Output to the output semantics converter module 68. For example, the modulator 72 outputs behavior information to the output semantics converter module 68 as an abstract behavior command.

The output semantics converter module 68 gives an output corresponding to the behavior information from the modulator 72 to the signal processing modules 61 to 67. Thereby, the robot apparatus 1 outputs the action determined by the action determination system 70 as an actual action.

The behavior determination system 70 has been described above. By the action determination system 70 as described above, the internal state model unit 71 can change the internal state such as the instinct and the emotional state of the robot apparatus 1 based on the recognition result in the input semantics converter module 59. In addition, the action selection unit 80 can select one action in which the robot apparatus 1 appears from a plurality of actions based on the recognition result.

Then, the internal state obtained by the internal state model unit 71 and the action selecting unit 8 are output from the modulator 72.
On the basis of the one action obtained by 0, action information to which an emotion is added is generated. And modulator 7
According to 2, the action information to which the emotion is added is output to the output semantics converter module 68.

(4) Processing in Output Semantics Converter Module The output semantics converter module 68 holds information such as the type (for example, bipedal or quadrupedal type) and the shape of the robot apparatus 1. The signal processing modules 61 to 67 control the behavior information from the modulator 72 according to the information. For example, in the case of the robot device 1 of the present embodiment that walks on four legs, the output semantics converter module 68 knows that the robot device 1 is a four-legged robot device, and thus “goes forward”. When such behavior information is transmitted from the modulator 72, a command is output to a signal processing module that controls the four legs in order to realize the “forward”. Here, the output semantics converter module 68 sends commands to each of the corresponding signal processing modules 61-67 in response to the abstract action command from the modulator 72.

Each of the signal processing modules 61 to 67 controls a corresponding device based on a command from the output semantics converter module 68. Thus, the action determined (selected) in the action determination system 70 as described above has appeared as an actual action of the robot apparatus 1.

The appearance of such an action of the robot apparatus 1 is performed under the control of the posture and the action. Although each part of the robot apparatus 1 can operate independently in principle, it is prohibited to independently perform a predetermined operation by managing the posture and the operation.

[0319] As shown in FIG.
For the body unit 2, the leg units 3A to 3D,
The head unit 4 and the tail unit 5 are connected to each other. Thereby, basically, the robot apparatus 1
Is configured such that each unit can operate independently by individually controlling each signal processing module according to the selected action. However, in some cases, the operation may be inappropriate due to interference between the operations of the units. Also, the transition to the desired posture or motion is
There are times when it is impossible from the current attitude.

Therefore, the robot apparatus 1 manages the posture and the operation by matching the units so as to prevent an unreasonable posture and interference between parts (units). The robot apparatus 1 manages such postures and movements by a posture management module 61 shown in FIG.

Specifically, the posture management module 61
When a command such as “walk forward” is received while sitting, the robot apparatus 1 searches for a posture transition path for transition from the “sitting state” to the “walking state”. For example, a transition path from a “sitting state” to a “walking state” is searched for via a plurality of postures and actions. Then, based on the result of the search for the transition path from the “sitting state” to the “walking state”, in order to execute the posture and motion on such a transition path, an instruction is issued according to the order of the path to be transitioned. Output to the signal processing module. Thereby, the robot apparatus 1 can prevent an unreasonable posture and interference between parts and reach a desired desired posture and movement, that is, the behavior determined by the behavior determination system 70 described above. .

The configuration of the robot device 1 and its processing have been described above. With the above-described configuration, the robot apparatus 1 can cause the behavior determined by the ethological approach to appear. As a result, the biological feeling of the robot device 1 further increases, and the user (owner) feels a closer feeling and a more satisfying feeling with the robot device 1.

(4) Other Embodiments The best actual embodiment of the robot apparatus 1 has been described above. However, the present invention can be realized as another embodiment described below.

[0324] In the above embodiment, the action determination system 70 determines the final action selection with reference to the motivation information. For example, in the example shown in FIG. 6B, “forward” is selected with reference to the motive information. However, the action selection at the last stage may be determined excluding the motivation information.

That is, for example, as shown in (A) and (B) of FIG. 26, “approach water”, which is a lower action (action) of “ingestion action”, and further, an action (operation) of the lower order Is selected with reference to information excluding the motive information, for example, perceptual information such as the distance to the object. For example, when trying to perform a certain action (for example, a vague action), the motive greatly affects the action selection, and after narrowing down the desired action to some extent, the action selection processing is separated from the motive. It is like (selective thinking) switching to one to achieve such behavior. That is, the point that the operation finally selected is determined without being influenced by the motive. Then, when finally determining such an operation, for example, perceptual information is used. In addition, the mode layer has 0 or more layers.
The module layer may be defined as one layer.

For example, as shown in FIGS. 27 and 28, an operation generation unit 100 is provided as a part for determining an operation without being based on the above-mentioned motive information. The action generation unit 100 selects, from the action selected by the action selection calculation unit, an action that implements an action of “approaching water” or “downward” as an action based on perceptual information or the like. . Then, the operation generation unit 100 outputs the selected operation to the modulator 72. As described above, the modulator 72 outputs a behavior modulated by the emotion sent from the internal state model unit 71.

Specifically, the operation generation unit 100
As shown in the figure, the perceptual information acquisition unit 90, the action selection calculation unit 1
02 is provided. For example, using the target information storage unit 94 in which various types of information from the output semantics converter module 68 are stored, the action selection calculation unit 102
Select the operation in. Thereby, the operation generation unit 1
00 is, for example, in the case of “moving forward”, the distance (for example, information in which the distance to the object is 10 cm) and the direction (for example, information in which the distance to the object is 10 cm) which is the information stored in the object information storage 94. The action selection operation unit 102 selects an action using only the information of (the information indicating that the object is in the right direction).

In the above-described embodiment, the case where one action is selected from a plurality of actions by the action selecting section 80 has been described. For example, the action selecting unit 80 holds information on a plurality of actions, and determines one action based on the data of these actions. However, it is not limited to this.

For example, the action determination system 70 can design a part for determining an action in an object-oriented manner. It should be noted that even when a system that determines an action as object-oriented is constructed, a hierarchical structure such as a higher-order action and a lower-order action is used as it is. The action is selected such that one action is selected for each object from an action group configured as an object. More specifically, as shown in FIG. 30, in the behavior system, a behavior is selected by a plurality of behavior selecting units (objects or threads) 80 ₁ , 80 ₂ , 8.
0 ₃ is provided as a hierarchical structure.

In this example, as shown in FIG. 30, a system is constructed in which the action selecting section, which is an object, has two layers, but it goes without saying that the present invention is not limited to this.

[0331] each action selection unit _{80 1,} 80 2, 80 _3,
Like the above-described action selection unit 80 provided solely by the action determination system 70, the action determination system 70 includes a perceptual information acquisition unit 90, a motive information acquisition unit 81, and an action selection calculation unit 82.

[0332] With such a configuration, based on the action the action selecting section 80 ₁ of the upper selects, for selecting a behavior in the lower action selector section 80 _2, 80 _3. here,
The selection of the higher-level action is the selection of the lower-level one-action selection section. Then, the selected one action selecting unit selects a further lower action.

Then, the plurality of action selecting units 80
_1, 80 _2, 80 action selection unit which is located at the bottom of the system consisting of _3, the information of the selected action, passes the motion generating unit 100 as described above.

By constructing such a system for determining an action as object-oriented, it is not necessary to always process or grasp the entire system for selecting an action, and the processing load of the action selection is reduced. Become like Further, even when a new action is added later, it is only necessary to add such a new object, and it is not necessary to rewrite all data for selecting an action. For example, a case where a new action is added means that a new action is acquired by learning,
This refers to the case where a new action is added due to a change in the growth level.

Also, FIG. 6 (B) or FIG. 26 (B)
Subsystem layer (SUBSYSTEM) and mode layer (MOD
E1, MODE2) and specific examples of the configuration of each action group in the module layer (MODULE) are as shown in FIGS.

[0336]

The robot apparatus detects external or internal information by the detecting means, and obtains, by the causal factor obtaining means, the external or internal information detected by the detecting means, which is the information affecting the behavior. Then, based on the causal factor acquired by the causal factor acquiring means, the appearance tendency of the action affected by the causal factor is acquired by the appearance tendency acquiring means, and the appearance tendency acquiring means corresponds to the two or more behaviors acquired by the appearance tendency acquiring means. The appearance tendency, which is the same group, is compared by the appearance tendency comparison means, and one action is selected by the action selection means based on the comparison result of the appearance tendency by the appearance tendency comparison means. Based on the action you choose,
By controlling the operation unit by the operation unit control unit and causing the selected action to appear, by comparing the appearance tendency determined by being influenced by the causal factor, one action is selected, The behavior emerges as a behavioral approach.

In such a method for determining the behavior of a robot device, the external or internal information of the robot device is detected by a detecting means in an information detecting step, and the behavior of the robot device of the external or internal information detected in the information detecting step is detected. In the causal factor acquisition step, the causal factors that are considered to be the influential information are acquired, and based on the causal factors acquired in the causal factor acquisition step, the appearance tendency of the behavior affected by the causal factor is determined in the appearance tendency acquisition step The appearance tendency corresponding to two or more actions acquired in the appearance tendency acquisition step, and the appearance tendency in the same group is compared in the appearance tendency comparison step. Based on the comparison result, one action is selected in the action selection step, and based on the action selected in the action selection step, the operation unit of the robot device is controlled in the operation unit control step, By causing the selected behavior to appear in the robot apparatus, the robot apparatus selects one behavior by comparing the appearance trends determined by the causal factors, and the ethological approach As a behavior comes to appear.

[Brief description of the drawings]

FIG. 1 is a perspective view illustrating an external configuration of a robot device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a circuit configuration of the robot device described above.

FIG. 3 is a block diagram showing a software structure of the robot device described above.

FIG. 4 shows the software of the above-mentioned robot device.
3 is a block diagram illustrating a configuration of a wear layer.

FIG. 5 is a block diagram illustrating a configuration of an upper behavior determination system.

FIG. 6 is a diagram used to explain a robot device that determines an action by the ethological approach.

FIG. 7 is a diagram showing an action selection system constructed in the action selection unit and having a plurality of actions in a hierarchical structure.

FIG. 8 is a first half diagram used to explain the function of each unit constituting the above-described action determination system.

FIG. 9 is a second half diagram used to explain the function of each unit constituting the above-described action determination system.

FIG. 10 is a block diagram used to describe a case where behavior is determined by perception and motivation by an ethological approach.

FIG. 11 is a characteristic diagram showing a causal factor state space constituted by causal factors and an appearance tendency space in which an appearance tendency of an action defined by the causal factor state space is mapped.

FIG. 12 is a diagram used to explain a causal factor state space.

FIG. 13 is a characteristic diagram showing a causal factor state space of eating behavior and water consumption behavior, which is used for describing an ethological approach as a mathematical expression.

FIG. 14 is a characteristic diagram showing an appearance tendency space composed of an eating tendency and an eating tendency used in the description of formulating the ethological approach.

FIG. 15 shows a case where the value mapped from the causal factor state space is in the eating behavior selection region ((A) in the figure) and a case where the value mapped from the causal factor state space is in the watering behavior selection region ( (B) in the figure.

FIG. 16 is a diagram used for explaining exclusive control.

FIG. 17 is a characteristic diagram showing an appearance tendency space used for explaining selection of an action realized by a calculation formula based on an ethological approach.

FIG. 18 is a characteristic diagram illustrating an appearance tendency space used for describing selection of an action by an actual process of the robot apparatus.

FIG. 19 is a block diagram illustrating a flow of information among a perceptual information acquisition unit, a motivation information acquisition unit, and a behavior information selection unit that constitute the behavior selection unit.

FIG. 20 is a diagram used to explain calculation of an appearance tendency before exclusive control.

FIG. 21 is a diagram used to explain calculation of an appearance tendency by exclusive control.

FIG. 22 is a flowchart illustrating a procedure for calculating an appearance tendency.

FIG. 23 is a block diagram illustrating a configuration of a perceptual information acquisition unit.

FIG. 24 is a flowchart showing a processing procedure in the above-mentioned perceptual information acquisition unit.

FIG. 25 is a diagram used to explain a motivation information acquisition unit.

FIG. 26 is a view showing another embodiment of the robot apparatus, which is used to describe a case where the action selection in the lower layer is not influenced by the motive in the action selection system having a hierarchical structure.

FIG. 27 is a diagram used to explain functions of a behavior selection calculation unit and a motion generation unit of the robot device according to the other embodiment described above.

FIG. 28 is a block diagram showing a configuration of an action determination system for a robot device according to another actual embodiment described above.

FIG. 29 is a block diagram illustrating a configuration of an operation generation unit.

FIG. 30 is a block diagram illustrating a plurality of action selection units that are set as objects.

FIG. 31 is a diagram illustrating a specific example of the first half of each action group configuration in a subsystem layer, a mode layer, and a module layer.

FIG. 32 is a diagram illustrating a specific example of the latter half of each action group configuration in the subsystem layer, the mode layer, and the module layer.

[Description of Signs] 1 robot apparatus, 70 action determination system, 71 internal state model, 72 modulator, 80 action selection unit,
81 motivation information acquisition section, 82 action selection section, 90 perception information acquisition section

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Rika Hasegawa 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Kotaro Sabe 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo No. Sonny Corporation (72) Inventor Ronald Cleg Arkin 30332-0280 Atlanta Georgia Institute of Technology, Georgia USA

Claims (20)

[Claims] 1. A robot device for controlling a motion unit to cause an action to appear, wherein the detection means detects external or internal information, and the external or internal information detected by the detection means affects an action. A causal factor acquiring means for acquiring causal factors to be information, based on the causal factors acquired by the causal factor acquiring means,
An appearance tendency acquisition means for acquiring an appearance tendency of an action affected by the causal factor; and an appearance tendency corresponding to two or more actions acquired by the appearance tendency acquisition means, wherein An appearance tendency comparing means to be compared; an action selecting means for selecting one action based on the result of the appearance tendency comparison by the appearance tendency comparing means; and an operation section controlled based on the action selected by the action selecting means. And an operation unit control means for causing the selected action to appear, wherein the appearance tendency of the one action selected by the action selection means changes according to a causal factor which changes due to the actual appearance of the action. A robot device characterized by the above-mentioned. 2. The robot apparatus according to claim 1, wherein the causal factor acquiring means acquires at least a causal factor related to perception and a causal factor related to motive. 3. The method according to claim 2, wherein the causal factor acquiring means acquires at least an instinct element or an emotional element as the causal factor relating to the motive.
The robot device as described. 4. The instinct element comprises: fatigue,
Fever or body temperature, pain, appetite or hunger, thirst, affection
ction), curiosity, elimination
Or at least one of sexuality, wherein the emotional element is happiness, sadness
s), anger, surprise, disgus
t), fear, frustration, boring
redom), sleep (somnolence), sociability (gregariousnes)
s), patience, tension, relaxed, alertness, guilt, spite, spy, loyalty, submission
4. The robot apparatus according to claim 3, wherein the robot apparatus is at least one of jealousy. 5. An action selection system, wherein a plurality of actions that can appear are constructed as a hierarchical structure, and a plurality of lower actions in the same group indicate specific actions of an upper action. The tendency comparing means compares the appearance tendency of a plurality of lower-level actions in the group corresponding to the higher-order behavior, and the action selecting means determines one based on the comparison result of the appearance tendency by the appearance tendency comparing means. The action unit control means controls the operation unit based on the lowest action when the action selected by the action selection means is the lowest action. The robot device according to claim 1, wherein 6. The causal factor acquiring means acquires a causal factor relating to perception and a causal factor relating to motivation, and the appearance tendency acquiring means, at least for the behavior of the lowest layer, is based on the causal factor relating to perception. The robot apparatus according to claim 5, wherein the tendency of appearance of the behavior of the lowest layer is obtained. 7. It has a plurality of objects for action selection corresponding to each action, and the causal factor acquisition means, the appearance tendency acquisition means, and the action selection means are realized by each object. The robot apparatus according to claim 1, wherein 8. The robot apparatus according to claim 1, wherein the appearance tendency comparison means performs an exclusion process between the appearance tendencies of the actions to be compared, and compares a plurality of appearance tendencies. 9. The robot apparatus according to claim 1, wherein said detecting means is a sensor. 10. The causal factor acquiring means, based on sensor information that is external or internal information detected by the sensor,
The robot apparatus according to claim 9, wherein a causal factor for evaluating the behavior is acquired. 11. An information detecting step of detecting external or internal information of the robot apparatus by a detecting means, and a causal factor which is information which affects the behavior of the robot apparatus of the external or internal information detected in the information detecting step. A causal factor acquiring step to acquire, an appearance tendency acquiring step of acquiring an appearance tendency of an action affected by the causal factor based on the causal factor acquired in the causal factor acquiring step, An appearance tendency comparison step of comparing the appearance trends corresponding to the two or more acquired behaviors, wherein the appearance tendency is compared with the same group; Selecting an action, and controlling the operating unit of the robot apparatus based on the action selected in the action selecting step, and And an operation unit control step of causing the action to appear, wherein an appearance tendency of the one action selected in the action selection step changes according to a causal factor that changes due to the actual appearance of the action. A method for determining the behavior of a robot device. 12. The method according to claim 11, wherein the causal factor acquiring step acquires at least a causal factor related to perception and a causal factor related to motive. 13. The method according to claim 12, wherein the causal factor acquiring step acquires at least one of an instinct element or an emotional element as the causal factor relating to the motive. 14. The instinct element may include fatigue
e) heat or body temperature, pain
n), appetite or hunger, thirst, affection, curiosity, excretion
ation) or at least one of sexuality, wherein the emotional element is happiness, sadness, anger, surprise, disgust, fear. ), Frustration,
Boredom, sleep (somnolence), sociability (grega)
riousness, patience, tension, relaxed, alertness, guil
t), spite, loyalty, obedience (sub)
14. The method according to claim 13, wherein the action is at least one of a mission and a jealousy. 15. The robot apparatus according to claim 1, wherein information of a plurality of actions that can appear is constructed in a hierarchical structure, and a plurality of lower actions in the same group indicate a specific action of the upper action. In the appearance tendency acquiring step, the appearance tendency is compared for a plurality of lower behaviors in the group corresponding to the upper behavior, and in the behavior selection step, the comparison result of the appearance tendency of the lower behavior is provided. In the operation unit control step, when the action selected in the action selection step is the lowest action, the robot based on the lowest action is selected. The method for determining an action of a robot device according to claim 11, wherein the operation unit of the device is controlled. 16. The causal factor acquiring step acquires a causal factor relating to perception and a causal factor relating to motivation. In the appearance tendency acquiring step, at least the behavior of the lowest layer is determined based on the causal factor relating to perception. 16. The method for determining an action of a robot device according to claim 15, wherein the appearance tendency of the action of the lowest layer is acquired. 17. The robot apparatus according to each of the actions,
The robot apparatus according to claim 11, further comprising a plurality of objects for selecting an action, wherein the causal factor obtaining step, the appearance tendency obtaining step, and the action selecting step are executed by each object. Action decision method. 18. The method according to claim 11, wherein in the appearance tendency comparison step, a plurality of appearance tendencies are compared by performing an exclusion process between the appearance tendencies of the actions to be compared. . 19. The method according to claim 11, wherein said detecting means is a sensor. 20. The robot according to claim 19, wherein in the causal factor acquiring step, a causal factor for evaluating behavior is acquired from sensor information detected as external or internal information detected by the sensor. How to determine the behavior of the device.